FilmSpark.AI Tutorial Guide

Welcome to the FilmSpark.AI Tutorial Guide. Click any section below to get started.

1. 🚀Getting StartedCreate your account, learn the interface, and understand how credits work.→
2. 📽️Scripted ProjectThe core FilmSpark workflow — script to storyboard to finished video, in ten steps.→
3. 🖼️Editing ImagesEdit, upscale, reframe, remove backgrounds, build 3D worlds and browse history for any image in FilmSpark.→
4. 📰Ad TemplatesDrop a product image into a pre-made narrative and get a video ad in three steps.→
5. 🎬Scene ComposerBuild scenes and explore locations from multiple angles without writing a script.→
6. 🎨Asset StudioGenerate standalone images and videos with any model, then bring them into your projects.→
7. 🤖The FilmSpark AgentA director's assistant inside your project. It reads your whole film, plans the work, and asks before it touches a frame.→
8. 💪Team WorkspacesPooled credits, shared billing, role-based access and real-time collaboration for production teams.→
9. 🖼️My LibraryYour central store of images, videos and audio across every project.→
10. 🎬First & Last FrameControl where a shot starts and ends by supplying both frames, and extend shots indefinitely with frame extraction.→
11. ⚖️Compare ModeGenerate the same shot or asset across multiple models at once and pick the winner side by side.→
12. 🎨Color Grading & LUTsApply a consistent film look across your project with built-in or custom LUTs over the Cineon Log style.→
13. ⏫Video UpscalersTurn 720p or 1080p clips into polished 4K with Topaz and Crystal, and know which model to pick.→
14. 📚Prompt LibrarySave prompts that worked and reuse them across shots, assets, projects and teammates.→
15. 💳Credits & SubscriptionsCredit packs, subscription plans, and how to manage or cancel your subscription.→
16. 💡TipsPractical reminders for better results, fewer wasted credits and stronger prompts.→
17. ❔FAQQuick answers to the questions we hear most.→

FilmSpark.AI is currently a desktop web app. It is not yet optimized for mobile devices.

Getting Started

Create your account, learn the interface, and understand how credits work.

Introduction#

Welcome to FilmSpark.AI! This guide will walk you through the simple steps to transform your ideas into compelling visual narratives – from concepts and scripts to storyboards, actor and location visuals, AI-generated video clips, and finally, assembling and exporting your final movie.

Please note that FilmSpark.AI is currently a desktop web app. It is not currently optimized for mobile devices.

Let's get started!

Account Creation#

The first thing you will need to do is create an account. Just click on the SIGN IN button in the top right corner of the screen. Then, you have the option of continuing with a Google account, or manually signing up with an email address and password. We bet you have done this before!

Site Navigation#

• Explore - This is the central hub of FilmSpark.AI. Here you can find our Tutorial Guide & Videos, as well as our Gallery.
• My Library - This is a collection of your Projects, Scenes, Images, Actors and other assets you have uploaded or created on FilmSpark.AI. Everything you create is automatically saved here.
• Scripted Project - This is the starting point for bringing your ideas to life! Clicking this will start you on the process of creating your new story project.
• Create Ad - A quick access point to create an ad for your product.
• Create Micro Drama - Create your own micro drama with the Hook-Point-Payoff-Hook cycle.
• Create Short Film - Bring your ideas for a short film to life.
• Scene Composer - Create individual scenes and shots using reference images and actors from My Library without using a script.
• Asset Studio - Generate images with your choice of image model. It's a quick and easy process!
• The FilmSpark Agent - Pro and Ultimate subscribers can use the in-project AI Agent to build and edit shots, assets and scenes from plain-language instructions. See the FilmSpark Agent guide.

Project Settings#

Once you have started your project, the Settings button can be found in the upper right of the UI. The button image is a Gear icon. The Project Settings modify many of the initial setup options for an existing project.

Format & Type#

• Story Type: Options include Custom, Short Film and Commercial/Ad.
• Output Target: Available choices include YouTube, Instagram, TikTok and more.
• Duration: Change the intended length of the story.

Creative Direction#

• Story Structure: Options include AI-Determined, Acts and Episodic.
• Genre & Tone: Genre will define the overall style or category of your script. Tone will influence the mood and emotional impact of your script.

Visuals & Models#

• Art Style: You can change your Art Style whenever you like. Note: You will have to regenerate any image prompts for this to take effect.
• Aspect Ratio: You can switch your aspect ratio, but will need to clear out all storyboard images and regenerate all shots.
• Default Image/Video/Lipsync Generation Model: You can globally specify your preferred Image, Video and LipSync generators here. For the full list of available models, see the Model Guide.
• Story/Prompt Suggestion LLM Model (Pro and Ultimate only) - Choose from a variety of different LLM Models to refine how your story's text and prompts are generated.
• You can also Transfer Ownership of your project to a friend. Note: you will not be able to work on it until ownership is transferred back to you.

Credits#

Credits are used to generate content, such as prompts, videos and images in the stories you create on FilmSpark.AI. Users are given a modest amount of credits after they first sign up and verify their accounts. Once out of credits, a user must purchase either Credit Packs or Subscriptions to continue using the app via the "Upgrade" button in the top right of the app.

See the Credits & Subscriptions section for information on how to purchase Credits and Subscriptions.

Scripted Project

The core FilmSpark workflow — script to storyboard to finished video, in ten steps.

Step 1 - Create New Project#

Scripted Projects let you upload or write a script, automatically generate a visual storyboard, and turn it into a final video. This is FilmSpark's core workflow for streamlining storytelling with precise creative control.

• Click on the Scripted Project button.
• Choose your Concept Source. You can choose from Develop Concept with AI or Upload Script or Brief. If you already have a script, you can upload it here to get started. (PDF and .txt files supported. Full screenplays are now supported — you can upload scripts of 100–200 pages). Note: When uploading a script many options below will vanish as they will be derived from your script. If you did, go ahead and skip directly to Step 6 - Create Your Assets, otherwise proceed down this list.
• Choose your Story Type. All options can be overridden.
  • Custom will let you determine the story type as the creation process unfolds.
  • Short Film is pre-selected when choosing Scripted Projects.
  • Commercial/Ad will bring up additional settings to add Ad Campaign Details. This is the same as "Create an Ad".
  • Micro Drama will craft an episodic, vertical 9:16 format series in the traditional Hook-Point-Payoff-Hook cycle.
• Style - You can upload an image to create your own style tag that will be used throughout your project, or you can choose from our pre-made styles including Realistic (16mm, 1970s Film, 1980s TV, Cyberpunk Neon-Noir, Documentary, Golden Hour), Shonen Anime, Claymation, Modern Comic, and more.
• Output Target - Selecting an output target other than "General" will automatically change the Aspect Ratio and Duration for the specified target. Available choices include YouTube, Instagram, TikTok and more.
• Choose your Format by first selecting your primary Aspect Ratio: 16:9 (Widescreen), 21:9 (Cinemascope), 4:3 (Horizontal), 3:2 (Widescreen), 9:16 (Vertical), 2:3 (Vertical), 3:4 (Vertical) or 1:1 (Square). Then, choose the Duration for the story. Note: not all image and video models support 21:9.
• Default Models - These allow you to pre-select the default video, image and lip sync models. This can be changed at any time in the Settings or for each individual shot via a pulldown menu. Pro and Ultimate subscribers can also select the default LLM at project creation, and the Create screen now remembers your last-used default models for new projects.
• Story Structure - You can choose from AI Determined to let the AI decide what will work best for the project you create, Acts or Episodic.
• Genre and Tone - Genre will define the overall style or category of your script. Tone will influence the mood and emotional impact of your script.

