langgenius/dify
1
0
Fork 0
mirror of https://github.com/langgenius/dify.git synced 2026-07-20 19:50:06 -04:00
Code Issues 775 Packages Projects Releases 10 Wiki Activity

tts cannot switch timbre #5739

New Issue
Closed
opened 2026-02-21 18:12:22 -05:00 by yindo · 10 comments
yindo commented 2026-02-21 18:12:22 -05:00
Owner
Copy Link
Copy Source

Originally created by @zscxc on GitHub (Sep 22, 2024).

Self Checks

  • This is only for bug report, if you would like to ask a question, please head to Discussions.
  • I have searched for existing issues search for existing issues, including closed ones.
  • I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
  • [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）
  • Please do not modify this template :) and fill in all the required fields.

Dify version

0.8.3

Cloud or Self Hosted

Self Hosted (Docker)

Steps to reproduce

image
Open and use the male voice, but click to play is not a male voice but a female voice, and this switch is still a Chinese female voice, is there any place can be adjusted.
The model I used was CosyVoice-300M-Instruct
image
image

✔️ Expected Behavior

I want to be able to adjust the sound when calling through the API

❌ Actual Behavior

No matter how you switch the timbre, it doesn't work

Originally created by @zscxc on GitHub (Sep 22, 2024). ### Self Checks - [X] This is only for bug report, if you would like to ask a question, please head to [Discussions](https://github.com/langgenius/dify/discussions/categories/general). - [X] I have searched for existing issues [search for existing issues](https://github.com/langgenius/dify/issues), including closed ones. - [X] I confirm that I am using English to submit this report (我已阅读并同意 [Language Policy](https://github.com/langgenius/dify/issues/1542)). - [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 the required fields. ### Dify version 0.8.3 ### Cloud or Self Hosted Self Hosted (Docker) ### Steps to reproduce ![image](https://github.com/user-attachments/assets/4f21fa8a-e2f6-4a88-a16c-36127479c1a2) Open and use the male voice, but click to play is not a male voice but a female voice, and this switch is still a Chinese female voice, is there any place can be adjusted. The model I used was CosyVoice-300M-Instruct ![image](https://github.com/user-attachments/assets/54a5dc99-613f-4cb8-afdc-6d131b02d242) ![image](https://github.com/user-attachments/assets/cfaec5b7-dba7-4377-8522-2642905aef18) ### ✔️ Expected Behavior I want to be able to adjust the sound when calling through the API ### ❌ Actual Behavior No matter how you switch the timbre, it doesn't work
yindo added the 🐞 bug label 2026-02-21 18:12:22 -05:00
yindo closed this issue 2026-02-21 18:12:22 -05:00
yindo commented 2026-02-21 18:12:23 -05:00
Author
Owner
Copy Link
Copy Source

@dosubot[bot] commented on GitHub (Sep 22, 2024):

Hello, @zscxc! I'm here to help you with any bugs, questions, or contributions you might have. Ask me anything! 😊

The issue you're experiencing with the timbre switching for the CosyVoice-300M-Instruct model might be related to how the payload is constructed and sent to the API. The logic for handling different voice types is implemented in the _build_tts_payload method. This method constructs the payload based on the model_type parameter, which can be one of the following: PresetVoice, CloneVoice, CloneVoice_CrossLingual, or InstructVoice.

Here is the relevant code snippet:

def _build_tts_payload(
        self,
        model_type: str,
        content_text: str,
        model_role: str,
        prompt_text: str,
        prompt_audio: str,
        instruct_text: str,
    ):
        if model_type == TTSModelType.PresetVoice.value and model_role:
            return {"tts_text": content_text, "role": model_role}
        if model_type == TTSModelType.CloneVoice.value and prompt_text and prompt_audio:
            return {"tts_text": content_text, "prompt_text": prompt_text, "prompt_audio": prompt_audio}
        if model_type == TTSModelType.CloneVoice_CrossLingual.value and prompt_audio:
            lang_tag = self._detect_lang_code(content_text)
            return {"tts_text": f"{content_text}", "prompt_audio": prompt_audio, "lang_tag": lang_tag}
        if model_type == TTSModelType.InstructVoice.value and instruct_text and model_role:
            return {"tts_text": content_text, "role": model_role, "instruct_text": instruct_text}

        raise RuntimeError(f"Invalid params for {model_type}")

This method is called within the _invoke method to generate the appropriate payload for the text-to-speech request [1].

To troubleshoot the issue, you can check the following:

  1. Ensure that the model_type parameter is correctly set to the desired voice type.
  2. Verify that the other parameters (model_role, prompt_text, prompt_audio, instruct_text) are correctly provided based on the selected model_type.

If the parameters are correctly set and the issue persists, there might be a bug in the implementation or a misconfiguration in the model settings. You may need to review the configuration files and ensure that the model is correctly initialized and configured to handle the specified voice types.

To continue talking to Dosu, mention @dosu.

@dosubot[bot] commented on GitHub (Sep 22, 2024): <!-- Greeting --> Hello, @zscxc! I'm here to help you with any bugs, questions, or contributions you might have. Ask me anything! 😊 <!-- Answer --> The issue you're experiencing with the timbre switching for the CosyVoice-300M-Instruct model might be related to how the payload is constructed and sent to the API. The logic for handling different voice types is implemented in the `_build_tts_payload` method. This method constructs the payload based on the `model_type` parameter, which can be one of the following: `PresetVoice`, `CloneVoice`, `CloneVoice_CrossLingual`, or `InstructVoice`. Here is the relevant code snippet: ```python def _build_tts_payload( self, model_type: str, content_text: str, model_role: str, prompt_text: str, prompt_audio: str, instruct_text: str, ): if model_type == TTSModelType.PresetVoice.value and model_role: return {"tts_text": content_text, "role": model_role} if model_type == TTSModelType.CloneVoice.value and prompt_text and prompt_audio: return {"tts_text": content_text, "prompt_text": prompt_text, "prompt_audio": prompt_audio} if model_type == TTSModelType.CloneVoice_CrossLingual.value and prompt_audio: lang_tag = self._detect_lang_code(content_text) return {"tts_text": f"{content_text}", "prompt_audio": prompt_audio, "lang_tag": lang_tag} if model_type == TTSModelType.InstructVoice.value and instruct_text and model_role: return {"tts_text": content_text, "role": model_role, "instruct_text": instruct_text} raise RuntimeError(f"Invalid params for {model_type}") ``` This method is called within the `_invoke` method to generate the appropriate payload for the text-to-speech request <sup>[[1]](https://github.com/langgenius/dify/blob/main/api/core/model_runtime/model_providers/sagemaker/tts/tts.py)</sup>. To troubleshoot the issue, you can check the following: 1. Ensure that the `model_type` parameter is correctly set to the desired voice type. 2. Verify that the other parameters (`model_role`, `prompt_text`, `prompt_audio`, `instruct_text`) are correctly provided based on the selected `model_type`. If the parameters are correctly set and the issue persists, there might be a bug in the implementation or a misconfiguration in the model settings. You may need to review the configuration files and ensure that the model is correctly initialized and configured to handle the specified voice types. <!-- Dosu Comment Footer --> *To continue talking to [Dosu](https://dosu.dev), mention @dosu.*
yindo commented 2026-02-21 18:12:23 -05:00
Author
Owner
Copy Link
Copy Source

@zscxc commented on GitHub (Sep 22, 2024):

Should I implement this timbre logic in dify

@zscxc commented on GitHub (Sep 22, 2024): Should I implement this timbre logic in dify
yindo commented 2026-02-21 18:12:23 -05:00
Author
Owner
Copy Link
Copy Source

@dosubot[bot] commented on GitHub (Sep 22, 2024):

The timbre switching logic is already implemented in Dify within the _build_tts_payload method. This method constructs the payload based on the model_type parameter and other related parameters like model_role, prompt_text, prompt_audio, and instruct_text [1].

To ensure the timbre switching works correctly, you should:

  1. Verify that the model_type parameter is set to the correct voice type (e.g., PresetVoice, CloneVoice, CloneVoice_CrossLingual, or InstructVoice).
  2. Ensure that the other parameters (model_role, prompt_text, prompt_audio, instruct_text) are correctly provided based on the selected model_type.

If these parameters are correctly set and the issue persists, it might be due to a bug or misconfiguration. You should review the configuration files and ensure the model is correctly initialized and configured to handle the specified voice types. The configuration file sagemaker_tts.yaml specifies the available voice types and their corresponding parameters [2].

If everything seems correct and the issue still exists, you might need to debug the payload construction and API request handling to identify any discrepancies.

To continue talking to Dosu, mention @dosu.

