Text-to-Speech (TTS) and Speech-to-Text (STT)

The TTS and STT engines are optional components for AarhusAI, fully supported by Open WebUI. These components are optional because they require a dedicated GPU to run.

Installation

Deploy these services from the following repositories or use any solution that supports the OpenAI Audio API:

• https://github.com/AarhusAI/whisper-api - Speech-to-Text
• https://github.com/AarhusAI/piper-tts - Text-to-Speech

See the links above for detailed installation instructions and configuration options.

Configuration

Create and Seal the Secrets

First, add the secrets for the TTS and STT API keys to local-secrets/openwebui-secrets.yaml and seal it:

kubectl create -f local-secrets/openwebui-secrets.yaml --dry-run=client -o yaml | kubeseal --format yaml > templates/sealed-openwebui-secrets.yaml

Update Open WebUI Configuration

Edit applications/openwebui/values.yaml in the Helm chart and add the following environment variables:

# STT (whisper) - OPTIONAL IF YOU WANT TO USE SPEECH-TO-TEXT (https://github.com/AarhusAI/whisper-api)
- name: AUDIO_STT_ENGINE
  value: "openai"
- name: AUDIO_STT_OPENAI_API_BASE_URL
  value: "<YOUR_WHISPER_API_URL>"  # e.g., https://whisper.example.com/
- name: AUDIO_STT_OPENAI_API_KEY
  valueFrom:
    secretKeyRef:
      name: openwebui-secrets
      key: STT_API_KEY

# TTS (piper) - OPTIONAL IF YOU WANT TO USE TEXT-TO-SPEECH (https://github.com/AarhusAI/piper-tts)
- name: AUDIO_TTS_ENGINE
  value: "openai"
- name: AUDIO_TTS_OPENAI_API_BASE_URL
  value: "<YOUR_TTS_API_URL>"  # e.g., https://tts.example.com/
- name: AUDIO_TTS_OPENAI_API_KEY
  valueFrom:
    secretKeyRef:
      name: openwebui-secrets
      key: TTS_API_KEY

Replace <YOUR_WHISPER_API_URL> and <YOUR_TTS_API_URL> with your deployed service URLs.

Deployment

Commit and push your changes. Argo will deploy the new version of Open WebUI.