Isolate Objects Quickly with peter119lee's ComfyUI-YOLOE26 Tool

By vramkickedin | | 2 min read
A large mask with text next to it that reads YOLO

ComfyUI-YOLOE26 is a custom node pack that segments objects in images using text prompts. Users can type simple descriptions like "person," "car," or "red apple" to isolate objects without needing predefined categories.

The tool was created by developer peter119lee as an alternative to SAM (Segment Anything Model) for specific use cases. It prioritizes speed and efficiency over precision, making it useful for quick iterations and batch processing on consumer hardware.

Speed-focused segmentation features

  • Text-based object segmentation without predefined classes.
  • Real-time processing speed for faster workflows.
  • Lower VRAM usage compared to SAM.
  • Multiple mask output options including per-instance and per-class.
  • Basic mask refinement tools without re-running detection.
  • Structured JSON metadata for automation workflows.

Artists and developers working on batch processing tasks or older hardware may find this tool helpful when perfect edge detection is not critical. It works well for initial dataset preparation or rapid prototyping where waiting for SAM would slow things down.

Developer limitations and comparisons

The developer openly states this is not a SAM replacement and advises users about specific trade-offs. SAM still wins on precision, particularly for edge quality and handling obscure objects. YOLOE-26 also lacks point and box prompting features, limiting it to text-only inputs.

'Before you get too excited: this is NOT a SAM replacement,'

peter119lee writes in the project notes. The edges will not be as clean as SAM on complex objects, and rare items may not detect properly. This is the developer's second node pack, and they welcome feedback on failure cases.

Get ComfyUI-YOLOE26 on GitHub.