TUTORIAL

Integrating Groq API with Scribis: Fast Speech-to-Text with Whisper

personJH LAI
calendar_today

Learn how to set up and use Groq API in Scribis for cloud Whisper speech recognition, including API key setup, model selection (whisper-large-v3 / whisper-large-v3-turbo), file limits, and troubleshooting.

If you want to leverage high-speed, transparently priced, and cost-effective cloud speech recognition within Scribis, the Groq API is an ideal choice. Groq provides an OpenAI-compatible Speech-to-Text API running Whisper models under the hood, enabling ultra-fast inference to convert audio or video into accurate transcripts.

This guide provides a comprehensive walkthrough on how to generate and link a Groq API key in Scribis, choose the appropriate Whisper model, and navigate file specifications, pricing, prompt hints, and common troubleshooting steps.

Groq Whisper Core Advantages & Use Cases

Powered by custom LPU (Language Processing Unit) architecture, Groq delivers lightning-fast Whisper inference. It is particularly well-suited for processing pre-recorded long and short media content—such as interviews, online courses, podcasts, meeting notes, and YouTube videos—where rapid transcript delivery is required.

Groq currently offers two primary multilingual speech recognition models:

  • whisper-large-v3-turbo: Focused on extreme speed and cost efficiency, ideal for high-volume daily transcription and fast content extraction.
  • whisper-large-v3: Focused on high accuracy, suited for noisy environments, heavy accents, technical jargon, or formal scenarios demanding top precision.
Model Name Best For Official Pricing Translation API Support
whisper-large-v3-turbo Fast daily transcription, bulk audio, cost efficiency US$0.04 / hour Transcription only (no dedicated translation endpoint)
whisper-large-v3 Complex acoustic environments, multi-speaker, technical proofreading US$0.111 / hour Supports both transcription & translation

All usage and billing are managed via the Groq Console based on actual processed audio duration.

Prerequisites

Before starting configuration, ensure you have the following:

Item Description & Purpose
Scribis Application Download and install the macOS or Windows version from the Scribis Official Releases
Groq Account Used to generate API Keys and monitor usage limits
Groq API Key Credentials authorizing Scribis to invoke Groq's cloud speech recognition service
Test Audio / Video Prepare a 1–2 minute audio file to confirm connection functionality

Security Note: API keys are sensitive credentials. Store them securely in a password manager. Never upload API keys to public GitHub repositories, blog posts, or screenshots.

1. Create a Groq API Key

  1. Log in to Groq Console
    Go to Groq Console and sign in (or register a new account).
  2. Navigate to API Keys
    Open the Groq API Keys management page.
  3. Create New Key
    Click Create API Key, giving it an easily recognizable name (e.g., Scribis-Whisper).
  4. Save the Key
    Copy the generated API Key immediately. The key is displayed in full only once. If lost or compromised, revoke it and create a new key.

2. Configure Groq API in Scribis

  1. Enter API Key
    Open Scribis, navigate to Settings → API, paste your key into the Groq API Key field, and save.
  2. Switch to Cloud ASR Mode
    Return to the main transcription view in Scribis, and in Model Settings, turn off "Use Local ASR" (when enabled, local hardware is used instead of calling the Groq API).
  3. Select Groq Model
    Expand the Cloud ASR Model dropdown to select your desired Groq model:
    • whisper-large-v3-turbo (Groq)
    • whisper-large-v3 (Groq)

If the Groq options do not appear in the dropdown after saving the key, restart Scribis to reload your configuration.

3. Scribis Speech Recognition Model Comparison

Mode / Model Architecture Core Advantage Resource Usage
whisper-large-v3-turbo (Groq) Cloud Groq LPU Ultra-fast, low cost ($0.04/hr) No local compute resource usage
whisper-large-v3 (Groq) Cloud Groq LPU Highest noise reduction & term precision No local compute resource usage
Local ASR (Whisper Local) Local CPU/GPU Works offline, completely free, private Consumes local RAM & hardware performance

4. Transcription Workflow

  1. Load File: In Scribis transcription view, click Local File to import your audio or video file.
  2. Language Settings:
    • Default is Auto Detect.
    • If the audio language is known, manually setting the language (e.g., English, Chinese) reduces latency and improves first-sentence accuracy.
  3. Start Transcription: Confirm the model is set to Groq Whisper and click Start Transcription. Scribis will send the audio to the Groq API and stream back the transcript with timestamps within seconds.

5. File Formats & Audio Constraints

  • Supported Formats: flac, mp3, mp4, mpeg, mpga, m4a, ogg, wav, webm.
  • File Size Limit: Up to 25 MB for Groq Free tier, and 100 MB for Dev tier.
  • Minimum Billing Duration: Minimum charged request duration is 10 seconds (requests shorter than 10 seconds are billed as 10 seconds).
  • Multi-track Notes: For media with multiple audio tracks or languages, the Groq API transcribes the first audio track by default. Verify track settings prior to uploading.

Audio Pre-processing & Compression Tips

If your file exceeds the size limit, converting it to 16 kHz, Mono FLAC is recommended to balance file size and speech recognition quality.

Example FFmpeg conversion command:

ffmpeg -i input.mp3 -ar 16000 -ac 1 -c:a flac output.flac

6. Advanced Configuration

1. ISO Language Codes

When specifying languages manually, Groq uses ISO-639-1 standard codes:

  • Chinese (Traditional / Simplified): zh
  • English: en
  • Japanese: ja
  • Korean: ko
  • Spanish: es

2. Custom Term Prompts

Groq Whisper supports a prompt parameter up to 224 tokens, which can be used to inject context, proper names, or technical terms to prevent model hallucinations.

If your Scribis version supports Prompt settings, you can supply keywords such as:

Scribis, Groq, Whisper Large v3 Turbo, Speech-to-Text, API

3. Timestamps & Subtitle Export

The Groq API returns detailed segment- and word-level timestamps. Once transcription completes, export directly to SRT, VTT, or plain text formats via Scribis.

7. Independent Groq API Verification via curl

If you experience connection issues in Scribis, run this terminal command to test your API key and Groq endpoints directly:

export GROQ_API_KEY='YOUR_GROQ_API_KEY'

curl "https://api.groq.com/openai/v1/audio/transcriptions" \
  -H "Authorization: Bearer ${GROQ_API_KEY}" \
  -F "file=@test_audio.mp3" \
  -F "model=whisper-large-v3-turbo" \
  -F "language=en" \
  -F "response_format=verbose_json"

If curl returns a valid JSON transcript, your API Key is working, indicating the issue may be unsaved configuration in Scribis or Local ASR being enabled.

Frequently Asked Questions

Q1: Getting HTTP 401 or 403 errors during transcription?

This indicates authentication failure. Verify your API key in Scribis, check if the key was deleted in the Groq console, or confirm your account balance status. Generate a new key at Groq API Keys if needed.

Q2: Missing text in multi-track or bilingual video files?

The Groq API currently transcribes only the primary (first) audio stream. Extract or mix the target voice track into a single stream before uploading.

Q3: Does Groq Whisper support Speaker Diarization?

The standard Groq Whisper transcription endpoint outputs text and timestamps without native speaker diarization parameters. Speaker labeling requires post-processing algorithms or external tools within Scribis.

Related Links