[GH-ISSUE #24613] feat: Custom chunking/preprocessing for TTS #107353

New Issue

GiteaMirror · 2026-05-18T06:09:15-05:00

GiteaMirror commented

2026-05-18 06:09:15 -05:00

Originally created by @ccdv-ai on GitHub (May 12, 2026).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/24613

Check Existing Issues

I have searched for all existing open AND closed issues and discussions for similar requests. I have found none that is comparable to my request.

Verify Feature Scope

I have read through and understood the scope definition for feature requests in the Issues section. I believe my feature request meets the definition and belongs in the Issues section instead of the Discussions.

Problem Description

Current implementation is limited to punctuation, paragraph or None. In some situations, a custom splitting strategy is required to improve the generated speech (without sacrifying the latency).
For example, splitting on punctuation doesn't separate titles from paragraphes.

Desired Solution you'd like

Allow to define a custom function that takes as input a string and output a list of chunks.

import re
def tts_chunking_function(text: str):
    # Very basic custom chunking function
    text = re.sub(r'\n+', '\n', text)
    return text.split('\n')

Alternatives Considered

No response

Additional Context

No response

Originally created by @ccdv-ai on GitHub (May 12, 2026). Original GitHub issue: https://github.com/open-webui/open-webui/issues/24613 ### Check Existing Issues - [x] I have searched for all existing **open AND closed** issues and discussions for similar requests. I have found none that is comparable to my request. ### Verify Feature Scope - [x] I have read through and understood the scope definition for feature requests in the Issues section. I believe my feature request meets the definition and belongs in the Issues section instead of the Discussions. ### Problem Description Current implementation is limited to punctuation, paragraph or None. In some situations, a custom splitting strategy is required to improve the generated speech (without sacrifying the latency). For example, splitting on punctuation doesn't separate titles from paragraphes. ### Desired Solution you'd like Allow to define a custom function that takes as input a string and output a list of chunks. ```python import re def tts_chunking_function(text: str): # Very basic custom chunking function text = re.sub(r'\n+', '\n', text) return text.split('\n') ``` ### Alternatives Considered _No response_ ### Additional Context _No response_

GiteaMirror commented

2026-05-18 06:09:16 -05:00

@owui-terminator[bot] commented on GitHub (May 12, 2026):

🔍 Related Issues Found

I found some existing issues that might be related. Please check if any of these are duplicates or contain helpful solutions:

🟣 #9493 enh: Audio splitting per punctuation should also split per paragraph
This is the closest match: it requests changing TTS splitting behavior to better handle paragraphs and latency, which overlaps directly with adding custom chunking/preprocessing strategies. Both are about making TTS chunk boundaries configurable beyond the default punctuation-based approach.
by Mavyre
🟣 #5924 BUG: "punctuation" split should also split at newlines
This issue specifically reports that punctuation splitting ignores newlines and bullet-list formatting. It is related because the new feature request is motivated by the same TTS chunking limitations and also seeks more flexible splitting behavior.
by thiswillbeyourgithub · bug
🟣 #9066 enh: Add TTS Response Splitting option for splitting on markdown (bold/italics)
This is another TTS response-splitting enhancement, but for markdown markers. It is related because it also asks for more user-controlled preprocessing/splitting before text is sent to TTS, which is the same general area as a custom chunking function.
by TheRealPSV
🟢 #19861 issue: Sentence skipping in voice mode
Although this is a bug report about skipped sentences, it concerns the voice-mode/TTS dispatch pipeline and sentence segmentation behavior. It is adjacent to custom preprocessing because both touch how text is chunked and fed to TTS during streaming.
by QuitHub · bug

💡 If your issue is a duplicate, please close it and add any additional details to the existing issue instead.

This comment was generated automatically. React with 👍 if helpful, 👎 if not.

@owui-terminator[bot] commented on GitHub (May 12, 2026):  🔍 **Related Issues Found** I found some existing issues that might be related. Please check if any of these are duplicates or contain helpful solutions: 1. 🟣 [#9493](https://github.com/open-webui/open-webui/issues/9493) **enh: Audio splitting per punctuation should also split per paragraph** *This is the closest match: it requests changing TTS splitting behavior to better handle paragraphs and latency, which overlaps directly with adding custom chunking/preprocessing strategies. Both are about making TTS chunk boundaries configurable beyond the default punctuation-based approach.* *by Mavyre* 2. 🟣 [#5924](https://github.com/open-webui/open-webui/issues/5924) **BUG: "punctuation" split should also split at newlines** *This issue specifically reports that punctuation splitting ignores newlines and bullet-list formatting. It is related because the new feature request is motivated by the same TTS chunking limitations and also seeks more flexible splitting behavior.* *by thiswillbeyourgithub · `bug`* 3. 🟣 [#9066](https://github.com/open-webui/open-webui/issues/9066) **enh: Add TTS Response Splitting option for splitting on markdown (bold/italics)** *This is another TTS response-splitting enhancement, but for markdown markers. It is related because it also asks for more user-controlled preprocessing/splitting before text is sent to TTS, which is the same general area as a custom chunking function.* *by TheRealPSV* 4. 🟢 [#19861](https://github.com/open-webui/open-webui/issues/19861) **issue: Sentence skipping in voice mode** *Although this is a bug report about skipped sentences, it concerns the voice-mode/TTS dispatch pipeline and sentence segmentation behavior. It is adjacent to custom preprocessing because both touch how text is chunked and fed to TTS during streaming.* *by QuitHub · `bug`* --- 💡 If your issue is a duplicate, please close it and add any additional details to the existing issue instead. *This comment was generated automatically.* React with 👍 if helpful, 👎 if not.

Sign in to join this conversation.

Branches Tags

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: github-starred/open-webui#107353