The timbre stability

[smile-yushu]

Self Checks

• [x] This template is only for bug reports. For questions, please visit Discussions.
• [x] I have thoroughly reviewed the project documentation (installation, training, inference) but couldn't find information to solve my problem. English 中文 日本語 Portuguese (Brazil)
• [x] I have searched for existing issues, including closed ones. Search issues
• [x] I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
• [x] [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）
• [x] Please do not modify this template and fill in all required fields.

Cloud or Self Hosted

Self Hosted (Source)

Environment Details

Windows 3090Ti

Steps to Reproduce

By default, the model will only learn the speaker's speech patterns and not the timbre. You still need to use prompts to ensure timbre stability

python tools/llama/merge_lora.py \
    --lora-config r_8_alpha_16 \
    --base-weight checkpoints/fish-speech-1.5 \
    --lora-weight results/$project/checkpoints/step_000000010.ckpt \
    --output checkpoints/fish-speech-1.5-yth-lora/

What should I do now？

✔️ Expected Behavior

No response

❌ Actual Behavior

No response

[smile-yushu]

I used parameters, but for the same statement, the generated voice tone may not be completely the same, sometimes there may be deviations, even male or female voices

python -m tools.api_client --text "请问您期望的安装日期是几月几日，您可以说8月3日，温馨提示" --reference_audio "D:/AI/fish-speech/fish-speech-main/references/2月21日/2月21日.WAV" --reference_text "开工后这几天冷空气持续发力气温一天天下降今天清晨全区气温都是个位数" --streaming False --output D:/code/video\195567e5 --format wav

[Whale-Dolphin]

We recommend users to use WSL instead of the original Windows. Windows has a very complex environment and is difficult to develop on. We will not test on window in this and future versions.

[shengzhou1216]

The same issue occurs on Linux (Ubuntu 22.04).

[github-actions[bot]]

This issue is stale because it has been open for 30 days with no activity.

[github-actions[bot]]

This issue was closed because it has been inactive for 14 days since being marked as stale.