Add your scene
Upload a short video with one clearly visible subject. The current Space accepts source clips from about 0.21 to 15 seconds.
Your video. A different character.
Bring a new character into an existing scene. Start with a video and a reference image, then guide the replacement with a simple prompt.
Upload your source video and character image in the community demo.
Have a short video and a reference image ready. Load the demo to upload your files and start creating.
Free access is subject to Hugging Face quota and queues.Start small. Try 3 seconds, 4 Turbo steps, and a fast resolution. This Space uses two GPU stages from your Hugging Face quota.
Your files go to the Space. Hugging Face and the demo operator process uploads. If prompted to sign in, open the original Space in a new tab.
Two inputs, a clear prompt, and a short first test.
Upload a short video with one clearly visible subject. The current Space accepts source clips from about 0.21 to 15 seconds.
Upload a clear reference image, then describe who to replace. Use <Video 1> for the scene and <Picture 1> for your character.
Start with 3 seconds and 4 steps at a fast resolution. Review the face, movement, and background before downloading the result.
Describe the subject, the replacement, and what should stay in the scene.
In <Video 1>, replace the main person with the character from <Picture 1>. Match their appearance to the reference image while following the movement in the video. Keep the camera angle, setting, and other people consistent.
Two ways to try AI video character replacement. Choose the controls that fit your clip.
| What to compare | Free MiniMax H3 Character SwapThe demo on this page | Wan2.2 AnimateCharacter Swap workflow |
|---|---|---|
| Workflow | Prompt-guided character replacement | Character Swap mode with subject masking |
| Model | MiniMax H3 + Character Swap LoRA + Larry Turbo LoRA | Wan2.2 Animate |
| Inputs | Source video, character image, and prompt | Input video and reference image |
| Starting settings | 3 seconds, 4 steps, 960 × 544 for landscape | 2 seconds, Low Res, Character Swap mode |
| Subject control | Describe the person and the intended change | Use Advanced Masking to adjust the selected subject |
| Free access | Hugging Face ZeroGPU quota; two GPU stages per request | Public demo quota and queue limits apply |
What to expect from the free community demo.
The public Hugging Face Space can be tried with available ZeroGPU quota. Wan2.Video does not charge for opening the demo. Shared queues, account limits, and two GPU stages per request mean free generation is limited and availability can change.
You need an original video and a reference image of the replacement character, plus a prompt describing the change. Choose a short, clear scene and a reference that shows the character well. Upload only media you have permission to use.
This is a community Space published by Jit2024. It combines MiniMax H3 with the Character Swap and Larry Turbo LoRA adapters. Wan2.Video provides a guide and access to the third-party demo; it is not the Space operator.
The prompt can ask the model to follow the source performance and retain the setting, but generated video may change faces, timing, clothing, or background details. Inspect the whole clip, especially fast motion and moments when the subject is obscured.
Open the original Space in a new tab, sign in to Hugging Face if prompted, and check your remaining quota. Try a shorter output or a fast resolution. Reloading the embedded demo clears its current form, so save any result first.
Both workflows use a video and a reference image. This H3 Space offers prompt-based direction and Turbo step settings, while the Wan2.2 Animate demo offers Character Swap mode and Advanced Masking. The comparison describes their controls; it is not a quality or speed benchmark.
Space details checked . Generation results depend on the Space and your available quota.