model: minimax-h3-remake to POST /api/v1/videos/generations. The main person takes on the face, hair, body and outfit from the photo, and everything else stays closer to the original than in the standard version: other people keep their exact clothing colors, objects such as a ball keep their color, outfits stay consistent across cuts, and close-ups stay clean. Output is 720p with the original audio.
swap_target to say which person to replace.
minimax-h3-swap, so switching is a one-word change to model; the price is 2.4 tokens ($0.048) per second, 20% above the standard version. Outputs of about 13.7 seconds or less use the Pro model; 14 to 15 second outputs are generated with the standard Character Replace model.
duration is set.480p or 720p (default 720p). The video is always generated at 720p at the same price; aspect ratio follows the guide video.swap_details: optional head-to-toe description of the photo's person, up to 800 characters. custom_callback_url: optional public HTTPS URL for status updates.duration, a 14.6-second guide video produces 14 seconds and costs 33.6 tokens ($0.672).duration to render a shorter opening section of a long clip.cost in the response is the amount actually charged.Authorization: Bearer <key>.duration for anything longer than 15 seconds.{"model": "minimax-h3-remake", "image_url": "…", "video_url": "…"} to /api/v1/videos/generations, adding swap_target for multi-person shots. Store img_uuid.GET /api/v1/jobs/detail or use custom_callback_url; download the video before link_expired_at.model value, for 20% more per second.Same inputs; the price is 20% higher. The Pro version keeps other people's clothing colors and object colors, keeps outfits consistent across shots and cuts, and keeps close-ups clean. In our tests a render takes about 1 to 1.5 times as long.
The scene, camera motion, movement and audio track. Logos, watermarks and on-screen text are also kept.
Always 720p. 480p is accepted as a value but still renders at 720p for the same price.
At least 5 seconds. Over 15 seconds is rejected unless you set duration (5 to 15) to use only the first N seconds.
Outputs of about 13.7 seconds or less use the Pro model. Longer outputs are generated with the standard Character Replace model and are still billed at the Pro rate.
Set swap_target to a position or action, for example "the person in the center of the frame".
2.4 tokens ($0.048) per second of output: 12 tokens ($0.24) for 5 seconds, up to 36 tokens ($0.72) for 15 seconds.
A clear, front-facing half-body or full-body photo of a single person.