Step 2 - Concept#

Craft Your Story with the Co-Pilot

If you did not upload a script, you now will work with the FilmSpark.AI Co-Pilot "Sparky" to craft your story and create a Synopsis, Scene Breakdown and then Script.

• Describe Your Idea: In the "Concept" section, work with Sparky and type out your initial story idea or concept. It can be as simple or complex as you like.
• Go With The Spark: Next Sparky will lead you through a series of questions to craft your actors, setting, plot points or product features for an Ad that you want to highlight. Pick from Sparky's suggestions, combine them, or add your own idea.

Step 3 - Synopsis#

When you're ready and happy with your answers, click (Generate) Synopsis in the top right corner. Sparky will create a concise summary of your core concept under the Synopsis tab. Read it carefully, make any edits if needed or re-gen it if you like. Small changes here can make a big difference in your story. You can also go back to edit your initial Concept. Note: If you do this, you will need to regenerate the Synopsis to see the updates.

Once satisfied with the Synopsis, click (Generate) Scenes.

Step 4 - Scene Breakdown#

Your story is taking shape! Based on your Synopsis, Sparky has generated a Scene-by-Scene Breakdown of your story. The Story Type you selected will determine the structure (3 Acts, 1 Scene, 60 1-minute episodes).

• Review & Edit Scene-by-Scene Breakdown: Read through the scene-by-scene outline. You can make edits, add a new scene, or delete a scene.

Step 5 - Script#

Generate Full Script: When you're happy with the Scene-by-Scene Breakdown, click (Generate) Full Script in the top right corner. The Co-Pilot will write dialogue, action lines, and descriptions for each scene. You can then review and edit the generated script within our editor. Use the fullscreen button to expand the script.

Once your script feels solid, click (Generate) Assets.

Step 6 - Create Your Assets#

Next, based on your script, FilmSpark will identify the key actors and locations and create descriptions and prompts for them. You will also have the option of generating images for every Actor and Location at once, if you choose to do so. Other options include:

Import Assets - Import the Actors, Locations and Props from an existing project into this one, using the Import Assets option in the Assets UI.

Duplicate Actor - Duplicates an Actor along with its image and details. To duplicate an Actor, click on the button in the top left corner of the Actor image while in the main Actors menu.

Quick Gen an Actor Image - Quickly generate an image for an Actor without having to leave the main Actors menu. To quickly generate an Actor image, click on the button in the bottom left corner of the Actor image.

Finally, if you would like to see a full screen view of an Actor, click on the Expand View button in the top right corner of the Actor image.

You can also take a closer look at the details for each Actor and Location by clicking on them to enter the Editing section.

Generating Actor Images#

For each actor, you have two powerful options to create their look:

• Generate from Story: FilmSpark will suggest an image prompt based on the actor's details and your chosen project style. You can edit this prompt for more control, or edit the actor details and the Image Prompt will be updated automatically. Tip: Start with the AI's suggestion, then tweak it!
• Create from Photo: You can use a photo of yourself, or your friends, etc. and bring them into your story (as long as you have permission). Upload a reference photo. Sparky will generate an image prompt to transform it into your chosen style.

Keep Photo Outfit: Selecting this will keep the outfit the person in the photo is wearing.

Actor Description#

When clicking on an Actor, you will see the description of that Actor on the left side of the screen. This includes the Actor's name, what they are wearing, their age and other details. These details will be automatically filled in by FilmSpark based on your script. You can create new Actors, or edit any of the details below, which are automatically carried over to your pre-written prompt by FilmSpark.

• Species, Sex, Body Type, Height, Hair Color, Eye Color, Age Category, Ethnicity
• Background / Bio, Outfit / Clothes, Key Features

Image References - Reference images can be added to help create the look of your Actor. To upload a reference image or add an image from My Library, click on the Add icon.

Prompt Tab#

• Image Prompt - The details from your Actor Description are automatically used in the prompt as long as Auto Prompt is checked. Then, any changes made in the Actor Description are automatically carried over to the prompt. The prompt is the final copy sent to the AI Model to create your image.
• Auto Prompt - Any changes you make in your Actor Description are automatically carried over to your prompt if the "Auto Prompt" button is checked.
• Changing Style - If you wish to change the style of an existing project you do so in the Project Settings.

Voice#

You can specify a voice for each actor. Just click on the Voice tab to see and hear the options. Press the play button to hear a sample.

Voice Cloning: This feature allows you to create your own voices simply by uploading a sample of the voice you want to generate.

• Kling Voice: Used with Kling v3.0 and Kling v2.6 Pro video models for in-video dialogue.
• Voice Import (Eleven Labs): For use with Eleven Labs Voice Generation Models only. Enter a Voice ID for an Eleven Labs voice.

Please note that private ElevenLabs voices will not work. Only public or shared voice IDs are supported.

Image Generation Model#

This is where you can select from a variety of model options for image generation.

• Number of Images: Choose from 1, 2 or 4 images to generate.
• Adv. Settings: Some image models have advanced settings available.
• Generate Image: Once your text prompt is ready, click the (Generate) Actor button.

Once you have generated Actor images, you will have the option to Edit, Create Turnarounds, Upscale, Remove Background, Create a 3D World from the image, and Adjust (Re-Crop).

Editing Actor Images#

Note: The various features listed below have been captured in greater detail in the Editing Images section.

Actor Turnarounds: Generate images showing different angles of an Actor.

Expression Sheets: Generate a one-sheet of an Actor's expressions, with new emoji expression icons to choose from.

Auto-Use Turnaround Sheets in Shots: When generating Turnaround Sheets, a pop-up lets you automatically use those angles in your shots for stronger character consistency.

Upscale Actor Images:

Remove Background:

3D (best used for Location images):

Adjust Actor Images:

Generation History:

Locations#

FilmSpark identifies key locations automatically. Edit prompts and descriptions, generate from reference photos, or upload your own images.

Props#

Add specific props. Click Create Prop, add description, generate prompt and image. New description fields are available for Locations and Props, along with Generate Details / Prop Details buttons that auto-generate descriptive details. Location descriptions also update when you use a reference image.

Miscellaneous#

Similar to Props — generate images for any miscellaneous items via Create Misc.

Once happy with your Assets, click (Generate) Storyboards.

Step 7 - Storyboards & Video#

The Storyboard is where your story comes to life and you will likely spend the most time in. FilmSpark.AI generates scene descriptions, storyboard images and video clips for your entire Story.

Please be mindful of generating videos and the associated credit cost.

Tip: Run a shot across multiple models at once — see Compare Mode. The FilmSpark Agent can also read your current selection and build or edit shots for you — see The FilmSpark Agent.

Overview#

When you first enter the Storyboard section, your Scenes and Shot Descriptions will be pre-generated based on your script. For maximum control, we recommend allowing FilmSpark to generate a first pass of the images, then go in as needed and re-gen or edit them.

Grid View and List View: Switch between layouts using icons in the upper left corner.

• Scene Sidebar: Navigate quickly to specific scenes from the sidebar in the Storyboard, Scenes Script and Shot List views.
• Edit Shot/Scene Order: Click the reorder icon next to Preview (upper-left) to re-order shots or scenes — useful for longer stories.
• Display Toggles: New toggles to the left of the size slider hide or show dialogue and shot details in the Storyboard.
• Edit Scene Titles & Script Links: Edit a scene or episode title directly in the Storyboard, and open the script for each scene via its Script link.

Anchor Shots: When creating a Shot List, you can choose to generate only the Anchor Shots first.

