Groq Whisper API Setup in OpenTypeless

Configure the built-in Groq Whisper transcription route, verify API access, then test a real recording. The reference below distinguishes released app behavior from the broader Groq API.

Verified provider setup

TL;DR configuration

Checked against OpenTypeless v1.1.61 on 2026-10-08.

OpenTypeless setup summary for stt-groq-whisper
Provider route	Groq Whisper
Settings area	Settings → Speech Recognition
Credential	Groq API key
Request format	HTTPS multipart audio upload
Built-in model	whisper-large-v3-turbo

Built-in speech recognition reference

In Settings → Speech Recognition, select groq-whisper and enter the provider API key. The model, endpoint, and extra fields below are built into OpenTypeless v1.1.61 and are read-only for this provider setup.

config_mode=built_in_stt
select_provider=groq-whisper
enter_api_key=provider_api_key
built_in_model=whisper-large-v3-turbo
built_in_endpoint=https://api.groq.com/openai/v1/audio/transcriptions

Observable success check

With Groq Whisper (groq-whisper) selected, Test sends a small silent WAV to check HTTP access. Success here means HTTP access; a green result does not verify microphone capture, transcript accuracy or insertion. Complete a separate short recording in a disposable text field.

Troubleshooting

The Test button stays disabled.

Likely cause: No Groq API key is available for the groq-whisper provider.

Fix: Enter the Groq key in Speech Recognition settings, then run Test.

Test reports an authentication failure.

Likely cause: The saved credential is not accepted by the Groq transcription route.

Fix: Re-enter the key for groq-whisper and retry without changing the built-in endpoint or model.

Test cannot reach the transcription service.

Likely cause: The network cannot connect to the configured Groq HTTPS endpoint.

Fix: Allow outbound HTTPS to api.groq.com and run the same provider test again.

Official sources and review status

Reviewed by Codex official-source review on 2026-10-08. Recheck by 2026-11-07 because provider endpoints, models, and authentication details can change.

Groq Speech-to-Text documentation
The transcription endpoint, authentication, upload format, and supported model IDs.
Groq API key console
Creation and management of the credential used in OpenTypeless.

Next Docs

Download OpenTypeless

Official references

Groq speech-to-text API

Current endpoint, supported models, limits and API parameters.

https://console.groq.com/docs/speech-to-text

Groq API keys

Create the key used by this provider.

https://console.groq.com/keys

Groq API errors

Interpret the returned HTTP status.

https://console.groq.com/docs/errors

Released OpenTypeless provider configuration

Versioned source for the built-in endpoint and model.

https://github.com/tover0314-w/opentypeless/blob/v1.1.61/src-tauri/src/stt/config.rs

Create and configure the key

Create a Groq API key in your account. Keep it out of public screenshots and issue reports.
In OpenTypeless Settings → Speech Recognition, choose Groq Whisper. Enter the key for that STT provider.
The v1.1.61 built-in route uses whisper-large-v3-turbo at the fixed endpoint below. This route does not expose every Groq API parameter as an app setting.
Run Test, then make a separate short recording containing an actual sentence. The released connection test uploads a small silent WAV and checks HTTP success; it does not test your microphone, vocabulary or text insertion.

Released built-in request

The recording is buffered, converted to WAV, and sent after recording ends. A selected language is sent when it is not “multi”. The built-in extra-field list is empty; this route does not send the custom dictionary as the Groq API prompt.

Provider: groq-whisper
Endpoint: https://api.groq.com/openai/v1/audio/transcriptions
Method: POST multipart/form-data
Authorization: Bearer <GROQ_API_KEY>
Model: whisper-large-v3-turbo
Audio field: file (WAV)
Language: selected language, except multi

API capabilities and app settings

Groq supports additional transcription parameters and other model choices in its API. Consult its current documentation for account limits and billing. Those capabilities do not imply a model switch, arbitrary upload, timestamp export, streaming partial transcript, or prompt field exists in the built-in OpenTypeless route. Its speech buffer is capped at 24 MiB in this release, and the transcription request occurs on disconnect.

Verify a complete dictation

Focus a disposable text document and record a non-sensitive sentence. Stop recording and wait for transcription.
Check the transcript’s names, numbers, language and negation. Run the same sentence again if you are evaluating a setting.
If AI Polish is enabled, its separate LLM can edit the transcript. A Groq STT key does not by itself configure the LLM or make the two processing routes local.
Confirm the destination field and undo. Record correction and complete response time; a successful connection test alone does not establish quality or end-to-end latency.

Troubleshooting by stage

Test is disabled or authentication fails.

Check that the key is saved for Groq Whisper. For a 401 response, confirm the credential in Groq; do not repeatedly change the fixed endpoint or model.

The service returns 429.

Inspect the account’s current rate or usage limits and the returned error. Wait according to the service response before another request; repeated rapid tests do not resolve an account limit.

A request is invalid or audio is too large.

Check the error response and the current Groq upload rules. OpenTypeless buffers short dictation; it does not provide a general large-file chunking workflow for this built-in provider.

Test passes, but my recording is silent or inaccurate.

The Test request is a connection check. Check microphone permission and input selection with a real phrase, then compare language, vocabulary and the transcript.

Transcription works but the final draft or insertion fails.

Check the LLM polishing stage and the focused destination separately. Verify identifiers after polishing and test insertion and undo in a short disposable draft.

Continue to the workflow

Groq dictation workflow

Test a real recording, language, correction effort and insertion.

/blog/groq-whisper-setup

Provider index

Keep STT and LLM configuration separate.

/docs/provider-index

Custom dictionary

Use vocabulary hints with the correct processing-stage boundary.

/docs/custom-dictionary