MiniMax H3 Recast replaces the main people in a source video with people from reference photos, and tracks each person across shots. Upload a 5–30 second video and one photo for each new person. The published description says the model aims to preserve the original motion, gestures, camera, shot timing, and soundtrack. The output is the recast video, with the source sound.
Create with MiniMax H3 RecastCreative controls
MiniMax H3 Recast takes a source video, reference photos, an optional prompt, a resolution, and an enhance-realism switch.
MiniMax H3 RecastUpload a source video of 5 to 30 seconds. Keep each shot to 15 seconds or less. MiniMax H3 Recast accepts MP4, MOV, and WebM on this page, and it aims to keep that clip’s motion, gestures, camera, shot timing, and soundtrack.
Source clipAdd 1 to 4 photos, one for each new person. MiniMax H3 Recast uses them to replace the main people from left to right, in every shot those people appear in. A prompt can name a different match. Photos do not add a separate credit charge.
The prompt is optional. Turn it on to say which photo replaces which person, or what to keep or change. Leave it off and MiniMax H3 Recast still replaces the main people from left to right. The photos are labeled Image 1, Image 2, and so on, so the prompt can point at a specific photo.
Choose 768P or 1080P before you generate. MiniMax H3 Recast starts at 768P. 1080P is the sharper option and uses more credits for the same clip.
Enhance realism asks for more natural skin textures and better blending of the new people into the scene. MiniMax H3 Recast starts with it on. Turn it off when you want the recast without that extra pass.
Credits depend on how long the video is and whether you choose 768P or 1080P. There is no separate duration control. A 10-second clip at 768P costs about 250 credits, and the same clip at 1080P costs about 375 credits. Photos do not change that cost.
Made with Clipin
MiniMax H3 Recast replaces people in a video you already have. Use the clip, the photos, and the optional prompt to set up the change.
Upload one photo for the person who should change, and add a prompt if more than one person is in the source. It aims to keep the original motion, gestures, camera, shot timing, and soundtrack. The finished video includes the source sound, so the prompt is for who appears and what else to keep or change.
Upload one photo per person, up to four. A prompt can say who becomes whom. With the prompt off, the main people are replaced from left to right, each in every shot they appear in. The photo slots are labeled Image 1 onward when you want the prompt to name one.
The optional prompt can name details to keep or change, such as clothing or the background. It still aims to keep the motion, gestures, camera, shot timing, and soundtrack. Compare the result with your source clip. The two videos on this page are the published example, so a later generation can look different.
One credit balance for video, image, and voiceover creation.
For occasional creation
Billed $288 yearly · Save $60
Models included
MiniMax H3
Seedance
FLUX 3Seedream
Nano BananaFor regular content creation
Billed $780 yearly · Save $168
Models included
MiniMax H3SeedanceFLUX 3SeedreamNano BananaFor high-volume creation
Billed $2,184 yearly · Save $204
Models included
MiniMax H3SeedanceFLUX 3SeedreamNano BananaCredits are shared across images, videos and voiceovers. Monthly credits don’t roll over.
It aims to keep the original motion, gestures, camera, shot timing, and soundtrack, including the cuts and the source audio. The video you get includes that source sound. You supply the new people with reference photos.
Use a source video of 5 to 30 seconds, and keep each shot to 15 seconds or less. This page accepts MP4, MOV, and WebM.
Add one photo per new person, from 1 to 4. Without a prompt, those photos replace the main people from left to right, in every shot they appear in.
The prompt is optional. Without one, the main people are replaced from left to right, each in every shot they appear in. Turn it on when you need to say who becomes whom, or what else to keep or change.
The form starts at 768P, and you can switch to 1080P before you generate. 1080P is sharper and uses more credits for the same clip.
It asks for more natural skin textures and better blending of the new people into the scene. It starts on, and you can turn it off.
Credits depend on the video length and the resolution. A 10-second clip costs about 250 credits at 768P and about 375 credits at 1080P. Photos do not add a separate charge.
No. They are the published source clip and the published result for this model. A new generation can look different.
Upload a source video, add one photo per person, and check the credit estimate before you generate.
Create with MiniMax H3 Recast