Scene and Story Quick Generate Buttons:

• (Generate) Full Story Descriptions / Full Story Images
• (Generate) Scene Descriptions / Scene Images

Duplicate Shot:

Quick Gen Image:

Quick Gen Video (after image exists):

Generating Storyboard Images#

Click any shot to enter the Generate Shot view.

Details:

• Image Shot Description, Choose Assets, Dialogue, Image Shot Details (Shot Type, Focal Length, Lighting, Perspective)
• Staging & Composition, Action & Performance, Set & Lighting, Decisive Moment, Atmosphere, Color Palette, Technical Finish

Disconnect from Story:

Generate Image Shot Details Button:

Prompt Tab:

Image Prompt is auto-generated when Auto Prompt is checked.

Scene Context:

Image Models:

• Number of Images (1, 2, or 4), Adv. Settings, Shot Notes
• Continue Shot: Create a direct continuation of the current shot, or a new next-shot candidate that uses the current shot as a reference for consistent lighting and staging.
• Note: When generating shots, the AI now sets a specific duration for each shot automatically.
• Click (Generate) Shot to create the image.

Editing Shot Images#

See the Editing Images section for full details. Features include Edit, Specialized Models (Luma Reframe, Qwen Multiple Angles), Upscale, Remove Background, 3D, Adjust, and Generation History.

Generate Video#

• Click the Video button above the Shot Image.
• Video Shot Description, Disconnect from Story, Choose Actors for Dialogue, Camera Movement, Video Action, Video Performance

Prompt Tab:

Video Models:

• Adv. Settings, Shot Notes
• Click (Generate) Video and wait for generation.
• Multi-Model Video: Generate multiple videos for a single shot using different video models, then compare them in Compare Mode.

Editing Videos#

Edit, Upscale, Reframe, and Generation History available for all generated videos.

Storyboard Preview Modes#

• Mouse-over shots to preview videos from the main storyboard
• Preview Mode — click Open Preview in upper left
• Shot Names/Descriptions, Auto-Play toggle, Edit button, Timeline navigation

Step 8 - Adding Sound#

Lip Sync Via Video Models#

There are multiple ways to create dialogue and lip sync with your characters in AI and with FilmSpark.

• Image + Prompt — Models like Kling 3.0, Veo 3.1 create dialogue and sound automatically. Best visual lip sync but voice may vary shot to shot. Use Voice Change to maintain consistency.
• Image + Audio File — Models like Kling Avatar Pro and OmniHuman. Upload or create a voiceover clip for longer clips.

SFX/Voice Tab#

Lip Sync (Video + Audio = Video):

• Select Model, verify Actor assignment, ensure Actor has a voice assigned
• Keep original video audio option
• LipSync Modes (v2/v2 Pro): Cut Off, Loop, Bounce, Silence, Remap
• Upload voice files (.wav/.mp3) or click Generate Voice Line. You can now generate multiple voice lines at once, and a history of uploaded and generated voice lines is kept.
• Click (Generate) Lip Sync

Creating Voiceover:

• Enter Assets and create a "Narrator" actor (add a voice, no image needed)
• In Storyboards, assign this actor for dialogue on any shot
• Generate image and video for the shot
• In SFX/Voice tab, click Voiceover, Generate Voice Line, then generate

Dubbing — Change spoken dialogue to another language.

Voice Change#

Replace an actor's voice with an Eleven Labs voice matching the original cadence and performance.

Sound FX#

• In the Storyboards - SFX/Voice tab: Generate sound effects for any shot that has a video.
• In the Video Editor: Generate both sound effects and music not tied to any one shot.

Models: Pixverse, Mirelo SFX v1. Options: Keep Original Video Audio, Sound Effect Prompt, Duration.

Step 9 - Video Editor#

Make basic trims and edits, add Sound Effects or Music, then export.

Timeline#

Storyboard images and videos are sequentially laid out. Trim clips to refine pacing.

Assets Section#

• Add/Delete Audio Tracks, Upload images/videos/audio, Generate SFX and Music
• Volume controls per clip and per track

SFX and Music in the Editor#

• Generating Sound Effects (Eleven Labs): Description, Duration, Loop, Prompt Influence, Output Format, Quantity (1-4)

• Generating Music (Minimax Music, plus Lyria 3 and Lyria 3 Pro): Prompt, Lyrics, Tags, Prompt Strength, Balance, Output, Bit Rate, BPM, Seed, Quantity (1-4)

Once finished editing, click Export.

Step 10 - Export & Publish#

Export#

• XML Editor — Export timeline as .XML for Adobe Premiere, Final Cut Pro, etc., with selectable resolution and framerate.
• Individual Files — Download .zip of source assets
• Export Assets — Export all active Actors, Locations and Props in the project.
• Full Movie — Render all scenes into a single video file
• Full Story — Export as .CSV (including a new Asset & Shot History CSV), PDF, or PDF with images

Publish#

Movie Poster:

• Generate a 2:3 poster in the Export section under the Poster tab
• Select image model, add Image Prompt (use Generate Prompt), select Cover Image Actors
• Click (Generate) Cover

YouTube Thumbnail:

• Generate under the Thumbnail tab
• Select image model, add Image Prompt, select Actors
• Click (Generate) Thumbnail

Editing Images

Edit, upscale, reframe, remove backgrounds, build 3D worlds and browse history for any image in FilmSpark.

Overview#

This section covers various changes or edits you can make to an image. You can edit images for Actors, Locations, Shots and any image you generate in FilmSpark.

Image Editor#

To make edits to an Actor or Shot Image, click on the Edit button under the image. This will open a menu that will let you generate a similar image with a change to it, such as changing the Actor's hair color or clothing.

• Image Edit Model: Click to open a list of image edit models you can use. New models (HiDream, GPT Image 2, and more) are added regularly — see the Model Guide.
• Specialized Image Edit Models:
  • Luma Reframe / Luma Reframe (Flash) - Adjust the Aspect Ratio of an image. Options include Aspect Ratio selection, optional Prompt, and Grid Position for precise canvas placement.
  • Qwen Multiple Angles Edit - Change the camera view of an existing image.

• Adv. Settings: Some models have advanced settings like output resolution and seed numbers.
• Paint: Use in-painting and text placement for precise edits. Click the Paint button to open the Paint Toolbar. Select an area with pencil or eraser, then use the Prompt to add, remove, or change content within that spot.

• Prompt: Describe the edit you want to make. For example, "Change hair color to red".
• Including Reference Images: Click the Add icon in the bottom left corner of the Prompt field to add reference images.
• Number of Edits: Generate 1, 2, or 4 image edits at once in the Image Edit UI.

• Generate Edit: Click to generate the edited image. The credit cost is shown on the button, which now always sits in the lower-right panel. Hover tooltips are available throughout the Image Edit UI.

Turnarounds (Actors Only)#

This feature lets you generate images showing different angles of an Actor. Turnarounds are only available for Actors.

Expression Sheets are also available for Actors — generate a one-sheet of an Actor's expressions, with emoji expression icons to choose from.

• Image Edit Model: Choose the model for generating turnaround images.
• Images At A Time: Select 1, 2, or 4 images per angle.
• Angles: Select which angles to generate (all 5 selected by default):
  • Front 3/4: Actor at 45 degrees, showing more of the left side.
  • Left: 90 degrees, left side profile.
  • Right: 90 degrees, right side profile.
  • Back: 180 degrees, viewed from behind.
  • Isometric: Approximately 30 degrees above eye level.
• Click Generate Turnarounds when ready.

Upscaling Images#

Click the Upscale button under an image to upscale it.

