Downloads · 30 days
0
Wan2.2-IS2V
rzgar/Wan2.2-IS2V
Wan2.2-IS2V is a machine learning model from rzgar. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for wan2.2. The card lists the license as apache-2.0.
<video controls width="100%" height="480" <source src="https://huggingface.co/rzgar/Wan2.2-IS2V/resolve/main/Demo/ComfyUI00007-audio.mp4" type="video/mp4" Your browser does not support the video tag. </video
Access
Public
Updated Jul 12, 2026
Repo size
140 GB
Likes
6
Public
Repo makeup
Click a slice to open those files.
.safetensors140 GB · 100%
At a glance
- Library
- wan2.2
- License
- apache-2.0
- Access
- Public
- Created
- Jul 11, 2026
- Updated
- Jul 12, 2026
- SHA
- 386841ed
Base models
Model card
From the Hugging Face model README
Wan2.2 IS2V
<video controls width="100%" height="480"> <source src="https://huggingface.co/rzgar/Wan2.2-IS2V/resolve/main/Demo/ComfyUI__00007-audio.mp4" type="video/mp4"> Your browser does not support the video tag. </video>Similar to Bernini-R-S2V, added speech-driven lip-sync to Wan 2.2 image-to-video checkpoints. Painted masks let you target one face or two speakers in sequence for simple dialog without audio bleeding across the whole frame.
It is not state-of-the-art audio-to-video, but it removes the need for post-processing lip-sync or a separate speech model just to get a talking head from Wan I2V. For basic talking-head clips, or longer videos built from short image+audio segments, it is a handy all-in-one option on top of Wan 2.2 I2V.
Wan IS2V Dual Speaker custom node handles the full I2V path (start_image, concat mask/latent, optional CLIP vision) plus masked S2V audio.
models
| File | Role |
|---|---|
wan2.2_is2v_high_noise_14B_fp16.safetensors | High noise FP16 |
wan2.2_is2v_low_noise_14B_fp16.safetensors | Low noise FP16 |
wan2.2_is2v_high_noise_14B_fp8_scaled.safetensors | High noise FP8 |
wan2.2_is2v_low_noise_14B_fp8_scaled.safetensors | Low noise FP8 |
wan2.2_is2v_high_noise_14B_int8_convrot.safetensors | High noise int8-convrot |
wan2.2_is2v_low_noise_14B_int8_convrot.safetensors | Low noise int8-convrot |
Usage
- Place the models in
ComfyUI/models/diffusion_models/ - Download and save the wav2vec2 audio encoder in
ComfyUI/models/audio_encoders/ - Download and Unzip ComfyUI-WanIS2V.zip into
ComfyUI/custom_nodes/, then restart ComfyUI. - Search for the Wan IS2V Dual Speaker node or use one of the example workflows
LightX2V LoRA
- Wan2.2-I2V-A14B-4steps-lora-rank64-Seko-V1-high_noise
- Wan2.2-I2V-A14B-4steps-lora-rank64-Seko-V1-low_noise
Motion Enhancer
Node: Wan IS2V Dual Speaker
Inputs
| Input | Required | Description |
|---|---|---|
audio_1 | no | Encoded mono audio for speaker 1 |
mask_1 | if audio_1 | Painted mask on the input image. Painted over = lip-sync region for speaker 1. |
audio_2 | no | Encoded mono audio for speaker 2 (dialog) |
mask_2 | if audio_2 | Lip-sync region for speaker 2 |
speaker_2_start_frame | no | When speaker 2 begins (default -1 = auto after speaker 1 ends) |
mask_crossfade_frames | no | Soft blend between speaker masks (default 4, 0 = hard cut) |
audio_inject_scale | no | Strength of audio injection inside the mask (default 1.0) |
- Library
- wan2.2
- License
- apache-2.0
- Created
- Jul 11, 2026
- Updated
- Jul 12, 2026