How to Use the ArtCraft AI Video Generator

ArtCraft can generate video from a text prompt, but its real strength is image-to-video with a first frame you designed yourself. You stage the shot by placing the set, posing the characters, and framing the camera, render a still, and then let a video model animate it. The result is far more predictable than pure text-to-video.

Video models in ArtCraft

The official README lists 25 video models. Models marked † are currently disabled, and ‡ marks restricted models that are not available in the desktop app.

As a rule of thumb, variants named Fast, Lite, Turbo, or Draft trade some quality for speed and lower cost. Use them for blocking out motion, then switch to the full model for the final take.

Step-by-step: from first frame to finished clip

1. Plan the shot

   Decide what happens in the clip, where the camera is, and how long it lasts. Short clips with one clear action work best with today's video models.

2. Build the first frame

   Use the image tools to create the opening frame. For maximum control, stage it in 3D: place the backdrop and props, pose the character, and set the camera angle. Render the still with an image model. See the image generator guide for details.

3. Open the Video tool

   Choose Video from the start screen in the web or desktop app, and load your first frame as the input image.

4. Pick a video model

   Choose a model for the job, for example Seedance 2.0 for cinematic motion or a Fast/Lite variant for quick tests. Check the credit cost before generating.

5. Write a motion prompt

   Describe what moves and how the camera moves, not what the scene looks like. The image already defines the look. Use the template below.

6. Review and refine

   If the motion is wrong, adjust the prompt. If the composition is wrong, go back and fix the first frame. Regenerating the still is much cheaper than regenerating video.

Writing motion prompts

When you start from an image, the prompt should focus on action, camera, and pacing:

[subject] [does a specific action], camera [movement: slow push-in / pans left /
static / orbits], [speed and mood], [what must stay consistent]

Example:

The clockmaker slowly lifts the pocket watch to his eye and smiles,
camera performs a slow push-in, calm and quiet mood,
keep the workshop and lighting unchanged

For text-to-video without an input image, add the visual description first, then the motion:

[scene and subject description], [action], camera [movement],
[lighting and atmosphere], [style]

Multi-shot sequences with consistent characters

ArtCraft's tools are designed for sequences where the set and the character must not drift between shots:

1. Use Image to Location to establish one environment you can shoot from several angles.
2. Use character posing and identity transfer to place the same character in each shot.
3. Set a different camera for each shot and render a first frame for each one.
4. Animate each frame with the same video model and similar prompt wording.
5. Edit the clips together in your video editor.

Keeping video costs under control

Video is the most expensive thing you can do in ArtCraft. Based on the official estimates, 1,000 credits cover about 63 seconds of Seedance 2.0 or 45 seconds of Kling 3.0 Pro. To get more out of your credits:

• Finalize composition with still images before animating anything.
• Test motion with Fast, Lite, or Turbo variants.
• Generate shorter clips and extend only the good ones.
• Connect your own provider API key if you already pay a provider.

See the full pricing and credits breakdown, or go deeper on ByteDance's flagship model in Seedance 2.0 in ArtCraft.