Source video
Golden-Hour Coastal Drive
This short clip supplies the driver's movement, vehicle motion, timing, composition, and recognizable coastal setting.
MiniMax H3 Face Swap Video Generator
Reimagine one person in a short video with a new facial identity. Add a clear portrait, identify the subject in your prompt, and use the original clip to guide the performance, camera movement, and setting.
Video and portrait reference workflowIdentity guided by a clear face imageFlexible 5–15 second output
Add up to 12 references: 9 images, 3 videos, and 3 audio clips. Each video or audio clip must be 2–15 seconds; videos and audio each have a separate 15-second total limit. Refer to them as Image 1, Video 1, and Audio 1 in your prompt.
320 / 7000
Preview video - Generate your own video above
Real workflow example
MiniMax H3 Face Swap from Inputs to Video
The source clip establishes the driving performance and camera path, while the portrait guides the new facial identity in the generated result.
Face reference
Target Identity Portrait
A clean portrait provides visible identity cues such as facial structure, eyes, freckles, hair, and overall appearance.
Generated result
New Identity, Familiar Performance
The generated driver reflects the portrait while following the action and scene direction established by the source video.
This is an unretouched MiniMax H3 multimodal reference result. It regenerates the shot rather than compositing a face onto every frame, so hair, clothing, expression, framing, and background details may also change.
What Is MiniMax H3 Face Swap?
MiniMax H3 Face Swap is a reference-guided video generation workflow. A source video directs motion and scene structure, while a portrait introduces the target identity. Because MiniMax H3 generates a new clip, the result behaves more like a controlled reshoot than a traditional face overlay.
- The portrait guides identityA sharp reference image gives the model readable facial structure, skin details, eyes, hair, and other distinguishing features.
- The clip guides performanceThe source video provides the action, pose changes, camera movement, timing, setting, and subject placement.
- The scene is regeneratedThe model creates fresh frames instead of tracking and pasting a face, so nearby appearance and scene details can shift.
How to Improve a MiniMax H3 Face Swap
Separate the job of each input: use the video for motion, the portrait for identity, and the prompt to identify the subject and protect the details that matter most.
Start with a Readable Face
Choose a short shot where the target person's face is visible at a useful size. Stable lighting and limited occlusion make identity guidance easier to follow across the clip.
- Prefer one clear subject over a crowded group scene
- Avoid hard cuts, heavy motion blur, and long face obstructions
Choose a Strong Portrait Reference
Use a well-lit image that shows the target identity without filters or covering accessories. A front or three-quarter view usually supplies the clearest facial cues.
- Keep the face sharp, unobstructed, and large in the image
- Match the source angle or include enough detail for profile turns
Protect the Rest of the Shot
Identify the person by position or clothing, connect the new face to Image 1, and list the performance, camera, setting, props, and other people that should remain consistent.
- Say which single person should receive the new identity
- Name preservation priorities before adding stylistic changes
Create a MiniMax H3 Face Swap in Four Steps
A concise source clip, recognizable portrait, and precise subject instruction give the model a clearer identity-editing target.
01
Upload a Source Video
Select a short continuous clip with one visible target face and a performance you want the generated video to follow.
02
Add the Target Portrait
Upload a sharp front or three-quarter portrait with even lighting, natural facial detail, and minimal obstruction.
03
Describe the Identity Swap
Name the person to change, assign the new identity to Image 1, and state which motion, wardrobe, setting, and camera details should remain.
04
Generate and Inspect
Review facial similarity, expressions, hair, motion continuity, and scene stability. Refine one input or instruction at a time.
Creative Uses for MiniMax H3 Face Swap
Explore new on-screen identities when an existing clip already has the performance, pacing, and camera direction your concept needs.
Casting Previsualization
Test how a different original character identity reads in a planned shot before committing to a new production.
Creator Avatar Videos
Use a performance reference to explore short videos starring an avatar or identity you have permission to use.
Narrative Variations
Recast a brief story moment while retaining the broad action, setting, and emotional beat of the reference.
Campaign Concepts
Compare approved spokesperson or fictional-character directions against the same motion and framing concept.
Short-Form Experiments
Create identity-led variations for social clips, visual tests, and personal creative projects.
Alternate Identity Takes
Generate several permitted casting ideas from one concise performance reference without reshooting each take.
MiniMax H3 Face Swap Questions Answered
Learn what this reference-driven workflow changes, which inputs work best, and where generated results differ from conventional face compositing.
Is MiniMax H3 Face Swap a traditional face replacement tool?
No. MiniMax H3 regenerates the video using the source clip and portrait as references. It does not track a face and composite it onto the original frames, so surrounding details may change.
What inputs does MiniMax H3 Face Swap need?
Use one short source video, one clear target portrait, and a prompt that identifies the subject to change. The prompt should also name the motion and scene details you want retained.
How is face swap different from character swap?
Face swap centers on facial identity while aiming to retain the source person's broader role and scene. Character swap intentionally redefines the full performer, including hair, outfit, silhouette, or visual style. MiniMax H3 may still reinterpret those surrounding details in either workflow.
Which portraits work best for MiniMax H3 Face Swap?
Start with a sharp, well-lit front or three-quarter portrait. Avoid strong filters, sunglasses, hands over the face, extreme crops, and tiny faces inside a wide image.
Can MiniMax H3 Face Swap handle several people?
A single clearly identified subject is the most dependable setup. Crowded scenes make it harder to keep identities separated, especially when faces overlap or move out of frame.
Will the original body, clothing, and background stay exact?
Not exactly. They can remain recognizable references, but the output is newly generated. Clothing, hair, hands, expressions, framing, lighting, and background details can be reinterpreted.
Should I get permission before swapping a face?
Use portraits and videos you own or are authorized to transform. Make sure your result respects the depicted person's consent, privacy, publicity rights, and the rules that apply where you publish it.
Create Your MiniMax H3 Face Swap Video
Bring together a short source clip, a permitted portrait, and a focused identity prompt to generate a new take.