Whisper Cloud for Home Assistant (Marketplace integration)

Installs: 753 sharing analytics
GitHub stars: 111
Connection: Uses the vendor's cloud
Upkeep: Updated in the last 6 months
Version: 3.1.1, October 26, 2026
Category: AI and voice, AI models
Kind: Marketplace integration
Brand: OpenAI
Needs: Needs the Marketplace (HACS on 2026.10 and older)
Domain: openai_whisper_cloud
Source: github.com/fabio-garavini/ha-openai-whisper-stt-api
Issues: Report an issue

Independent directory, made by the developer of Vulpo. Not affiliated with Home Assistant or the Open Home Foundation.

About

Whisper Cloud is a custom integration that sends Assist speech-to-text audio to a cloud transcription API. It supports OpenAI, GroqCloud, Mistral AI, Cohere and Xiaomi MiMo, plus a custom endpoint for any OpenAI-compatible transcription server such as whisper-asr-webservice or faster-whisper-server. Moving transcription off the local machine reduces the workload on the Home Assistant server. Each provider entry has settings for model, temperature and an optional prompt, though providers that do not accept these hide them. An option can upload audio while it is still being recorded to lower latency, and the integration falls back to a buffered upload if the provider refuses streaming. Setup is done in the UI with an API key for the chosen provider.

Automatic summary of the README and code of fabio-garavini/ha-openai-whisper-stt-api. Something wrong, or are you the maintainer and want your own text here? Get in touch

Supports

Uses the vendor's cloud Needs an internet connection and an account with the vendor.
From the Marketplace Installed through the Marketplace, formerly HACS.
Actively maintained Its author released or pushed a change in the last six months.

Setting it up

Install it from the Marketplace and restart Home Assistant, then add it under Settings, Devices and services.

Whisper Cloud

API Provider

Provider setup

Name
API keyAPI key for the selected provider.
Model
Temperature
Prompt (Optional)
Stream audio while it is recorded (lower latency)Upload the audio to the provider while it is still being recorded instead of after the recording completes. If the provider does not accept streamed uploads, the integration automatically falls back to a buffered upload and remembers not to try again.

Custom endpoint

Name
URLYour server URL. A URL without a path gets the standard endpoint appended (https://host and https://host/v1 both become https://host/v1/audio/transcriptions); any URL that already includes a path is used as-is, e.g. http://host:8000/asr (whisper-asr-webservice).
API key (Optional)Leave empty if your server does not require an API key.
ModelModel name accepted by your server (e.g. whisper-1, Systran/faster-whisper-large-v3).
Temperature
Prompt (Optional)
Audio file field nameMultipart field used to upload the audio. Most OpenAI-compatible servers use file; whisper-asr-webservice needs audio_file. If unsure, keep file: the integration detects and retries automatically on a 422 response.
Stream audio while it is recorded (lower latency)Upload the audio to the provider while it is still being recorded instead of after the recording completes. If the provider does not accept streamed uploads, the integration automatically falls back to a buffered upload and remembers not to try again.

Reconfigure

Name
URLYour server URL. A URL without a path gets the standard endpoint appended (https://host and https://host/v1 both become https://host/v1/audio/transcriptions); any URL that already includes a path is used as-is, e.g. http://host:8000/asr (whisper-asr-webservice).
API key (leave blank to keep the current one)Leave blank to keep the current API key.
Model
Temperature
Prompt (Optional)
Audio file field nameMultipart field used to upload the audio. Most OpenAI-compatible servers use file; whisper-asr-webservice needs audio_file. If unsure, keep file: the integration detects and retries automatically on a 422 response.
Stream audio while it is recorded (lower latency)Upload the audio to the provider while it is still being recorded instead of after the recording completes. If the provider does not accept streamed uploads, the integration automatically falls back to a buffered upload and remembers not to try again.

After setup

Model
Temperature
Prompt (Optional)
Stream audio while it is recorded (lower latency)Upload the audio to the provider while it is still being recorded instead of after the recording completes. If the provider does not accept streamed uploads, the integration automatically falls back to a buffered upload and remembers not to try again.

Texts from the code of fabio-garavini/ha-openai-whisper-stt-api.

Troubleshooting

DiagnosticsDownloads a diagnostics file to attach to a bug report.

From the README

This HA custom integration lets you use any compatible OpenAI API (OpenAI, GroqCloud, Mistral AI, others coming...) for computing speech-to-text in cloud, reducing workload on Home Assistant server.

From the README of fabio-garavini/ha-openai-whisper-stt-api, GNU AGPL v3.0. Shortened.