Comfyui_TTP_Toolset  by TTPlanetPig

Semantic tiling and upscaling toolkit for ComfyUI

Created 1 year ago
1,012 stars

Top 36.2% on SourcePulse

GitHubView on GitHub
Project Summary

Summary

This ComfyUI toolset addresses advanced image tiling for detailed img2img upscaling and modification. It targets ComfyUI users seeking greater control over high-resolution image generation by enabling object-aware, variable-size tiling, thereby overcoming limitations of traditional fixed-grid approaches. The primary benefit is enhanced detail and precision in complex image manipulation workflows.

How It Works

The core innovation is Smart Tile 2.0, an object-aware tiled img2img workflow. Instead of fixed grids, it allows users to define variable-size tiles around specific visual structures like faces, text, or subjects. This approach processes one tile at a time through the sampler, enabling granular detail enhancement where needed. The workflow supports automatic tile generation, per-tile prompts, and sophisticated assembly with mask-aware blending, semantic priority, and color correction, ensuring critical details are preserved and enhanced.

Quick Start & Requirements

This repository provides custom nodes for ComfyUI. Installation typically involves placing the repository's contents into the ComfyUI custom_nodes directory. Key requirements include a working ComfyUI installation and specific AI models (e.g., SAM3/SAM3.1 for auto-tiling, QwenVL3 for object detection, diffusion models, VAEs, text encoders). Example workflows, such as examples/smart_tile_qwen_sam_loop_example_v2.json, are provided for guidance.

Highlighted Details

  • Smart Tile 2.0: A primary feature offering object-aware, variable-size tiling for detailed img2img upscaling. It supports automatic layout generation, per-tile prompts, loop processing, semantic priority, mask-aware assembly, and final-only saving.
  • Visual Tile Editor: The TTP Smart Tile Interactive Crop node enables manual, painted, or auto-detected tiling with interactive layout management and saving.
  • TeaCache for Hunyuan Video: Integrates TeaCache for accelerated video sampling in Hunyuan Video workflows, offering significant speedups (e.g., ~x2.1 on NVIDIA 4090).
  • Coordinate and Conditioning Utilities: A suite of helper nodes for managing tile-specific data, coordinates, and conditioning batches.

Licensing & Compatibility

The provided README does not specify a software license. This absence may present compatibility concerns for commercial use or integration into closed-source projects.

Limitations & Caveats

The TTP Tile 1.0 is designated as a legacy workflow. Acceleration via TeaCache for Hunyuan Video may require careful tuning to avoid impacting image quality or motion. Auto-tiling features necessitate the correct installation and connection of specific AI models. The lack of explicit licensing information is a notable caveat for adoption.

Health Check
Last Commit

4 days ago

Responsiveness

Inactive

Pull Requests (30d)
0
Issues (30d)
0
Star History
22 stars in the last 30 days

Explore Similar Projects

Feedback? Help us improve.