Back homeStudio guide

How Rumpelstiltskin AI Video Generator works

The product is built for a fast first render: describe a moment, improve the prompt if you want, choose the format and review a private preview.

1. Start with the fixed character

The studio sends two reference images for the fictional Rumpelstiltskin lead so the model can keep the face, hair and outfit coherent. Your prompt controls what happens around that character.

2. Describe movement, not just a subject

Useful prompts say what the character does, how the camera moves, what the light is doing and how the scene changes. For example: “Rumpelstiltskin tiptoes across the hay floor, pauses under a lantern and turns toward the camera as gold dust rises.”

3. Refine the prompt in your language

The optional DeepSeek prompt optimizer expands a rough idea with visual details while keeping the original language. You can edit the optimized version before generation.

4. Add a co-star only with permission

An optional upload can add a supporting or lead character. The form requires a confirmation that you own the image or have permission from the person shown.

5. Choose resolution and frame

The first release creates 15-second scenes in 9:16, 16:9 or square framing. 768P is the default MVP resolution; higher resolutions use more credits.

6. Review a private result

Completed videos are stored in a private Supabase bucket. The account receives a short-lived signed preview URL, and the cleanup task removes outputs after seven days during the MVP validation period.