Magnific Creative#

• Optimized For: Provides optimal starting points for different image types.
• Scale: Choose from 2x to 16x.
• Engine: Choose from Automatic, Illusio (digital art), Sharpy (realistic photos), or Sparkle (general-purpose).
• Prompt (optional): Add details for the upscale.
• Creativity: Controls how much the AI generates new details.
• HDR: Adjusts visual appearance.
• Resemblance: Controls fidelity to the original.
• Fractality: Controls density of fine, repeating micro-textures.

Magnific Precision#

• Sharpen: Adjusts clarity and edge definition.
• Smart Grain: Adds realistic fine-grained texture.
• Ultra Detail: Controls fine, intricate detail while maintaining fidelity.

Topaz Upscale#

• Model: Each model handles different types of images and issues.
• Upscale Factor: 1x to 4x.
• Subject Detection: All, Foreground, or Background.
• Crop to Fill: Automatically resizes and crops to match an aspect ratio.
• Face Enhancement / Creativity / Strength: Controls for facial detail enhancement.

Removing Background from an Image#

Quickly remove the background from an image by clicking the Remove BG button under an image.

Generating a 3D World from an Image#

Generate a 3D world from an image by clicking the 3D button under an image. This is useful for viewing different camera angles and perspectives.

• 3D Model: Select the model to use.
• Generate 3D: Click to generate. Credit cost shown on button.
• Navigating in 3D Space:
  • W / A / S / D: Horizontal movement
  • R / F: Up / Down movement
  • Space / Shift: Faster / slower movement
  • Mouse drag + wheel: Look and move
• Capture Screenshot: Click the Camera icon to save a screenshot.
• Enhance Image: Generate a high-res image from your screenshot.
• 3D View: Return to the 3D World.

Adjust Images#

This tool lets you manually zoom in and crop images. Great for creating character close-ups:

1. Duplicate your Actor (preserves the full body view)
2. Rename to something like "Name Close-up"
3. Click Edit → Adjust
4. Use the slider to zoom in on the character's face
5. Click Use Crop
6. Click Upscale
7. Now you have a close-up of your actor

Image History#

Once you've generated at least one image, a History button appears under the image. Click to browse all generated images for a Shot or Asset.

• Details: Shows the prompt, resolution, image model, and edit details.
• Images in this Generation: Thumbnails of all images generated in a batch.
• Shot Navigation Arrows (storyboard only): Cycle through shots.
• Image Thumbnails: Click to display in the image window.
• Download Image: Green download icon in top right.
• Delete Image: Red trashcan icon (permanent — use with caution).
• Use this Generation: Select this image for your Asset or Shot.

Ad Templates

Drop a product image into a pre-made narrative and get a video ad in three steps.

Overview#

Ad Templates let you create video ads quickly by adding an image of the product you want to showcase into pre-made narrative structures. To begin, click on the Ad Templates link on the left side menu of the Explore page, or scroll down the Explore page to the Ad Templates section and choose the Ad Template you wish to use.

Once you have chosen an Ad Template, you're only three short steps away from having a video ad for your product!

Step 1 - Upload Your Product Image#

Upload a photo of the product you wish to showcase.

Step 2 - Choose an Aspect Ratio#

Choose an Aspect Ratio for the ad you wish to create. By default, the Aspect Ratio is 16:9.

Step 3 - Generate Your Ad Video#

When you're ready, click Generate to create a video ad for your product!

Additional Options#

• Customize the prompts, switch models, and more to get the best results.
• You can also edit your ad's shots in the Storyboard and use the Video Editor to finalize your work.

Scene Composer

Build scenes and explore locations from multiple angles without writing a script.

Overview#

Scene Composer lets you generate scenes without using a script. Users create images and then videos and compose a scene using the same familiar tools such as Assets and Storyboard, Editor and Export. Scene Composer has "Scene Coverage" checkboxes which allow you to explore a scene/location from many different angles and shot types.

Step 1 - Creating a Scene#

• To get started, click on the Scene Composer button on the Explore page or on the Scene Composer link on the left side of the screen.
• Click the Create a Scene button to begin.
• In the following menu, type in a description of the scene or location you want to explore. Then click Create.

Step 2 - Scene Description and Coverage#

Once you are in the scene, you will see the Scene Description, the AI-generated Scene Coverage, Assets, Storyboard and Editor sections.

• Scene Description - Describes the scene and location.
• Scene Coverage - Choose from a selection of shots/camera angles to explore the scene from different perspectives. Select the checkboxes for the shots you want and then click the Generate button in the Storyboard section to create your shots.

1. Describe the scene you want to create by entering text into the Scene Description field or add reference images by clicking on the Plus button.
2. Click "Smart Fill" and the AI will fill in the fields below based on your description or image. You can override or set these manually as well.
3. Select the Location, Mood (eg. Dramatic, Comedic), Lighting and Time of Day.
4. Click on the Choose Actors button to select any Actors you wish to add to the scene. The list of Actors you can choose from comes from Actors you have created in other projects in FilmSpark.AI. You can also choose to create new Actors by clicking on the Assets tab.
5. Under Scene Coverage, select the type of shot you want to generate. There are multiple types to choose from including Establishing Shot, Wide Shot, Profile Shot and many more. You can also choose to enter a Custom Shot Description to describe the image you want to generate for more control.

6. Select the image model you want to use and the number of images you want to generate. When you are ready, click on the green Generate button to generate images for a Scene.
7. If needed, you can add new scenes by clicking on the Add Scene button to create an additional scene. New images will always appear in the latest scene.

Asset Studio

Generate standalone images and videos with any model, then bring them into your projects.

Overview#

Asset Studio lets you generate and edit images that can be later imported into projects or scenes. To get started, click on the Asset Studio button on the Explore page or on the Asset Studio link on the left side of the screen.

• Choose an Image Model - Select the image model you want to use to generate an image. There are many different options available, so we recommend trying different ones to find the style best suited for your project.
• Image Prompt - Type a description of the image you want to create in the Prompt Field.
• Aspect Ratio - You can select the aspect ratio for the image.
• Negative Prompt - Type what you don't want to appear in the image.
• Number of Images - Select the number of images you want to generate.
• Generate - When you are ready to generate an image, click on the green Generate button. This button will also show the credit cost.

Generating Videos#

You can now generate videos directly in the Asset Studio, in addition to images. Choose a video model, set your options, and click Generate — the green Generate button shows the credit cost.

The FilmSpark Agent

A director's assistant inside your project. It reads your whole film, plans the work, and asks before it touches a frame.

Meet the FilmSpark Agent#

The FilmSpark Agent is the most powerful tool in the console — a director's assistant that lives inside your project. It reads your entire film, plans multi-step work like a crew, and asks permission before it touches a frame. It runs on your API key and on Skills — playbooks that teach it professional craft, which you can extend with your own. You stay in the chair. It does the clicking.

Tell it "change the actor's jacket to red," "storyboard the whole scene," or "swap the lead across the whole project," and it figures out the steps, picks the right model, and shows you the plan before anything touches your storyboard.

Availability: The Agent is live for Pro and Ultimate subscribers. Enter your own API key to use it. Credit estimates are shown for new image and video generations.

Your API key#

The Agent runs on your own API key — your account, your control, your spend.

1. Open the Agent on the far right side of any project and enter your API key. If in a Team Workspace, you can enter this in your Workspace Settings.
2. Paste the key from your provider dashboard and save. You can add one or more provider keys — OpenAI, Gemini, and Anthropic — and the Agent runs on whichever the job needs.
3. A green Connected state confirms it. Every request the Agent makes runs through your account, and your key never leaves your workspace.

