What Happens When a Video Has No Audio Stream in Claude Video

When a video has no audio stream, the Claude Video /watch skill detects the missing stream during metadata extraction, skips the Whisper transcription step to avoid unnecessary API calls, and proceeds with frame extraction and timestamp generation.

The bradautomates/claude-video repository implements a robust video processing pipeline that gracefully handles videos without audio tracks. Instead of failing or attempting to process a non-existent audio stream, the system identifies the absence of audio early in the workflow and adjusts its behavior accordingly.

How Audio Stream Detection Works

The detection mechanism relies on ffprobe metadata analysis performed during the initial video processing stage.

Metadata Extraction in frames.py

In skills/watch/scripts/frames.py, the code analyzes video streams using ffprobe or ffmpeg to determine whether an audio track exists. At line 118, the script evaluates the presence of an audio stream and sets a boolean flag:

has_audio = audio_stream is not None

This has_audio flag becomes a critical piece of metadata that downstream components use to determine whether transcription is possible.

The has_audio Boolean Flag

When audio_stream is None, indicating the video container lacks an audio track, the has_audio variable is set to False. This flag is passed through the processing pipeline to watch.py, where it drives conditional logic for the Whisper API integration.

Conditional Transcription Logic

The orchestration layer in watch.py uses the has_audio flag to determine whether to invoke the Whisper transcription workflow.

Watch Orchestration and Skip Mechanism

At lines 265-267 in skills/watch/scripts/watch.py, the code explicitly checks the audio flag before attempting transcription:

if not has_audio:
    print("[watch] no audio stream found — proceeding without transcription")
    # Skip Whisper processing and continue with frame-only workflow

This check prevents unnecessary API calls to Whisper and avoids potential errors from attempting to extract audio from a silent video file. When the flag is False, the skill bypasses whisper.py entirely and proceeds directly to result assembly.

Processing Outcome for Silent Videos

When processing a video without audio, the system completes the following steps:

1. Downloads the video from the provided URL
2. Extracts frames and generates timestamps normally
3. Omits transcription from the final output
4. Returns frame data without text transcripts

The absence of audio does not affect frame extraction, timestamp generation, or any other visual processing capabilities.

Practical Implementation Examples

The following examples demonstrate the difference between processing videos with and without audio streams:

# Process a video with audio - full transcription enabled

/watch https://example.com/video-with-audio.mp4

# Output includes: frames, timestamps, and Whisper-generated transcript

# Process a video without audio - transcription skipped

/watch https://example.com/video-no-audio.mp4

# Output includes:

#   [watch] no audio stream found — proceeding without transcription

#   Frames and timestamps generated, but no transcript returned

Summary

• Detection occurs early: frames.py identifies missing audio streams at line 118 by checking if audio_stream is not None
• Graceful degradation: Rather than failing, watch.py detects the has_audio=False condition and prints a notification at lines 265-267
• Resource efficiency: The system skips the Whisper API call entirely for silent videos, avoiding unnecessary processing costs and errors
• Functionality preserved: Frame extraction and timestamp generation continue normally regardless of audio presence

Frequently Asked Questions

Does the watch skill fail if a video has no audio stream?

No, the skill does not fail. According to the source code in skills/watch/scripts/watch.py, the system checks the has_audio flag and explicitly handles the missing audio case by printing a notification and continuing with frame extraction. The workflow completes successfully, just without generating a transcript.

How does the system detect whether a video contains audio?

The detection happens in skills/watch/scripts/frames.py at line 118, where the code evaluates audio_stream is not None to set the has_audio boolean. This uses ffprobe or ffmpeg metadata extraction to determine if an audio stream exists in the video container before any processing begins.

Which components are involved when a video has no audio stream?

Three main files handle this scenario: frames.py performs the initial detection and sets the flag, watch.py checks the flag and orchestrates the skip logic, and whisper.py is bypassed entirely. The SKILL.md file defines the /watch command contract that supports this behavior.

Can I still get timestamps for a video without audio?

Yes. The timestamp generation and frame extraction logic operates independently of audio processing. When has_audio is False, only the Whisper transcription step is skipped; all visual analysis including frame extraction and timestamp generation continues unaffected.

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