VTrim

VTrim is a lightweight, efficient video analysis and trimming tool. It automatically finds segments containing people and can output a trimmed video instantly—without re-encoding, preserving original quality at blazing speed.

• ⚡ Lossless • 🎥 Professional edit-ready XML

Installation

Install via pip:

pip install vtrim

Quick Start

Trim Video Directly

vtrim --input your_video.mp4 --detect-human --output output.mp4

• Uses FFmpeg stream copy (-c copy) → no re-encoding, no quality loss.
• Automatically merges nearby detections and adds padding for smooth transitions.

Export Edit Timeline to DaVinci Resolve / Premiere Pro

Preserve the full timeline (including gaps) as an FCP7 XML for professional editing:

vtrim --input your_video.mp4 --detect-human --export-xml timeline.xml

Audio and video are perfectly synchronized and split per segment.

You can combine both outputs:

vtrim --input your_video.mp4 \
      --detect-human \
      --output output.mp4 \
      --export-xml timeline.xml

Get Raw Detection Results (JSON)

Print detected time segments to stdout for scripting or integration:

vtrim --input meeting.mp4 --detect-human

Output:

{
  "segments": [
    { "start": 2.3, "end": 5.8 },
    { "start": 10.1, "end": 14.7 }
  ]
}

Segments are in seconds, ready for automation (e.g., with Python, Node.js, or shell scripts).

Full Option Reference

Parameter	Required	Description
--input	Yes	Path to the input video file (e.g., video.mp4).
--detect-human	No*	Enable human detection. If omitted, the tool returns an empty segment list and does nothing. (Required if you want analysis or trimming)
--output	No	Path to save the trimmed video (e.g., trimmed.mp4). If not provided, only JSON results are printed to stdout.
--export-xml	No	Path to save the FCP7 XML.
--conf-threshold	No	Confidence threshold for person detection (0.0–1.0). Lower = more sensitive. Default: 0.25.
--padding	No	Seconds of padding added before/after each detected segment. Default: 1.
--gap-tolerance	No	Maximum gap (in seconds) between detections to merge into one segment. Default: 4.

📌 Important: --detect-human must be specified to perform any analysis. Without it, the output will always be {"segments": []}.

Notes

• The underlying model is YOLOv8n (ONNX format), optimized for CPU inference.
• Video trimming uses FFmpeg stream copy (-c copy), so it’s fast and lossless—no quality degradation.
• Progress updates are printed to stderr during analysis (every 5% for known-length videos).
• For automation, set the environment variable ANALYZER_PROGRESS_JSON=1 to receive machine-readable progress messages on stderr.