Skip to content

Qwen-TTS v1.0.7 generate hollow and cracking sounds or crash my PC on ComfyUI #307

Description

@CoudPelle

Description

After testing various Qwen TTS tools, I found that none of them work on my computer. The Custom Voice and Design Voice tools generate silent audio files, sometimes with a crackling sound at the beginning. The Voice Clone tool causes the PC to freeze. Nothing responds after I launch the program. I have to restart the PC. No errors or warnings are generated when running the first two tools. I have tried with different language setup (english, french, auto) but nothing changed.

Here is a sample output of custom voice with the following settings:
text: What's the problem ? I attended to all the things you asked me.
speaker: Eric
model_choice: 1.7B
device: auto
precision: bf16
language: english
seed: random
instruct: Adult male and rough voice. He's tone is sad and angry
max_new_token: 2048
top_p: 0,80
top_K: 20
temperature: 1
repetition_penalty: 1.05
attention: flash_attn

ComfyUI_temp_bfjsw_00001_.zip

Reproduction

  • Clone the repo in comfyUI custom_nodes
  • Download the requirements in the repo with the comfyui pip venv
  • Run the comfyui exemple nodes graph
  • Crash or bug when trying to use it as described above

Logs

Environment Information

OS: Windows 11
Python version: 3.12.11
ComfyUI version: 0.19.5
GPU: AMD RX 6900 XT
CPU: AMD Ryzen 7 5800X

Known Issue

  • The issue hasn't been already addressed in Documentation, Issues, and Discussions.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions