[GH-ISSUE #12482] feat: Access to Input Audio File Without Transcription #16618

New Issue

GiteaMirror · 2026-04-19T22:30:48-05:00

GiteaMirror commented

2026-04-19 22:30:48 -05:00

Originally created by @SpectreSpect on GitHub (Apr 5, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/12482

Originally assigned to: @tjbck on GitHub.

Check Existing Issues

I have searched the existing issues and discussions.

Problem Description

I am working with Open WebUI and need to access the input audio file for processing, but I do not need the transcription feature. Currently, the system automatically transcribes any audio input, which is unnecessary for my use case and can consume additional resources.

Use Case:

My application requires access to the raw audio file for further processing, but I do not need the transcription output.

I want to be able to receive or handle the audio file as-is, without triggering transcription or any related processing.

Desired Solution you'd like

Provide an option to access the raw input audio file directly, without initiating transcription. This could be done by:

Disabling the transcription process while still allowing the audio to be captured and passed to the backend.

Providing a way to retrieve the audio file (e.g., as a raw file object or URL) without it being processed by the speech-to-text engine.

Alternatives Considered

No response

Additional Context

Benefits:

Resource Efficiency: By avoiding transcription, you save processing power and time for use cases that don’t require it.

Improved Flexibility: Allows users to handle the raw audio directly, making it easier to implement custom audio processing workflows without interference from the transcription process.

Simpler Integration: Users can seamlessly integrate the audio file into their systems or other applications without modifying the default behavior of Open WebUI.

This change would be highly useful for scenarios where audio is needed for purposes like playback, analysis, or storage, but transcription is not necessary.

Originally created by @SpectreSpect on GitHub (Apr 5, 2025). Original GitHub issue: https://github.com/open-webui/open-webui/issues/12482 Originally assigned to: @tjbck on GitHub. ### Check Existing Issues - [x] I have searched the existing issues and discussions. ### Problem Description I am working with Open WebUI and need to access the input audio file for processing, but I do not need the transcription feature. Currently, the system automatically transcribes any audio input, which is unnecessary for my use case and can consume additional resources. Use Case: My application requires access to the raw audio file for further processing, but I do not need the transcription output. I want to be able to receive or handle the audio file as-is, without triggering transcription or any related processing. ### Desired Solution you'd like Provide an option to access the raw input audio file directly, without initiating transcription. This could be done by: Disabling the transcription process while still allowing the audio to be captured and passed to the backend. Providing a way to retrieve the audio file (e.g., as a raw file object or URL) without it being processed by the speech-to-text engine. ### Alternatives Considered _No response_ ### Additional Context Benefits: Resource Efficiency: By avoiding transcription, you save processing power and time for use cases that don’t require it. Improved Flexibility: Allows users to handle the raw audio directly, making it easier to implement custom audio processing workflows without interference from the transcription process. Simpler Integration: Users can seamlessly integrate the audio file into their systems or other applications without modifying the default behavior of Open WebUI. This change would be highly useful for scenarios where audio is needed for purposes like playback, analysis, or storage, but transcription is not necessary.

GiteaMirror closed this issue

2026-04-19 22:30:48 -05:00

GiteaMirror commented

2026-04-19 22:30:49 -05:00

@surajssd commented on GitHub (Apr 13, 2025):

Yeah even I was wondering about this. I have deployed microsoft/Phi-4-multimodal-instruct using vllm. How do I tell Open Web UI to use the /v1/chat/completions API for audio files but right now it defaults to using the /v1/audio/transcriptions API. So here is what I can do using curl:

✗ curl -X POST "http://localhost:8000/v1/chat/completions" \
  -H "Content-Type: application/json" \
  --data '{
  "model": "microsoft/Phi-4-multimodal-instruct",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "What is in this audio?"
        },
        {
          "type": "audio_url",
          "audio_url": {
            "url": "https://ia800303.us.archive.org/28/items/HindSwaraj-Speech-03-1/tryst.mp3"
          }
        }
      ]
    }
  ]
}' | jq

I think when the input has an image, then Open Web UI is using the /v1/chat/completions API. But not sure how to do the same for audio!

Even though above image shows an URL but you can also pass raw audio as base64 data.

@surajssd commented on GitHub (Apr 13, 2025): Yeah even I was wondering about this. I have deployed `microsoft/Phi-4-multimodal-instruct` using vllm. How do I tell Open Web UI to use the `/v1/chat/completions` API for audio files but right now it defaults to using the `/v1/audio/transcriptions` API. So here is what I can do using `curl`: ```json ✗ curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "microsoft/Phi-4-multimodal-instruct", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "What is in this audio?" }, { "type": "audio_url", "audio_url": { "url": "https://ia800303.us.archive.org/28/items/HindSwaraj-Speech-03-1/tryst.mp3" } } ] } ] }' | jq ``` I think when the input has an image, then Open Web UI is using the `/v1/chat/completions` API. But not sure how to do the same for audio! Even though above image shows an URL but you can also pass raw audio as base64 data.

GiteaMirror commented

2026-04-19 22:30:50 -05:00

@andi-at-1 commented on GitHub (Aug 26, 2025):

that would be a great feature, voxtral enables you to work with the audio file directly as input, but that won´t work with openwebui

@andi-at-1 commented on GitHub (Aug 26, 2025): that would be a great feature, voxtral enables you to work with the audio file directly as input, but that won´t work with openwebui

GiteaMirror referenced this issue

2026-04-19 23:53:04 -05:00

[GH-ISSUE #16618] issue: Kokoro Browser-based TTS doesn't work on macOS #17978

GiteaMirror referenced this issue

2026-04-25 07:25:24 -05:00

[GH-ISSUE #16618] issue: Kokoro Browser-based TTS doesn't work on macOS #33507

GiteaMirror referenced this issue

2026-05-05 19:52:33 -05:00

[GH-ISSUE #16618] issue: Kokoro Browser-based TTS doesn't work on macOS #56644

GiteaMirror referenced this issue

2026-05-13 04:40:11 -05:00

[GH-ISSUE #16618] issue: Kokoro Browser-based TTS doesn't work on macOS #72582