video-crop-trim

ComfyUI Video Crop & Trim (FFmpeg)

A ComfyUI custom node for pixel-precise, interactive video cropping and time-range trimming - powered by FFmpeg. Draw your crop region directly on a live video preview inside the node, trim by seconds or frame number, and output both video frames and synchronised audio ready for downstream nodes like VHS Video Combine.

---

Features

Flexible input

• Upload a video directly from the node with the 📁 Upload Your Video button
• Connect a ComfyUI IMAGE tensor from any upstream node

Interactive crop editor

• Live first-frame preview rendered inside the node canvas
• Draw a new crop region by dragging on empty canvas
• Move the region by dragging inside it
• Resize from any corner handle
• Double-click or press Reset Crop to snap back to the full frame
• Rule-of-thirds grid always visible inside the crop region

Crop output preview

• Press 🔍 Preview Crop Output button to fill the canvas with only the cropped region, letterboxed to the correct aspect ratio, so you can verify framing before running the workflow
• Press ⊞ Back to Full Frame button to return to the editor

Playback

• Off-screen <video> element streams the source file directly in the node
• Play / pause button and a scrubbable progress bar with time and frame-number display

Trimming

• none - keep the whole file
• seconds - set trim_start / trim_end in seconds
• frames - set trim_start / trim_end as frame indices

Audio

• strip - discard audio, output video frames only

• extract - trim audio to exactly match the video clip and return it as a standard ComfyUI AUDIO tensor (compatible with VHS Video Combine and other audio nodes)

• The info bar shows the detected audio codec, channel count, and sample rate

Encoding

• Output formats: mp4, webm, avi, mov
• Codecs: libx264, libx265, libvpx-vp9, prores
• CRF quality control for H.264 / H.265 / VP9
• Extra FFmpeg argument passthrough for anything else (-preset fast -tune film -movflags +faststart etc.)

Node resizing

• The preview canvas scales with the node; drag any corner to give yourself more room
• Minimizing node width folds some menus

---

Installation

Required

1. Install ffmpeg if not already installed:

   pip install ffmpeg-python
   

2. Make sure ffmpeg is installed on your system and available on $PATH.

Via ComfyUI Manager (recommended)

Search for Video Crop & Trim and click Install.

Manual

1. Copy the comfyui-video-crop-trim folder into your ComfyUI/custom_nodes/ directory.

cd ComfyUI/custom_nodes
git clone https://github.com/RiverSide71/comfyui-video-crop-trim

2. Restart ComfyUI.

---

Requirements

| Package | Notes | |---|---| | ffmpeg-python | Python bindings for FFmpeg | | ffmpeg | System binary - must be on PATH | | Pillow | Frame extraction / tensor conversion | | numpy | Tensor math | | torch | Provided by ComfyUI |

Install Python dependencies (Only if not already installed):

pip install ffmpeg-python Pillow numpy

Install the FFmpeg system binary if you don't already have it:

# macOS
brew install ffmpeg

# Ubuntu / Debian
sudo apt install ffmpeg

# Windows - download a build from https://ffmpeg.org/download.html
# and add the bin/ folder to your PATH

---

Node reference

The node lives under video/processing in the ComfyUI node menu.

Inputs

| Name | Type | Description | |---|---|---| | crop_x | INT | Left edge of the crop region in pixels (even numbers only) | | crop_y | INT | Top edge of the crop region in pixels (even numbers only) | | crop_width | INT | Width of the crop region in pixels (even numbers only) | | crop_height | INT | Height of the crop region in pixels (even numbers only) | | trim_start | FLOAT | Start of the kept region | | trim_end | FLOAT | End of the kept region (0 = until end of file) | | output_format | dropdown | mp4 / webm / avi / mov | | codec | dropdown | libx264 / libx265 / libvpx-vp9 / prores | | quality | INT | CRF value - lower is better quality / larger file (0–51) | | fps (optional) | FLOAT | Output framerate; auto-detected from file when possible | | extra_ffmpeg_args (optional) | STRING | Raw FFmpeg output flags, e.g. -preset fast -tune film | | audio_mode (optional) | dropdown | strip / extract (default) |

Outputs

| Name | Type | Description | |---|---|---| | frames | IMAGE | Cropped / trimmed video as a (B, H, W, C) float32 tensor | | audio | AUDIO | Trimmed audio as { waveform: Tensor(1, C, T), sample_rate: int } - None when audio_mode is strip or the source has no audio |

---

Usage examples

Basic crop and save

📁 Upload Your Video
        ↓
Video Crop & Trim  →  frames  →  VHS Video Combine  →  saved file

1. Click 📁 Upload Your Video to load your file. The first frame appears in the node canvas automatically.
2. Drag on the canvas to draw your crop region, or adjust the crop_x / crop_y / crop_width / crop_height widgets by hand.
3. Click 🔍 Preview Crop Output to verify framing, then click ⊞ Back to Full Frame to continue editing.
4. Connect frames → VHS Video Combine and queue the prompt.

Crop + trim + audio for VHS Video Combine

Video Crop & Trim
    ├─ frames  →  [images]  VHS Video Combine  →  saved file
    └─ audio   →  [audio]   ↗

1. Set audio_mode to extract.
2. Set trim_mode to seconds and fill in trim_start / trim_end.
3. Connect both frames and audio outputs to VHS Video Combine. The audio is automatically trimmed to match the video segment.

Re-encode footage from an upstream node

Connect any ComfyUI IMAGE tensor (e.g., from an Interpolation or Upscale node) directly to the images input. The tensor is encoded to a temporary file first, then cropped, trimmed, and returned as a new tensor.

---

Keyboard shortcuts / mouse actions (canvas)

| Action | Result | |---|---| | Drag on empty area | Draw a new crop rectangle | | Drag inside crop rect | Move the region | | Drag a corner handle | Resize from that corner | | Double-click | Reset crop to full frame |

---

Tips

• FFmpeg requires even pixel dimensions for most codecs. Odd values are automatically rounded down by one pixel.
• trim_end = 0 means "until the end of the file" regardless of trim mode.
• Extra FFmpeg args override any conflicting built-in flags, so you can force specific encoder settings without touching the source code.
• When connecting an IMAGE tensor via the video input, audio extraction is not available (there is no audio in a tensor). Set audio_mode to strip in this case.
• If you get a widget-order validation error after updating the node, delete the node from your workflow and re-add it - this resets the saved widget slot positions.

---

License

MIT