# Bulin 47 AI — Create a freestyle video

Canonical HTML page: [https://bulin47-ai.com/](https://bulin47-ai.com/)
Relevant section: [Studio](https://bulin47-ai.com/#create)
Document type: public usage summary, not a verbatim page export.
Last substantive review: 2026-10-11 (Asia/Shanghai).
Purpose: Choose valid photos, assign the four crowd roles and submit a generation.

Use this guide to prepare photos and submit the built-in freestyle workflow. You upload ordinary single-person photos, not a seven-view character sheet.

## Required photos

- One clear lead-character photo for Original mode; that photo plus exactly four separate crowd-person photos for Custom mode.
- JPG, PNG or WebP, non-empty and at most 10 MB per image (10,485,760 bytes). Convert HEIC to JPG before uploading.
- Show face, hair, clothes and upper body. Prefer front-facing, well-lit photos with one person and no sunglasses.
- Every depicted person must be 18 or older and authorize this AI processing. Do not upload minors, celebrities or public figures. Original fictional adult images require the necessary rights.

## Steps

1. Open [the studio](https://bulin47-ai.com/#create) and add one lead photo.
2. Choose Original to keep the original crowd as the generation target, or Custom and upload four crowd photos through the multi-file box. Upload order assigns these four roles; changing the order changes the assignment:

   - 1: right-foreground phone/speaker holder, usually seen from behind.
   - 2: rear-left bystander, bald and in a dark shirt near the start of the source.
   - 3: rear-center patterned-shirt bystander, immediately behind the lead on the right.
   - 4: separate rear-right bystander in a dark collared jacket, behind the patterned-shirt person. This is not a distant stairway person.

3. Select 480p (10 credits) or 720p (22 credits). Both crowd modes cost the same.
4. Confirm photo permissions, sign in and submit with sufficient credits.
5. Follow the generation stages. Once submission succeeds, you can leave and return to [My creations](https://bulin47-ai.com/my-creations).
6. Watch the complete result and download the MP4. Each new attempt uses credits.

## Processing and limits
Photo selection alone is a local preview. Submission uploads the photos. KIE's third-party workflow creates a seven-view lead reference, generates two performance sections and assembles the final video with original audio. The current implementation uses Nano Banana 2 for the character reference and Wan 3.0 for video; providers and availability may change. No filming, user-written prompt, user-supplied source clip or uploaded seven-view sheet is required.

The generation display reflects task stages, not a guaranteed completion time. There is no user cancellation control; closing the page does not cancel the job. Four-person substitution is attempted, not guaranteed: occlusion and segment joins can change identity. Generation is not an automatic rights clearance service.

Related: [Costs and sign-in](https://bulin47-ai.com/pricing.md), [Results and troubleshooting](https://bulin47-ai.com/results.md), [Photo processing](https://bulin47-ai.com/privacy.md).
Next: [Choose photos in the studio](https://bulin47-ai.com/#create).