1. Prepare the source clip
fal documents a 5–30 second source video, with no individual shot longer than 15 seconds. Keep the movement, framing, cuts and sound you want in the source. Recast uses them as its target.
2. Add reference photos
Upload one clear image per replacement person, up to four. By default, the first photo replaces the leftmost main person, the second photo the next person to the right. If that mapping is wrong, describe the subjects and replacement order in the optional prompt.
3. Choose resolution and review price
Choose 768p or 1080p. fal publishes $0.30 per second for 768p and $0.45 per second for 1080p. Recast Studio presents its own exact credit quote before any submission once live billing is enabled.
4. Inspect the output
Look closely at identity across cuts, clothing, small props, and speech. The source soundtrack remains, so a newly pictured person can speak in the original performer's voice. Do not publish a result that misrepresents someone's consent or endorsement.
Quick reference
- Source video
- 5–30 seconds; no individual shot over 15 seconds
- Reference images
- One clear photo for each new person; up to four
- Default pairing
- Photo 1 maps to the leftmost main person
- Optional prompt
- Explain who becomes whom or what else should stay
- Resolution
- 768p or 1080p
- Output sound
- The source soundtrack remains, including voices
Sources
Limits and input behavior are from fal's API documentation. The audio and clothing caveat is supported by Fuser's published two-person test. Checked October 13, 2026.