Speed in practice: fast batch + Whisper voice-to-text

USE 0 - 10 min

Transcribe a voice recording with Whisper

Groq hosts Whisper large-v3 alongside its text models, on the same fast hardware. A one-minute audio file comes back as text in roughly 2 seconds. This is the voice-to-lab-notes pattern used later in the n8n sessions.

Log into n8n.32dots.de with the email and password you received when you signed up. Will be live on session day

1 Record a 30-second voice note on your phone or laptop — a quick verbal summary of a paper you read, for example. Export as .mp3 or .wav.
2 Send it to the Whisper endpoint. Replace YOUR_KEY and your filename: ` curl https://api.groq.com/openai/v1/audio/transcriptions \ -H 'Authorization: Bearer YOUR_KEY' \ -F 'file=@your_note.mp3' \ -F 'model=whisper-large-v3' `
3 Read the transcript in the response. Note how short the round-trip felt — the model is whisper-large-v3, the same model that takes 20+ seconds on a laptop CPU.
4 Try it in the Groq playground too. console.groq.com has an Audio tab — drag the file in, same model, no curl needed.

✓

You get back a JSON object with a 'text' field containing your spoken words, accurately transcribed.

BUILD 10 - 20 min

Batch-summarise 5 paper abstracts

The real payoff of Groq's speed is batch work: processing many short texts quickly. Here you loop over 5 abstracts and collect a one-sentence summary for each — the kind of literature-triage task that normally blocks on API latency.

Your task

Write a short Python script (or an n8n Loop node) that sends 5 PubMed abstracts to llama-3.3-70b-versatile one by one and collects a one-sentence summary for each.

1 Collect 5 abstracts. Search PubMed for a topic you know (e.g. 'CRISPR off-target effects'). Copy 5 abstract texts into a list in your script.
2 Use the openai Python SDK pointed at Groq. Install once with pip install openai, then: `python from openai import OpenAI client = OpenAI( base_url='https://api.groq.com/openai/v1', api_key='YOUR_KEY' ) for abstract in abstracts: r = client.chat.completions.create( model='llama-3.3-70b-versatile', messages=[{'role':'user','content':f'One sentence summary: {abstract}'}] ) print(r.choices[0].message.content) `
3 Run it and time it. All 5 calls should complete in under 10 seconds total — note the contrast with GPT-4o free tier, which would queue them.
4 Optional n8n path. Use a Loop Over Items node feeding an HTTP Request node (POST to https://api.groq.com/openai/v1/chat/completions, Bearer auth, JSON body). Same result, no Python needed.

Deliverable

A list of 5 one-sentence summaries printed to your terminal (or shown in n8n's output panel), with a rough total runtime.