Feedback
minimax h3 max recast
MiniMax H3 Max Recast keeps your footage and changes its cast — upload portraits and the actors on screen are swapped without touching timing, shots or sound.
All Tools
Discover our comprehensive AI-powered animation toolkit
MiniMax H3
MiniMax H3 AI Video Generator
Seedance 2.5
The Future of AI Video Is Here.

Seedance 2.0
The Future of AI Video Is Here.

Veo3.1
Create Stunning Videos with Veo3.1
FLUX 3 Video Generator

Kling 3.0
Next-Gen AI Video Generator
Grok Video Generator
Create Videos from Text or Images with AI
MiniMax H3 video generator
MiniMax H3 AI Video Generator
A Second Casting Pass, Powered by MiniMax H3 Max Recast
The video you bring is treated as the final take: it supplies the performance and the soundtrack, while the portraits you upload supply the people who appear in it.
- The edit you already approved stays intactMotion, framing, cutting rhythm and audio are all inherited from the source clip, so the result matches what you signed off on — only the identities on screen differ.
- Up to four portraits, mapped across the frameEach uploaded photo takes over one performer, matched from the left edge of the picture to the right. When the automatic pairing is wrong, a single short instruction reassigns who becomes whom.
- Performance and lighting come from the sourceEye lines, head turns, comic beats, expressions, set lighting and lens character are carried over from the original footage, while faces and clothing are drawn from your uploads — and each performer is followed through scene changes.
MiniMax H3 Max Recast: Three Steps to a New Cast
Pick a finished clip, attach one portrait for each performer, check the identities in a 768p preview, then render the delivery version at 1080p.
Inside MiniMax H3 Max Recast: What It Does Well
Built for casting rather than creation — the source keeps its movement, cutting rhythm and soundtrack, only the people are replaced, and a 768p pass lets you verify before spending on 1080p.
Cast up to four people, mapped left to right
You can hand over as many as four portraits; by default they claim the leading performers in each shot from left to right. One line of instruction overrides that pairing whenever the default reads wrong.
Gestures and micro-expressions carry over
Where the eyes go, how the head turns, comic timing, facial nuance, on-set lighting and lens behaviour are all inherited from the original take; only faces and wardrobe come from your photos.
Cuts and soundtrack are left alone
The existing cut pattern and audio bed are untouched, so the finished recast drops straight into your timeline instead of forcing a re-edit or a re-record.
Face-only mode when wardrobe must hold
Left unprompted, the tool tends to borrow hair and clothing from the portrait, even a tightly cropped headshot. Asking for the face alone keeps the original styling — confirmed in our 23 October 2026 run.
Low-cost 768p pass before the 1080p render
A quick 768p preview shows whether the casting reads correctly before you commit to 1080p. Runtime mirrors the source, so cost scales with how long the clip runs.
Proven across very different kinds of footage
Sample recasts range from a music-video close-up and a barista commercial to a fencing bout, a skate launch and a fish market at dawn — each keeping its own motion, cuts and sound.
MiniMax H3 Max Recast: Common Questions Answered
Short answers on recasting limits, reference photos, output resolution and the situations where this tool is not the right fit.
What separates a recast from a text-to-video generation?
Generation invents footage from scratch. Here the footage already exists: the people in your photos take over the on-screen performers, while movement, camera work, edit points and audio stay with the original clip.
What length of source video is supported?
Between 5 and 30 seconds, with no single shot longer than 15 seconds — split longer takes first. The recast runs exactly as long as the clip you put in.
How many performers can be swapped, and how are they matched?
Up to four, each with one photo. By default the portraits claim the leading performers from left to right across the frame; a short prompt can re-map the pairing if the automatic order is off.
Which reference photos give the best results?
Sharp, evenly lit, straight-on portraits, one person per image. Skip sunglasses, hats, hands over the face, group shots and low-resolution or heavily filtered pictures. If the original hair and wardrobe must remain, request a face-only replacement.
When should I preview at 768p instead of rendering 1080p?
Use 768p to confirm the casting reads correctly, then render at 1080p for delivery. Pricing follows the length of the clip.
In which situations does this tool fall short?
It cannot invent new action, new camera work or a different location — the source dictates all three. Results weaken in crowded scenes, with tiny or distant faces, fast movement and occlusion after hard cuts, and identity can drift back toward the original performer in wide shots or profile turns.
Try MiniMax H3 Max Recast on Your Own Footage
Upload a finished clip plus portraits of the people you want on screen, and let the swap run — timing, camera work, cuts and audio are preserved exactly.
