Home Assistant: Whisper API Integration for Speech-to-Text

Integration works for Assist pipelines.

Requirements:

• Installed Whisper.cpp server in network

Server setup (whisper.cpp):

1. Git pull whisper.cpp
2. Build the server (example for CUDA):

cmake -B build -DGGML_CUDA=1
cmake --build build --config Release

3. Run the server:

./ai/whisper.cpp/build/bin/whisper-server -m /ai/models/whisper/ggml-large-v3-turbo-q8_0.bin --host 192.168.0.55 --port 5045 -l ru  -sow -sns --vad --vad-model /ai/models/whisper/ggml-silero-v6.2.0.bin --inference-path /v1/audio/transcriptions

Configuration:

Configuration is done in the UI (YAML is no longer supported):

1. Install via HACS (or copy custom_components/whisper_api_stt to your config directory) and restart Home Assistant.
2. Settings → Devices & services → Add integration → Whisper.cpp API.
3. Fill in the form. Settings can be changed later via the integration's Configure button; the entry reloads automatically.

You can add several entries (e.g. different languages or servers).

Upgrading from the YAML version: remove the stt: - platform: whisper_api_stt block from configuration.yaml, restart, and add the integration through the UI.

Parameters:

• Server URL: full URL of your whisper.cpp or OpenAI-compatible endpoint, e.g. http://192.168.0.55:5045/v1/audio/transcriptions.
• API key (optional): API key if required by your server.
• Model: model name. Defaults to whisper-1 (whisper.cpp server ignores it).
• Language: language code (e.g., en-US, ru-RU).
• Prompt (optional): text to guide the model's style.
• Temperature: sampling temperature between 0 and 1. Defaults to 0.0.

The form checks that the server is reachable before saving.

Notes:

• The integration converts the language code to ISO-639-1 (e.g., ru-RU -> ru) for API compatibility.
• Ensure your server_url includes the full path to the endpoint, e.g., /v1/audio/transcriptions.

Used sources + thanks to:

• sfortis/openai_tts: https://github.com/sfortis/openai_tts