Your Skills#

Skills are the playbooks that teach the Agent FilmSpark's craft — the rules a working crew would know: how a scene breaks into beats, which references a model needs attached, how to keep a face and a voice consistent across a hundred shots.

The default pack comes preloaded. Open the Skills panel and every FilmSpark skill is already installed and ready — no setup.

Add your own. Built a technique your team relies on? Upload your own custom skill file — including custom GPT or Claude Skills (.md, .skill, .txt files) — and it runs exactly like ours. This is where the Agent becomes yours.

You never choose a skill by hand. The words in your request do the routing. Name a model ("in Seedance," "for Kling O3") and you get that engine's exact grammar. Name a format ("storyboard panel," "location board") and you get that deliverable. Say neither, and you get FilmSpark's portable house format that works on any model. The Agent names the Skill it's using as it works, so you always know where a result came from.

The default Skills pack#

Every skill in the pack and what it does. 🆕 marks skills added since the Skills Guide was written.

For phrasings that route cleanly every time — "Break down this scene into shots," "Make a Seedance ref prompt out of this shot," "Change X, keep everything else" — see the FilmSpark Skills User Guide.

Say it plainly#

No special syntax. Describe the outcome and the Agent works out the steps.

It knows your whole project#

The Agent reads your script, scenes, shot list, characters, and the current state of your storyboard. You don't re-explain what you're working on or paste context into every request.

That awareness runs deep. It can summarize the project, list every scene with descriptions and shot counts, pull script snippets for continuity, and break any shot down to its camera settings, lighting, and reference images. It tracks your actors, locations, and props, including their visual descriptions and key features.

It also reads what you currently have selected in the interface, so a request like "fix the lighting on this one" lands on the right shot.

Suggest, then approve#

Nothing gets written to your project automatically. Every change arrives as a proposal card you review and approve. The Generate button glows when something's ready for you.

If a request involves several moves, the Agent shows the full execution plan in a side panel before it runs — every step, the assets it will add, and the credit cost, all before anything touches your project. No black box.

When one step in a multi-step plan misses, you retry just that step instead of starting the whole thing over. And multiple chats can run at once, each working its own job.

Start with your cast and your world#

Character boards and location boards are the identity references for the whole project — every shot and storyboard afterward inherits them.

Character boards. One line — "Make a character board for April" — and the Agent builds an actor continuity sheet: full turnarounds, headshots, expressions, and wardrobe callouts, with multiple actors side by side on one sheet. Your actors hold across every shot that follows.

Location boards. Same trick for your world. The Agent maps the location — hero view, top-down geography with labeled callouts, a camera-angle strip, and blocking notes — so screen direction and reverse shots stay honest in every scene you shoot there.

Edit shots in place#

You can edit a finished shot by describing the change, and the Agent edits the existing image rather than regenerating from scratch. "Remove the boots, fill with mud." "Add a hat to this character." It picks the right model, runs the edit, and drops the result back into the shot.

The same works at scale. Tell it to "remove all the actors in this scene," or across the entire story, and it edits the shots instead of rebuilding them.

Edits use change-plus-preserve grammar: what changes, and an explicit inventory of what's locked. "Change only the lighting to dusk, keep everything else." The preserve list is what stops drift.

Generation that resolves its own dependencies#

The Agent works out the right order on its own.

Say "change the actor's jacket to red." It edits the character first, then regenerates or edits the shots that use them, so everything stays consistent. Need a shot that calls for a character who doesn't exist yet? It generates the missing reference image first, then builds the shot from it.

It also chains multi-step jobs from a single prompt: generate a missing asset, wait for the result, then composite it into a new shot.

Project-wide changes#

Big structural edits run from one instruction. Swap one actor for another across the whole film, and the Agent updates every place that character appears — shot images and shot descriptions both, in the right order, as one approval. It can add or remove single or multiple shots and assets at once, with full awareness of where they sit in the story.

Batch operations bundle a mix of these into one ordered proposal. Create a scene, add shots to it, attach assets, regenerate the prompts, and queue the images — all in a single approval card.

Storyboards, two ways#

From finished shots. Ask the Agent to convert a scene into a storyboard panel and it does two things in one pass: assembles your approved shots into one annotated board — numbered panels, timing, stage directions — and writes the Seedance prompt that animates it, with your board, characters, and location attached as tagged references. Nothing on that sheet you haven't signed off on, and the video prompt is ready before you've scrolled down.

From nothing (anchor mode). If the scene has no shots yet, the same ask boards the whole scene in one generation — missing references get created first (actor stills, location plate), then the board is anchored to the character sheet, the location board, and whatever anchor shot you do have. Panels born together are born consistent. It's faster and cheaper: you approve the sheet instead of every frame. Longer scenes split into multiple boards at beat boundaries.

Shots first for maximum control, board first for maximum speed — your call, scene by scene.

Board becomes motion#

One line turns the board into video: ask for a Seedance prompt built from the storyboard panel, approve it, and Generate glows. All needed shots run as one generation, continuity intact — the cuts land exactly where the board says they should. You get spillover effects and cross-cut continuity that separate clips can't give you.

Video continuation and wardrobe#

For video, the Agent carries information forward. It pulls the description and prompt from the previous shot into the next one and transfers specific details to hold continuity from shot to shot.

Wardrobe changes work the same way. Switch a character's outfit partway through the story to reflect a costume change. New outfits need new assets to pull from, and the Agent will flag that.

You can also point it at several shots and have it combine them into a single multi-shot video prompt.

Seedance 2.0 reference multi-shots#

Point the Agent at a range of shots and have it build a Seedance 2.0 multi-shot from them. Tell it "take shots 1 through 5, use them as references, and make a Seedance 2.0 multi-shot with timing."

It writes the prompt, pulls in your shot images as the references, and uses @Image tagging to assemble them into a single multi-shot sequence with the timing you asked for. You get one Seedance 2.0 prompt that strings your existing shots together, ready to generate.

Sharper prompts and descriptions#

The Agent helps tighten your image and video prompts when the output isn't quite landing. It can regenerate a shot's structured details, rebuild the image and video prompts from the current shot state, or refresh an asset's fields — either fully or filling only what's missing. It respects your auto-update settings, so manual prompts you've customized stay put unless you ask.

Compare models side by side#

Run the same shot across 2 to 12 image models at once and compare the results side by side. The same comparison works for assets. Default to the Agent's model choice, or call it: "Use Nano Banana Pro at 4K." You decide. (See also Compare Mode.)

Threads: run Agent chats in parallel#

The Agent doesn't make you wait in line. Open multiple Agent chats and each works its own job: a project-wide recast in one thread — "replace Hank with Ellen in every shot, descriptions included," executed in the right order as one approval — while a second thread boards the next scene and you're already typing a third ask.

You can also attach one or more reference images to any Agent prompt to make your intent unmistakable.

Memory and sessions#

The Agent remembers your conversation within a session, so closing the panel doesn't wipe the thread. It tracks whether it was open or closed per project and reopens the way you left it. When you want a clean slate, one click clears the chat — with a confirmation so you don't lose it by accident.

Getting the best out of it#

• Set up your cast and locations first. Character boards and location boards are the single best consistency investment in a project — every shot and storyboard afterward inherits them. Identity lives in references, not prose; shots only describe what's new.
• Name the destination up front. "Break down the scene into a storyboard" routes correctly in one step; deciding later can send work down the wrong path.
• Name the model when it matters. "Create a Seedance 2 prompt and assign the reference images." Say nothing and you get FilmSpark's portable format that works on any model.
• Give references jobs. Drag images straight into the chat — "this one is the outfit, this one is the location." Untagged refs are dead weight; references with jobs do far more work than adjectives.
• One beat per shot. If two actions compete, that's two shots. Boards and multi-shot exist for sequences.
• Your assets are safe. Re-running a sheet or board creates a new version — approved originals are never overwritten unless you ask.
• You pick the image quality. High, Medium, or Low before each generation, so you decide where the credits go.

