Wan2.1 InfiniteTalk LipSync (Native) — v1.5

Wan2.1 InfiniteTalk LipSync (Native) — v1.5 is a ComfyUI workflow for generating lip-synced talking-head videos from a face image and speech audio; it builds video segments to follow the audio and outputs an MP4 with the original audio.

In auto-calculated mode, the workflow uses speech duration to set the loop count, creates the first video block, then extends it with previous_frames as context and outputs trim_image for each extension.

ImageBatch appends each trim_image to the accumulated video, and Save Video [Eclipse] combines the final image batch with the original audio.

Provide a starting face image and a voice clip; the audio encoder converts the clip into phonemic feature representations.

The workflow outputs an MP4 from the accumulated image batch and original audio, trimming both to the shorter length before output.

It is designed for ComfyUI and requires Eclipse 4.1.22.

For local use, place the model files in ComfyUI folders for diffusion_models, text_encoders, model_patches, audio_encoders, and vae.

The workflow lists nsfw_wan_umt5-xxl_bf16_fixed.safetensors, wan/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors, and wan/Wan2_1-I2V-14B-720p_fp8_e4m3fn_scaled_KJ.safetensors.

It also lists Wan2_1_VAE_bf16.safetensors, wan2.1_infiniteTalk_multi_fp16.safetensors, and wav2vec2-chinese-base_fp16.safetensors.

Named custom node packages are ComfyUI_Eclipse, ComfyUI-Easy-Use, and ComfyUI-KJNodes.

The loop switch offers auto-calculated loops based on audio duration or manual loops controlled by loop_count; auto-calculated mode is described as the default.

The Settings group exposes output width and height, target frame rate, and loop_count, with examples of 480p or 720p and 24.0 or 30.0 fps.

Before use, verify that the six model files are in the expected ComfyUI folders.

Suggestion · not verified

Before use, verify that Eclipse 4.1.22, ComfyUI_Eclipse, and comfyui-easy-use are available in your ComfyUI setup.

Suggestion · not verified