CivArchive
    ← All articles
    Published September 4, 2025by AIXBG_fp8

    Advanced audio generation ultra-long video workflow

    629 views0 reactions0 comments on CivitAI6 collected
    digital humanworkflowswan2.2 s2v

    1. Introduction

    Just upload a picture and audio, and she can speak or sing. The Wan2.2 S2V model can generate continuous videos of over 15 seconds at a time. I optimized the workflow and the result was excellent. The workflow I built is in the attachment. Everyone can download and experience it.

    2. Model Links

    You can find the models inour repodiffusion_models

    audio_encoders

    vae

    text_encoders

    ComfyUI/

    ├───📂 models/

    │ ├───📂 diffusion_models/

    │ │ ├─── wan2.2_s2v_14B_fp8_scaled.safetensors

    │ │ └─── wan2.2_s2v_14B_bf16.safetensors

    │ ├───📂 text_encoders/

    │ │ └─── umt5_xxl_fp8_e4m3fn_scaled.safetensors

    │ ├───📂 audio_encoders/ # Create one if you

    can't find this folder

    │ │ └─── wav2vec2_large_english_fp16.safetensors

    │ └───📂 vae/

    │ └── wan_2.1_vae.safetensors

    3. Usage Method

    1. Load Image: Upload reference image

    2. LoadAudio: Upload your own audio

    3. Set the Duration: The duration of the generated video is measured in seconds.

      If bypassed, it will be generated based on the duration of the entire audio segment

    4. Set the resolution: The default is 400, which is a total resolution meaning 400,000 pixels. It is a resolution size of 480P.

      If you want to generate a 720P video, change it to 920.

    5. Set the Frame rate: The default is 16.

    6. Set the Chunk Length: The default is 77.

    7. Enter Prompt.

    8. Use Ctrl-Enter or click the Run button to execute the workflow.

    Follow me, If you have any questions, you can leave me a message:

    bilibili @AI_小兵哥
    YouTube @AIXBG_fp8

    Attachments