@dosubot[bot] commented on GitHub (Sep 22, 2024): <!-- Answer --> The timbre switching logic is already implemented in Dify within the `_build_tts_payload` method. This method constructs the payload based on the `model_type` parameter and other related parameters like `model_role`, `prompt_text`, `prompt_audio`, and `instruct_text` <sup>[[1]](https://github.com/langgenius/dify/blob/main/api/core/model_runtime/model_providers/sagemaker/tts/tts.py)</sup>. To ensure the timbre switching works correctly, you should: 1. Verify that the `model_type` parameter is set to the correct voice type (e.g., `PresetVoice`, `CloneVoice`, `CloneVoice_CrossLingual`, or `InstructVoice`). 2. Ensure that the other parameters (`model_role`, `prompt_text`, `prompt_audio`, `instruct_text`) are correctly provided based on the selected `model_type`. If these parameters are correctly set and the issue persists, it might be due to a bug or misconfiguration. You should review the configuration files and ensure the model is correctly initialized and configured to handle the specified voice types. The configuration file `sagemaker_tts.yaml` specifies the available voice types and their corresponding parameters <sup>[[2]](https://github.com/langgenius/dify/blob/main/api/core/tools/provider/builtin/aws/tools/sagemaker_tts.yaml)</sup>. If everything seems correct and the issue still exists, you might need to debug the payload construction and API request handling to identify any discrepancies. <!-- Dosu Comment Footer --> *To continue talking to [Dosu](https://dosu.dev), mention @dosu.*
yindo commented 2026-02-21 18:12:23 -05:00
Author
Owner
Copy Link
Copy Source

@ic-xu commented on GitHub (Sep 25, 2024):

Did you click the publish button again to publish this?

@ic-xu commented on GitHub (Sep 25, 2024): Did you click the publish button again to publish this?
yindo commented 2026-02-21 18:12:23 -05:00
Author
Owner
Copy Link
Copy Source

@zscxc commented on GitHub (Sep 25, 2024):

I clicked it, but it didn't have any effect.

@zscxc commented on GitHub (Sep 25, 2024): I clicked it, but it didn't have any effect.
yindo commented 2026-02-21 18:12:23 -05:00
Author
Owner
Copy Link
Copy Source

@ic-xu commented on GitHub (Sep 25, 2024):

How can I reproduce this bug? Is there any xinference that can be used remotely?

@ic-xu commented on GitHub (Sep 25, 2024): How can I reproduce this bug? Is there any xinference that can be used remotely?
yindo commented 2026-02-21 18:12:23 -05:00
Author
Owner
Copy Link
Copy Source

@ic-xu commented on GitHub (Sep 25, 2024):

I don't have any xinference models available. Do you have any for temporary use? It will be ready in a moment.

@ic-xu commented on GitHub (Sep 25, 2024): I don't have any xinference models available. Do you have any for temporary use? It will be ready in a moment.
yindo commented 2026-02-21 18:12:23 -05:00
Author
Owner
Copy Link
Copy Source

@zscxc commented on GitHub (Sep 25, 2024):

You can run the CosyVoice-300M-Instruct model remotely or use xinferencce to do replication

@zscxc commented on GitHub (Sep 25, 2024): You can run the CosyVoice-300M-Instruct model remotely or use xinferencce to do replication
yindo commented 2026-02-21 18:12:23 -05:00
Author
Owner
Copy Link
Copy Source

@zscxc commented on GitHub (Sep 25, 2024):

Do you use wechat? Or email you the identifier for the sunflower.

@zscxc commented on GitHub (Sep 25, 2024): Do you use wechat? Or email you the identifier for the sunflower.
yindo commented 2026-02-21 18:12:23 -05:00
Author
Owner
Copy Link
Copy Source

@ic-xu commented on GitHub (Sep 25, 2024):

OK, wait a moment.

@ic-xu commented on GitHub (Sep 25, 2024): OK, wait a moment.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
Knowledge Pipeline
Proposal
WIP
cloud
dependencies
feat:webapp
good first issue
outdated
performance
pull-request
Mirrored from GitHub Pull Request
 python:uv
refactor
size:L
stale
status: accepting prs
status: blocked
status: in discussion
stt & tts
vulnerablity
web
wont-fix
🌊 feat:workflow
🌚 invalid
🐞 bug
👻 feat:rag
💪 enhancement
📚 documentation
🔨 feat:tools
🙋‍♂️ question
🙏 help wanted
🤔 cant-reproduce
🤖 feat:agent
No Label 🐞 bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: langgenius/dify#5739
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?