Video Generation

TF Design supports video generation through the ToothFairyAI backend. Video clips are produced as .mp4 files saved to the project workspace.

Available Models

Model	Alias	Strength
Sorcerer Motion	sorcerer_motion	Text-to-video — standard quality, fast generation
Mystica Motion	mystica_motion	Text-to-video — balanced quality and speed
Mystica Motion Pro	mystica_motion_pro	Text-to-video — high quality, longer generation
Mystica Video SD	mystica_video_sd	Stable diffusion video — stylised, artistic output

How It Works

1. Provide a text prompt describing the desired video.
2. The Desktop app sends the request to the ToothFairyAI backend.
3. The model generates a video clip and returns it as an .mp4 file.
4. The output appears as a download chip in the chat.
5. The video can be embedded in artifacts or downloaded.

Video generation typically takes longer than image generation. The task runs asynchronously — you can continue working while the video renders.

Prompt Templates

Video prompt templates are available from the Video Templates tab on the entry view. Categories include:

Category	Examples
Product demos	Animated product showcases, feature highlights
Social content	Short-form vertical video, stories, reels
Motion graphics	Animated logos, text animations, transitions
Backgrounds	Looping video backgrounds, ambient motion
Explainers	Animated diagrams, process visualisations

Using Generated Video in Artifacts

Generated videos are saved to the project workspace and can be:

• Embedded in HTML — <video src="assets/clip.mp4" autoplay loop muted> .
• Used as backgrounds — CSS <video> element with object-fit: cover.
• Downloaded — click the download chip to save locally.
• Included in exports — video files are part of the ZIP archive.

Internal Endpoints

The following endpoints are exposed by the running Desktop app's embedded runtime for in-app tools and the agent:

Endpoint	Method	Description
/api/projects/:id/media/generate	POST	Generate video for a project (specify video model)
/api/projects/:id/media/tasks	GET	List media tasks (poll for completion)
/api/media/tasks/:id/wait	GET	Wait synchronously for a task to finish

Tips

• Be specific — describe camera movement, subject, lighting, and mood.
• Keep prompts concise — overly long prompts can reduce quality.
• Use prompt templates — start from a template and customise rather than writing from scratch.
• Check task status — video generation is asynchronous; use the media task indicator in the app to track progress.