[GH-ISSUE #16644] issue: Voice Mode not playing audio until entire audio file has been generated #72597

New Issue

GiteaMirror · 2026-05-13T04:41:57-05:00

GiteaMirror commented

2026-05-13 04:41:57 -05:00

Originally created by @Byrd910 on GitHub (Aug 15, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/16644

Check Existing Issues

I have searched the existing issues and discussions.
I am using the latest version of Open WebUI.

Installation Method

Docker

Open WebUI Version

v0.6.22

Ollama Version (if applicable)

No response

Operating System

Fedora 41

Browser (if applicable)

Chrome 139.0.7258.123

Confirmation

I have read and followed all instructions in README.md.
I am using the latest version of both Open WebUI and Ollama.
I have included the browser console logs.
I have included the Docker container logs.
I have provided every relevant configuration, setting, and environment variable used in my setup.
I have clearly listed every relevant configuration, custom setting, environment variable, and command-line option that influences my setup (such as Docker Compose overrides, .env values, browser settings, authentication configurations, etc).
I have documented step-by-step reproduction instructions that are precise, sequential, and leave nothing to interpretation. My steps:
Start with the initial platform/version/OS and dependencies used,
Specify exact install/launch/configure commands,
List URLs visited, user input (incl. example values/emails/passwords if needed),
Describe all options and toggles enabled or changed,
Include any files or environmental changes,
Identify the expected and actual result at each stage,
Ensure any reasonably skilled user can follow and hit the same issue.

Expected Behavior

I would expect Voice Mode to begin playing audio as soon as the TTS engine begins generating it, similar to how the "Read Aloud" button works.

Actual Behavior

I am using Orpheus-FastAPI by Lex-au as my TTS engine, and the Audio Text-to-Speech settings are configured to split response on Punctuation. I can see this is happening when reviewing the Orpheus-FastAPI logs (it is receiving/generating one sentence at a time).

When using the "Read Aloud" button after generating text in a conversation, audio streaming works as expected - i.e. it begins narrating audio as soon as Orpheus-FastAPI starts generating sentences. However, when using "Voice Mode", while response splitting on punctuation and sentence-by-sentence generation is still occurring on the Orpheus end exactly as with the "Read Aloud" button - no audio begins playing until the entire LLM response has been generated by Orpheus, resulting in a long delay (depending on the length of the LLM response).

Steps to Reproduce

Launch Open WebUI using the default Docker Compose file for open-webui:cuda
Use llama.cpp:server-cuda docker image to host any LLM model (I am using Gemma3 27B)
Launch Orpheus-FastAPI with the docker-compose-gpu.yml setup with any model (I am using Orpheus-3b-FT-Q8_0)
Configure Open WebUI's Audio->TTS settings for:
TTS Engine: OpenAI
URL: http://path-to-orpheus:5005/v1
API Key: not-needed
TTS Voice: tara
TTS Model: tts-1
Response Splitting: Punctuation
Ask the LLM in a text chat to "Create a short 200-word story." Once finished, click "Read Aloud." Audio narration begins very quickly, as soon as sentences start being generated by Orpheus.
Start a new chat and enter "Voice Mode." Ask the LLM to tell you a short, 200-word story. Audio narration takes much longer, waiting until the entire LLM response is generated by Orpheus instead of beginning once individual sentences are complete.

Logs & Screenshots

I can provide any logs requested.

Additional Information

No response

Originally created by @Byrd910 on GitHub (Aug 15, 2025). Original GitHub issue: https://github.com/open-webui/open-webui/issues/16644 ### Check Existing Issues - [x] I have searched the existing issues and discussions. - [x] I am using the latest version of Open WebUI. ### Installation Method Docker ### Open WebUI Version v0.6.22 ### Ollama Version (if applicable) _No response_ ### Operating System Fedora 41 ### Browser (if applicable) Chrome 139.0.7258.123 ### Confirmation - [x] I have read and followed all instructions in `README.md`. - [x] I am using the latest version of **both** Open WebUI and Ollama. - [x] I have included the browser console logs. - [x] I have included the Docker container logs. - [x] I have **provided every relevant configuration, setting, and environment variable used in my setup.** - [x] I have clearly **listed every relevant configuration, custom setting, environment variable, and command-line option that influences my setup** (such as Docker Compose overrides, .env values, browser settings, authentication configurations, etc). - [x] I have documented **step-by-step reproduction instructions that are precise, sequential, and leave nothing to interpretation**. My steps: - Start with the initial platform/version/OS and dependencies used, - Specify exact install/launch/configure commands, - List URLs visited, user input (incl. example values/emails/passwords if needed), - Describe all options and toggles enabled or changed, - Include any files or environmental changes, - Identify the expected and actual result at each stage, - Ensure any reasonably skilled user can follow and hit the same issue. ### Expected Behavior I would expect Voice Mode to begin playing audio as soon as the TTS engine begins generating it, similar to how the "Read Aloud" button works. ### Actual Behavior I am using Orpheus-FastAPI by Lex-au as my TTS engine, and the Audio Text-to-Speech settings are configured to split response on Punctuation. I can see this is happening when reviewing the Orpheus-FastAPI logs (it is receiving/generating one sentence at a time). When using the "Read Aloud" button after generating text in a conversation, audio streaming works as expected - i.e. it begins narrating audio as soon as Orpheus-FastAPI starts generating sentences. However, when using "Voice Mode", while response splitting on punctuation and sentence-by-sentence generation is still occurring on the Orpheus end exactly as with the "Read Aloud" button - no audio begins playing until the entire LLM response has been generated by Orpheus, resulting in a long delay (depending on the length of the LLM response). ### Steps to Reproduce 1. Launch Open WebUI using the default Docker Compose file for open-webui:cuda 2. Use llama.cpp:server-cuda docker image to host any LLM model (I am using Gemma3 27B) 3. Launch Orpheus-FastAPI with the docker-compose-gpu.yml setup with any model (I am using Orpheus-3b-FT-Q8_0) 4. Configure Open WebUI's Audio->TTS settings for: TTS Engine: OpenAI URL: http://path-to-orpheus:5005/v1 API Key: not-needed TTS Voice: tara TTS Model: tts-1 Response Splitting: Punctuation 5. Ask the LLM in a text chat to "Create a short 200-word story." Once finished, click "Read Aloud." Audio narration begins very quickly, as soon as sentences start being generated by Orpheus. 6. Start a new chat and enter "Voice Mode." Ask the LLM to tell you a short, 200-word story. Audio narration takes much longer, waiting until the entire LLM response is generated by Orpheus instead of beginning once individual sentences are complete. ### Logs & Screenshots I can provide any logs requested. ### Additional Information _No response_

GiteaMirror added the bug label 2026-05-13 04:41:57 -05:00

GiteaMirror closed this issue

2026-05-13 04:41:58 -05:00

GiteaMirror commented

2026-05-13 04:42:01 -05:00

@Byrd910 commented on GitHub (Aug 15, 2025):

Update: I also tested with a different TTS just now, and was able to still reproduce the issue.

I used Kokoro-FastAPI with the default docker/gpu/docker-compose.yml setup. Configured my Open WebUI Settings->Audio->TTS settings as follows:

TTS Engine: OpenAI
URL: http://path-to-kokoro:8880/v1
API Key: not-needed
TTS Voice: af_bella
TTS Model: kokoro
Response Splitting: Punctuation

Because Kokoro is much faster, I asked for a 750-word story. Using the "Read Aloud" button, audio narration starts almost immediately. Using "Voice Mode" asking for a 750-word story, it takes a very long time to start audio narration (waiting for the entire text to be generated instead of starting narration after one sentence).

@Byrd910 commented on GitHub (Aug 15, 2025): Update: I also tested with a different TTS just now, and was able to still reproduce the issue. I used Kokoro-FastAPI with the default docker/gpu/docker-compose.yml setup. Configured my Open WebUI Settings->Audio->TTS settings as follows: TTS Engine: OpenAI URL: http://path-to-kokoro:8880/v1 API Key: not-needed TTS Voice: af_bella TTS Model: kokoro Response Splitting: Punctuation Because Kokoro is much faster, I asked for a 750-word story. Using the "Read Aloud" button, audio narration starts almost immediately. Using "Voice Mode" asking for a 750-word story, it takes a very long time to start audio narration (waiting for the entire text to be generated instead of starting narration after one sentence).

GiteaMirror commented

2026-05-13 04:42:01 -05:00

@silentoplayz commented on GitHub (Aug 16, 2025):

Related PR - https://github.com/open-webui/open-webui/pull/16152

@silentoplayz commented on GitHub (Aug 16, 2025): Related PR - https://github.com/open-webui/open-webui/pull/16152

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#72597