Know the hard limits: video generations run 4–15 seconds, so a long scene is automatically split into segments — and stitched in the Editor. Seedance Reference takes 12 files max, Kling O3 takes 7 refs combined, and boards hold up to 9 panels. There's no text-to-video — everything starts from a still or a reference. The Agent handles the math; you approve the plan.

What it generates#

Putting it together, the Agent can:

• Edit any shot field: shot type, perspective, focal length, camera movement, lighting, color palette, action and performance, atmosphere, staging, dialogue, attached assets, duration, and per-shot model overrides.
• Create and delete shots, and create, edit, or delete scenes.
• Create, edit, or delete actors, locations, props, and other assets, down to their visual details and image prompts.
• Generate shot images, shot videos, both together, and asset images, with optional one-time model overrides.
• Build character boards, turnaround sheets, location boards, location coverage, staged shots, storyboard panel sheets, and Seedance multi-shot prompts.
• Update the project title and logline.

Everything above is a proposal. You approve it, and only then does it touch your project.

Team Workspaces

Pooled credits, shared billing, role-based access and real-time collaboration for production teams.

Overview#

Team Workspaces let multiple people collaborate inside a single FilmSpark account with pooled credits, shared billing, and role-based access. Designed for production teams, agencies, and studios running multiple projects at once.

Team Workspaces are currently available to users on an Enterprise plan or trial access. Contact sales@filmspark.ai to get set up.

Roles#

Every workspace member is assigned one of four roles:

• Owner — Full control. Manages billing, seats, members, and workspace settings. Every workspace has exactly one Owner.
• Admin — Can add and remove members, change roles, and view credit usage across the workspace.
• Member — Can create and work on projects, generate images and video, and use all FilmSpark features within the workspace's credit pool.
• Viewer — Read-only access. Can view projects but not generate or edit.

Inviting Members#

Owners and Admins invite members by email.

• If the invited user already has a FilmSpark account, they're added to the workspace immediately and can switch into it from their account menu.
• If they don't have an account, they'll be prompted to create one first, then automatically joined to the workspace.

Working Together in Real Time#

Team Workspaces support real-time collaboration — multiple people can open the same project at the same time and work together.

Teammate Status#

When other team members are in the project, their names appear at the top center of the screen so everyone can see who's active.

This makes it easy to coordinate — no more guessing whether a teammate is still working on a shot or has moved on.

Shot Notes#

Every shot has a Shot Notes field that's shared across the entire team. Use it to leave direction, feedback, or context that other collaborators will see when they open the shot.

Good uses for Shot Notes:

• "Don't regenerate — the client approved this one."
• "Needs a wider angle — try lens 35mm next pass."
• "I'm working on this shot right now."
• Direction or feedback from a director, producer, or reviewer.

Shot Notes are the most reliable way to communicate inside a project, since they live with the shot itself and don't disappear.

What to Watch Out For#

Real-time collaboration is powerful, but a few actions can cause conflicts if two people do them at the same time:

• Don't generate the full Shot List or Storyboard while others are working in the project. These actions regenerate large portions of the project and can overwrite work your teammates are doing. Coordinate before running them — ideally, do bulk generation when you're the only one in the project.
• Don't generate the same shot at the same time. If two people hit Generate on the same shot, the first person's generation wins, then the second one runs after — it gets confusing fast and wastes credits. The status indicator will tell you if someone else is in the same shot. Check for this before you generate.

• Communicate before bulk operations. Anything that affects multiple shots at once — bulk regeneration, script reimports, shot list updates — should be coordinated with the rest of the team first.

Best Practices#

• Divide shots before starting. Assign specific shots or scenes to each team member so no one steps on each other's work.
• Check who's in the project. Before you generate or edit, glance at the presence indicator at the top of the screen.
• Use Shot Notes for direction. They stay with the shot and everyone sees them.
• Talk in your team's channel. FilmSpark tells you who's there, but Discord or Slack is still the fastest way to coordinate who's doing what.

Seats#

Team Workspaces start at a minimum of 10 seats and can be expanded to 20 seats or more depending on your Enterprise contract.

• Seat count is controlled by the Owner. Contact your FilmSpark account rep to add more seats.
• Adding and removing members is handled in-app by the Owner or Admins. As long as you stay within your seat count, no billing change is needed.

Credits#

Credits in a Team Workspace are pooled across all members — everyone on the team draws from the same bucket.

• Owners and Admins can see how credits are being used across the workspace.

• Members see the shared credit balance but don't manage allocation.

Coming soon: Per-member credit allocations so Admins can cap individual usage.

What's Shared vs. Personal#

In the current release, the Asset Studio and its content are personal to each member — not shared across the workspace. Each member has their own working space inside the workspace.

The workspace provides the shared container — pooled credits, unified billing, member management, and role-based access — while individual creative work stays with the person who made it.

Broader sharing of projects and assets across workspace members is on the roadmap.

Permissions#

Current permissions in Team Workspaces:

• Owners and Admins manage members, seats, and role assignments
• Owners and Admins view credit usage
• Owners and Admins control which AI models and defaults are available to workspace members
• Only the Owner manages billing

Billing#

Team Workspaces are billed under an Enterprise contract, invoiced to the Owner. Individual members don't need their own FilmSpark subscription — their access is covered by the workspace seat.

Getting Started#

To set up a Team Workspace for your company or production, contact sales@filmspark.ai.

Available as either:

• Trial access — evaluate Team Workspaces with your team before committing
• Enterprise plan — full contract with invoice billing, dedicated support, and enterprise features

Coming Soon#

Admin controls to restrict which AI models specific members can use.

Note: currently, export and project deletion are not restricted by role — any Member can export or delete projects they've created.

Single Sign-On (SSO / SAML)#

Enterprise workspaces support SAML-based Single Sign-On so your team can log in using your company's identity provider. No separate FilmSpark password to manage.

Direct integrations:

• Microsoft Azure AD (Entra ID)
• Google Workspace
• Okta Workforce

Any SAML 2.0–compliant identity provider can also be configured as a custom connection.

Why SSO matters:

• One login for everyone. Team members sign in with their existing company credentials.
• Centralized control. When someone leaves your company, revoking their IdP access automatically cuts off FilmSpark access.
• Enterprise-grade security. SSO is required by most studio and agency security reviews and is standard across our Enterprise contracts.

My Library

Your central store of images, videos and audio across every project.

Overview#

My Library is your central store of images, videos, and audio across all projects. Recent updates made it much faster to find, upload, and reuse assets.

Finding and filtering#

• Filter images and videos by project to quickly locate an asset.
• Redesigned Projects tab for cleaner navigation across your work.

Uploading and reusing#

• Upload multiple images at once.
• Select multiple images to use as references together.
• Load from Library is standardized across Shots — add assets the same way everywhere.
• Drag and drop onto the + symbols to load images directly.

First & Last Frame

Control where a shot starts and ends by supplying both frames, and extend shots indefinitely with frame extraction.

Overview#

Some video models support a feature called First/Last Frame. This allows you to specify the Last frame of a video along with the first frame to better guide and control the shot.

