Griptape Audio Transcription Driver: Groq

This is the node that turns Groq's absurdly fast Whisper endpoints into a Griptape audio transcription driver. If you've got audio that needs turning into text - a voice memo, a lecture, a video's speech track - and you don't want to run Whisper locally on your already-busy GPU, this is the cheap, fast path. Groq's transcription is famously quick, the free tier is generous, and you get it all through one small node.

How it works

The node constructs an OpenAiAudioTranscriptionDriver and points it at Groq's OpenAI-compatible endpoint (https://api.groq.com/openai/v1/) instead of OpenAI's. Groq serves Whisper models on that endpoint, so the pack reuses OpenAI's transcription driver plumbing with a swapped base URL - a tidy trick. The result comes out the DRIVER output as an AUDIO_TRANSCRIPTION_DRIVER type.

That driver is what you feed into the "Griptape Audio Transcription Task" node's driver input (the task node falls back to whatever the agent's config has if you don't provide one). It's also what you'd use in an AudioTranscriptionClient tool if your agent is doing transcription as a tool call.

The inputs that matter

There are exactly two, both optional:

• audio_transcription_model (default whisper-large-v3-turbo, with distil-whisper-large-v3-en and whisper-large-v3 as alternatives) - Groq only serves these three Whisper flavors, so the dropdown is intentionally short.
• api_key_env_var (default GROQ_API_KEY) - the env var / settings key name holding your key. Get a free one at https://console.groq.com/keys.

That's the whole node. No model files to download, no local inference - all transcription happens on Groq's servers.

Installing it

It ships in the one pack:

cd ComfyUI/custom_nodes
git clone https://github.com/griptape-ai/ComfyUI-Griptape

or ComfyUI Manager → search "Griptape". Restart, then set your GROQ_API_KEY in Settings → Griptape (or as an env var). The pack's griptape[all] dependency drags in a torch build that can occasionally fight your ComfyUI install - the README's troubleshooting section covers the standard fix.

Common issues

• [WARNING]: GROQ_API_KEY is not set - the universal pack key warning; check Settings → Griptape and that the env var name matches.
• "Model not found" from the API - you picked a model Groq doesn't serve. Stick to the three in the dropdown.
• No driver wired anywhere - remember the driver only does something when it's connected to an Audio Transcription Task or an agent that uses transcription. A bare driver floating in the graph just sits there.

Worth knowing: if you're already using the Groq Drivers config for your agent, that config bundles an audio transcription driver too - you only need this standalone node when you want the driver separate (e.g. to expand, swap, or feed a specific task).