Claude Video Whisper API Preference: Why Groq Wins Over OpenAI

Claude Video prefers the Groq Whisper API by default, automatically selecting Groq's whisper-large-v3 model for faster, cheaper transcription while only falling back to OpenAI's whisper-1 when explicitly forced or when no Groq API key is present.

Claude Video is an open-source video analysis tool that transcribes audio using external Whisper APIs. According to the source code in the bradautomates/claude-video repository, the application implements a strict hierarchy for its Claude Video Whisper API integration that prioritizes cost-effectiveness and speed.

Why Groq is the Preferred Backend

The developers explicitly designed Claude Video to prioritize Groq as the default transcription service. In skills/watch/scripts/setup.py, the installation script notes that "Groq is preferred: it runs whisper-large-v3 at a fraction of OpenAI's price".

This preference is reinforced in the user-facing documentation at skills/watch/SKILL.md, which states that the default backend is Groq (whisper-large-v3) because it is cheaper and faster than alternatives. OpenAI's whisper-1 serves strictly as a fallback mechanism when Groq credentials are absent or when users manually override the configuration.

Backend Resolution Logic in watch.py

The backend selection logic resides in skills/watch/scripts/watch.py. When processing a video file, the application calls load_api_key(args.whisper) to determine which provider to use.

backend, api_key = load_api_key(args.whisper)          # selects Groq when both keys exist

...
response = _post_whisper(GROQ_ENDPOINT, api_key, GROQ_MODEL, audio_path)   # Groq path

response = _post_whisper(OPENAI_ENDPOINT, api_key, OPENAI_MODEL, audio_path)   # OpenAI path

The load_api_key function automatically returns Groq credentials when both GROQ_API_KEY and OPENAI_API_KEY environment variables are present. This ensures that Groq receives priority without requiring explicit user intervention.

CLI Options for Forcing a Specific Provider

Users can override the default Claude Video Whisper API selection using command-line flags. The --whisper argument accepts specific values to force a particular backend or disable transcription entirely.

# Automatic (preferred) selection – uses Groq if a GROQ_API_KEY is present,

# otherwise falls back to OpenAI if an OPENAI_API_KEY is present.

watch video.mp4

# Force the Groq backend explicitly

watch video.mp4 --whisper groq

# Force the OpenAI backend explicitly

watch video.mp4 --whisper openai

# Disable Whisper entirely (frames-only mode)

watch video.mp4 --no-whisper

When you explicitly specify --whisper openai, the application bypasses the Groq preference logic and routes requests directly to OpenAI's whisper-1 endpoint.

Model Specifications and Constants

The specific model versions are defined in skills/watch/scripts/whisper.py. Groq utilizes the whisper-large-v3 model, while OpenAI uses the standard whisper-1 model.

• Groq: whisper-large-v3 via GROQ_ENDPOINT
• OpenAI: whisper-1 via OPENAI_ENDPOINT

These constants feed into the _post_whisper function, which handles the HTTP POST requests to the respective API endpoints using the resolved API key.

Summary

• Claude Video Whisper API integration prioritizes Groq over OpenAI by default.
• The selection logic in watch.py automatically chooses Groq when both API keys are available.
• Groq provides access to whisper-large-v3 at lower cost and higher speed than OpenAI's whisper-1.
• Users can force OpenAI using --whisper openai or disable audio processing with --no-whisper.
• Fallback to OpenAI occurs only when no Groq key exists or when explicitly requested.

Frequently Asked Questions

Which Whisper API does Claude Video use by default?

Claude Video uses the Groq Whisper API by default. When both GROQ_API_KEY and OPENAI_API_KEY environment variables are present, the load_api_key function in watch.py automatically selects Groq's whisper-large-v3 model.

How do I force Claude Video to use OpenAI instead of Groq?

Pass the --whisper openai flag when running the watch command. This bypasses the automatic preference logic and routes transcription requests to OpenAI's whisper-1 endpoint instead of Groq.

What Whisper models does Claude Video support?

According to skills/watch/scripts/whisper.py, Claude Video supports whisper-large-v3 via Groq and whisper-1 via OpenAI. These model constants are hardcoded in the source and passed to the respective API endpoints.

Why does Claude Video prefer Groq over OpenAI for transcription?

The developers chose Groq as the preferred backend because it runs the whisper-large-v3 model at a fraction of OpenAI's price while delivering faster inference speeds. This cost-performance advantage is documented in both setup.py and SKILL.md.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}