Here are the current Models that use this feature:

• Kling v2.5 Turbo Pro
• Kling O3 Standard
• Kling O3 Pro
• Kling v3.0
• Kling v3.0 Pro
• LTX-2.3 Fast
• LTX-2.3 Pro
• Seedance v1.5 Pro
• Veo 3.1 Fast
• Veo 3.1 Quality

Here is a quick tutorial on using this feature and its benefits:

Adding an Image#

In a Storyboard shot, video models that use First/Last Frame will show a sub menu for this feature under the video preview window. Click on it to open the First/Last Frame section.

In the First/Last Frame section you will see two image windows, one for the First Frame and one for the Last Frame. If you have already generated an image for the shot, the First Frame image window will show that image automatically.

To add an image for the Last Frame, you have several options:

• Copy Image - Click on the icon in the upper left corner of the First Frame. This will copy the image to the Last Frame. From here you can Edit this image.
• Flip Images - If you wish to move the First Frame to Last, or switch the Last frame with the First, click on the double arrows in-between the images.
• Adding an image from My Library: Click on the empty image window to bring up My Library. Select the image you want to use and then click the Confirm Selection button. You will have the option to crop the image.
• Uploading an image: You can drag and drop any image from your computer into the image window to upload it.
• Extract Frame: This very powerful feature allows you to extract frames from videos, allowing you to extend shots indefinitely. Click on the Extract button under the Last Frame image window.

Extracting a Frame#

In the Extract Frame window, you can select a video source and frame position to add an image frame from that video for either the First or Last frame.

• Select the Video Source:
  • Last Video Shot: Extract a frame from the video generated for your previous shot.
  • Current Video Shot: Extract a frame from the video for the current shot.
• Select the Frame Position:
  • First Frame: Extract the first frame of a video.
  • Middle Frame: Extract the middle frame of a video.
  • Last Frame: Extract the last frame of a video.
• When ready, click on the green Extract button.

Editing a Frame#

• If you have an image in the First or Last Frame window, click the Edit button to open the Image Editor.
• After generating an edited image, click Use this Generation to apply it. See the Editing Images section for more guidance.

Additional Features#

Duplicating an image: Copy an image from either frame to use in the other. Hover over the image and click the Duplicate button in the top left corner.

Swapping an image: Swap the First Frame and Last Frame images. Click the Swap button between both image windows.

First Last Frame Prompt#

Make sure your video prompt reflects the shot you want to create with the First and Last Frames. Describe the action.

Generating a First/Last Frame Video#

When you have both the First and Last frame images set up the way you want them, just click the (Generate) Video button.

Compare Mode

Generate the same shot or asset across multiple models at once and pick the winner side by side.

Overview#

Compare Mode lets you generate the same shot or asset across multiple models at once and view the results side by side, so you can pick the best one before committing. It works for images and for video.

Multi-model image generation#

When generating an Actor, Location, Prop, or Shot image, you can select multiple image models and generate with all of them in a single pass. Each model returns its own result into a shared comparison view.

• Select 2 or more image models in the generation panel.
• Set the number of images and any per-model advanced settings.
• Click Generate — the credit estimate covers all selected models.

Multi-model video generation#

You can now generate multiple videos for a single shot using different video models and compare them in Compare Mode — the same side-by-side workflow as images.

Compare View#

When a Shot or Asset has multiple generated images or videos, click the Compare button in the History view to see them all on one screen.

• View every result for a shot/asset together at higher resolution.
• Pick the winner to set it as the shot's or asset's active image/video.
• For assets, you can update the actor/asset description directly from a multi-model generation.
• Per-model advanced settings are viewable for multi-model generations.

Tip: Compare via the Agent#

The FilmSpark Agent can run comparisons for you — "run this shot across Nano Banana Pro and GPT Image 2 and compare." It supports 2 to 12 models at once. See The FilmSpark Agent.

For model-by-model guidance on which to compare, see the Model Guide.

Color Grading & LUTs

Apply a consistent film look across your project with built-in or custom LUTs over the Cineon Log style.

Overview#

FilmSpark supports color grading with LUTs so you can apply a consistent film look across your project. This works in conjunction with the Cineon Log art style.

Using LUTs#

• Use one of FilmSpark's built-in LUTs, or upload your own .cube or .bin file.
• LUTs pair with the Cineon Log art style for a gradable, flat starting point.
• Adjust LUT intensity up to 200% for a stronger or subtler look, with improved live previews.

Related styles#

• Cineon Log art style — a flat, 8 bit log-style base intended for color grading with a LUT in FilmSpark or your editor of choice.
• Sunset lighting style — a lighting option in Image Shot Details with improved scene-lighting consistency.

Video Upscalers

Turn 720p or 1080p clips into polished 4K with Topaz and Crystal, and know which model to pick.

Overview#

Video Upscalers take a generated shot and increase its resolution, sharpen detail, and clean up artifacts — turning a 720p or 1080p clip into a polished 4K delivery.

FilmSpark offers two upscalers:

• Topaz — The industry-standard suite used in professional post-production pipelines. Multiple models tuned for different types of footage. Covered in detail below.
• Crystal — A second upscaler option available alongside Topaz. Use whichever gives you the best result for your shot.

This guide focuses on Topaz, since it's the more configurable of the two and most users will start there.

Topaz Quick Start: Which Topaz Model Should I Use?#

If you're not sure, use Proteus at 2x. It's the default for a reason — fast, clean, temporally stable, and works well on most AI-generated video from Seedance, Kling, Veo, Runway, and Wan.

For premium 4K delivery where you want the best possible result and don't mind waiting longer, use a Starlight model.

The Topaz Models#

Everyday Workhorses#

Proteus — The default. Best all-around upscaler for most footage. Sharpens detail and reduces compression artifacts without hallucinating new content. Use this for social media cuts, client previews, and anything where you want predictable, natural-looking results.

Artemis HQ / MQ / LQ — Denoise and sharpen, tuned to your source quality. Pick HQ when your source is already clean, MQ for average footage, LQ when the input is rough.

Gaia HQ / CG / Gaia 2 — Built for rendered content. Use Gaia CG or Gaia 2 on 3D world outputs, animated shots, or anything stylized. Gaia 2 is purpose-built for animation and motion graphics at 2x.

Denoise Specialists#

Nyx / Nyx Fast / Nyx XL / Nyx HF — Focused on removing noise and grain. If your footage has visible grain, compression noise, or was shot (or generated) in low-light conditions, Nyx cleans it up while preserving detail. Use Nyx Fast when speed matters more than maximum quality.

Premium (Diffusion-Based)#

The Starlight family uses diffusion models to rebuild detail — more like regenerating the image than enhancing it. Results are stunning but take longer.

Starlight HQ — The flagship. Best overall quality for 4K delivery.

Starlight Sharp — Tuned for facial detail recovery. Use this on character-heavy shots where faces need to hold up at 4K.

Starlight Mini — Lighter version of HQ. Faster, still uses diffusion.

Starlight Fast 1 / Fast 2 — Speed-optimized diffusion variants. Use when you want the Starlight look without the full wait.

Starlight Precise 1 / 2 / 2.5 — Tuned for tighter adherence to the source. Less hallucination, more faithful to what's actually in the frame. Good for shots where subjects or text need to stay consistent.

Recommended Settings by Use Case#

• Social media post (1080p, speed matters): Proteus, 2x
• Client preview or 4K delivery (live-action style): Starlight HQ, 2x
• 3D worlds or animated shots: Gaia CG or Gaia 2, 2x
• Character-heavy shot with facial close-ups: Starlight Sharp, 2x
• Noisy or grainy source: Nyx, 2x (run this first, then upscale with Proteus if needed)
• Motion graphics or 2D animation: Gaia 2, 2x

Upscale Factor#

You can upscale 2x, 4x, or higher. In practice:

• 2x — The sweet spot. A 1080p clip becomes 4K. Fastest and the quality jump is usually all you need.
• 4x — A 720p clip becomes close to 4K. Use this when your source is lower resolution.
• Higher factors — Available but rarely needed. Quality gains diminish fast.

