Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/HUANGCHIHHUNGLeo/claude-real-video/llms.txt

Use this file to discover all available pages before exploring further.

claude-real-video has two install tiers. The core install gives you scene-aware frame extraction and deduplication — everything you need to feed a video’s key frames to an LLM. The full install adds openai-whisper so the tool can also transcribe the audio and write a plain-text transcript alongside the frames.
Python 3.10 or later is required. Check your version with python --version before continuing.

Install the package

pip install claude-real-video
The core install pulls in yt-dlp (for URL sources) and Pillow (for pixel-level deduplication). The [whisper] extra additionally installs openai-whisper.

Install ffmpeg

ffmpeg and ffprobe are used for frame extraction and audio processing. They are not pip-installable — you must install them through your system’s package manager and ensure they are on your PATH.
The easiest path is Homebrew:
brew install ffmpeg

Verify ffmpeg is on your PATH

After installation, confirm ffmpeg can be found:
ffmpeg -version
You should see a version string like ffmpeg version 7.x .... If the command is not found, check that the bin\ folder (Windows) or the install path (macOS/Linux) is in your PATH.

Whisper

Transcription uses the whisper CLI, which is installed automatically by the [whisper] extra. You can also install it separately:
pip install openai-whisper
Whisper also relies on ffmpeg to extract audio from video files, so the ffmpeg installation above is a prerequisite for transcription as well. If whisper is not found on your PATH, claude-real-video skips transcription cleanly and notes the reason in MANIFEST.txt.

Verify the install

1

Run the help command

crv --help
You should see the full usage output listing all flags and their defaults.
2

Alternative entry point

The module entry point works as an alias if crv is not in your shell’s PATH (for example, in some virtual-environment setups):
python -m claude_real_video --help

Supported platforms

claude-real-video is tested on macOS, Linux, and Windows.

Build docs developers (and LLMs) love