Groq Core Workflow B: Audio, Vision & Speech

Groq Core Workflow B: Audio, Vision & Speech

Overview

Beyond chat completions, Groq provides ultra-fast audio transcription (Whisper at 216x real-time), multimodal vision (Llama 4 Scout/​Maverick), and text-to-speech. These endpoints use the same groq-sdk client.

Prerequisites

• groq-sdk installed, GROQ_API_KEY set
• For audio: audio files in supported formats
• For vision: image URLs or base64 images

Audio Models

Supported audio formats: flac, mp3, mp4, mpeg, mpga, m4a, ogg, wav, webm

Instructions

Step 1: Audio Transcription (Whisper)

import Groq from "groq-sdk";
import fs from "fs";

const groq = new Groq();

// Transcribe audio file
async function transcribe(filePath: string): Promise<string> {
  const transcription = await groq.audio.transcriptions.create({
    file: fs.createReadStream(filePath),
    model: "whisper-large-v3-turbo",
    response_format: "json",        // or "text" or "verbose_json"
    language: "en",                  // Optional: ISO 639-1 code
  });

  return transcription.text;
}

// With timestamps (verbose mode)
async function transcribeWithTimestamps(filePath: string) {
  const transcription = await groq.audio.transcriptions.create({
    file: fs.createReadStream(filePath),
    model: "whisper-large-v3-turbo",
    response_format: "verbose_json",
    timestamp_granularities: ["segment"],
  });

  return transcription;
  // Returns segments with start/​end times
}

Step 2: Audio Translation (to English)

// Translate any language audio to English text
async function translateAudio(filePath: string): Promise<string> {
  const translation = await groq.audio.translations.create({
    file: fs.createReadStream(filePath),
    model: "whisper-large-v3",
  });

  return translation.text;
}

Step 3: Vision (Image Understanding)

// Analyze images with Llama 4 Scout (up to 5 images per request)
async function analyzeImage(imageUrl: string, question: string) {
  const completion = await groq.chat.completions.create({
    model: "meta-llama/​llama-4-scout-17b-16e-instruct",
    messages: [
      {
        role: "user",
        content: [
          { type: "text", text: question },
          { type: "image_url", image_url: { url: imageUrl } },
        ],
      },
    ],
    max_tokens: 1024,
  });

  return completion.choices[0].message.content;
}

// Multiple images
async function compareImages(urls: string[], prompt: string) {
  const imageContent = urls.map((url) => ({
    type: "image_url" as const,
    image_url: { url },
  }));

  const completion = await groq.chat.completions.create({
    model: "meta-llama/​llama-4-scout-17b-16e-instruct",
    messages: [{
      role: "user",
      content: [{ type: "text", text: prompt }, ...imageContent],
    }],
    max_tokens: 2048,
  });

  return completion.choices[0].message.content;
}

// Base64 image input
async function analyzeBase64Image(base64Data: string) {
  return groq.chat.completions.create({
    model: "meta-llama/​llama-4-scout-17b-16e-instruct",
    messages: [{
      role: "user",
      content: [
        { type: "text", text: "Describe this image in detail." },
        {
          type: "image_url",
          image_url: { url: `data:image/​jpeg;base64,${base64Data}` },
        },
      ],
    }],
  });
}

Step 4: Text-to-Speech

// Generate speech from text
async function textToSpeech(text: string, outputPath: string) {
  const response = await groq.audio.speech.create({
    model: "playai-tts",          // or "playai-tts-arabic"
    input: text,
    voice: "Arista-PlayAI",      // See Groq docs for voice options
    response_format: "wav",       // wav, mp3, flac, opus, aac
  });

  const buffer = Buffer.from(await response.arrayBuffer());
  fs.writeFileSync(outputPath, buffer);
  console.log(`Audio saved to ${outputPath}`);
}

Step 5: Python Audio Transcription

from groq import Groq

client = Groq()

# Transcribe
with open("audio.mp3", "rb") as file:
    transcription = client.audio.transcriptions.create(
        file=("audio.mp3", file),
        model="whisper-large-v3-turbo",
        response_format="verbose_json",
    )
    print(transcription.text)
    for segment in transcription.segments:
        print(f"[{segment.start:.1f}s - {segment.end:.1f}s] {segment.text}")

Step 6: Model Benchmarking

// Compare models on same prompt for speed vs quality
async function benchmarkModels(prompt: string) {
  const models = [
    "llama-3.1-8b-instant",
    "llama-3.3-70b-versatile",
    "llama-3.3-70b-specdec",
  ];

  for (const model of models) {
    const start = performance.now();
    const result = await groq.chat.completions.create({
      model,
      messages: [{ role: "user", content: prompt }],
      max_tokens: 200,
    });
    const elapsed = performance.now() - start;
    const tps = result.usage!.completion_tokens / ((result.usage as any).completion_time || 1);

    console.log(
      `${model.padEnd(45)} | ${elapsed.toFixed(0)}ms | ${tps.toFixed(0)} tok/​s | ${result.usage!.total_tokens} tokens`
    );
  }
}

Vision Model Limits

• Maximum 5 images per request
• Supported formats: JPEG, PNG, GIF, WebP
• Images fetched from URL or embedded as base64
• Vision models also support tool use, JSON mode, and streaming

Error Handling

Resources

• Groq Speech-to-Text
• Groq Text-to-Speech
• Groq Vision
• Groq Models

Next Steps

For common errors and troubleshooting, see groq-common-errors.