The middle values beneath all the sliders represent the current selection.

Frame Rate#

You can also convert frame rate during upscale — turning 24fps footage into 60fps for smoother motion, or creating slow-motion.

Crystal#

Crystal is a second upscaler available in FilmSpark. It works differently from Topaz and is worth trying as an alternative when Topaz results don't match what you're looking for. A dedicated guide for Crystal will follow.

Tips#

• Upscale is the last step. Do all your editing, color work, and sequencing first. Upscale the final cut, not every intermediate version.
• Test on a single shot first. Run one shot through your chosen model before committing to a full project upscale. Every project's footage responds a little differently.
• Starlight hallucinates detail. That's a feature when you want a dramatic quality jump, but a bug if you need faithful reproduction. If the source is reference footage or has text/logos that must stay accurate, use Proteus or Starlight Precise.
• Match the model to the content. Using Gaia on live-action or Proteus on animation will give you worse results than picking the right specialist.
• Try both upscalers. If Topaz isn't giving you what you want on a particular shot, run it through Crystal and compare.

Prompt Library

Save prompts that worked and reuse them across shots, assets, projects and teammates.

Overview#

The Prompt Library lets you save your favorite prompts for easy re-use and share them with teammates — so a prompt that worked well becomes reusable across shots and projects.

Saving and reusing prompts#

• Save an image or video prompt to the library from the prompt field.
• Reuse a saved prompt on any shot or asset.
• Share saved prompts with teammates in your workspace.

What you can store#

The library covers prompts for Shots as well as Locations, Props, and Miscellaneous assets, so structural and asset prompts can both be saved and reused.

Credits & Subscriptions

Credit packs, subscription plans, and how to manage or cancel your subscription.

How to Buy Credits#

If you wish to purchase additional Credits, click the "Credits" button in the upper right corner of the screen.

The Credit Packs available for purchase are:

• Creator Top-Up Credits Pack ($10): 750 Credits
• Filmmaker Credits Pack ($25): 2,000 Credits
• Studio Credits Pack ($50): 4,500 Credits
• Producer Credits Pack ($99): 10,000 Credits
• Blockbuster Credits Pack ($249): 30,000 Credits

How to Purchase a Subscription#

If you wish to purchase a Subscription Plan, click the "Upgrade" button in the upper right corner of the screen.

Subscription Plans – Available in Monthly or Annual billing:

• Standard Plan – 650 credits/month. $15 billed monthly, or $12/month when billed annually ($144/year)
• Pro Plan – 2,500 credits/month. $50 billed monthly, or $40/month when billed annually ($480/year)
• Ultimate Plan – 6,500 credits/month. $125 billed monthly, or $100/month when billed annually ($1,200/year)
• Enterprise Plan – Tailor a plan to your organization's needs. Contact sales@filmspark.ai.

How to Cancel a Subscription#

To cancel an active subscription, click on the Account button in the top right corner of the app. Click on Profile to see your subscription details. From this screen, click Manage Billing & Payments to cancel. The canceled subscription will last until the end of the billing period.

Tips

Practical reminders for better results, fewer wasted credits and stronger prompts.

Helpful Tips & Reminders#

• Iterative Process: You can always go back to previous steps (Concept, Script, Assets, Storyboards, Video Clips) to make changes and regenerate elements. Your project saves automatically. Keep in mind that if you make changes, you may need to regenerate subsequent steps.
• Regeneration: Not happy with an AI-generated image or video clip? Try generating it again. The AI will try a slightly different approach each time. If you're not getting what you want, try editing the Prompt directly.
• Compare Models: Not sure which model to use? Generate the same shot or asset across multiple models at once and compare them side by side. See the Compare Mode guide.
• Use the Agent: Pro and Ultimate subscribers can ask the in-project AI Agent to make changes in plain language — it proposes the work for you to approve. See the The FilmSpark Agent guide.
• Credits: Keep an eye on your credit balance. Most AI generation actions consume credits. The cost will usually be displayed near the generation button.
• Generation Fails or Odd Results: If a generation fails or looks odd, try generating again. You can also add more details to the description for more precise results.
• Want to Change the Art Style? You can do this in the Settings Menu. Click the cog-shaped button in the upper right corner. If Auto Prompt is off for a Shot or Asset, you'll need to click Regenerate Prompt first.
• AI doesn't always know Left from Right. Sometimes prompts will place things on the opposite side. Try using the opposite direction in your prompt.
• Be specific: Clearly distinguish between primary and secondary subjects.
• Use descriptive language: Words like "dramatic," "tense," and "chaotic" set the mood effectively.
• Prioritize key elements: Mention the most important aspects early in the prompt.
• Avoid ambiguity: Use precise language to reduce interpretation errors.

Compositional Placement Tips#

Use compositional cues to place objects in specific screen locations:

• "The object remains in the bottom-left third of the image, close to the edge."
• "Keep the object anchored in the same screen position as before."
• "Centered in foreground"
• "In the distance, near the horizon"
• "Facing left, looking at camera"

Be Specific, But Not Overloaded: Too little info = vague output. Too much conflicting detail = confusion.

FAQ

Quick answers to the questions we hear most.

Frequently Asked Questions#

How do Credits work?#

FilmSpark.AI uses a credit system for AI generation tasks. Creating text, images, or video clips will consume credits. Your available credit balance is always visible, and the cost for an action is shown before you confirm it. Editing in the timeline editor typically does not consume additional credits unless you regenerate an asset.

Do Credits ever expire?#

Unused monthly credits received with a subscription plan do not carry over month to month. Credits purchased in Credit Packs never expire.

What image and video models does FilmSpark.AI use?#

Please refer to our Model Guide to see all of the image and video models available. There are many to choose from!

What is the FilmSpark Agent?#

The FilmSpark Agent is an in-project AI assistant for Pro and Ultimate subscribers. It reads your whole project and can build or edit shots, assets and scenes from plain-language instructions, proposing each change for you to approve. You enter your own API key to use it. See the FilmSpark Agent guide for details.

How do I generate videos?#

In a Scripted Project and in Scene Composer, videos can be generated in the Storyboard. You will first need to generate or upload an image for a Storyboard Shot. Then click on the Video button above the Shot Image, to the right of the Image button.

How do I switch the visual style of my Story?#

This can be done in the Settings menu. Click the cog-shaped button near the top right corner of the screen.

How do I change the name or description of an actor?#

Go to the Assets section and click on the actor you wish to edit. Update their name, appearance and features in the Description. Remember to regenerate the Prompt (if auto prompt is off) and then generate a new image.

Can I make an actor look like me or someone I know?#

Yes! In the Assets section, click on the actor and choose Create from Photo. Upload a reference photo. You can keep the outfit from the photo if you like. Edit the Description to match for best results.

Can I download my individual assets?#

Yes! Click the Export button (to the right of the Editor button) in a Scripted Project or Scene Composer. You'll find options to export and download your assets, including an XML file.

How do I know what was updated in FilmSpark.AI?#

Click on the Build Notes link in the Left Side Menu anytime you are in the app.

Is there a demo available?#

You can book a demo by clicking on the Book a Demo link in the Left Side Menu.

I don't see the answer to my question. What should I do?#

Check out our Tutorial Guide and Tutorial Videos. You can also ask our Chatbot by clicking the icon near the bottom right corner of the screen. If all else fails, click Contact Us.