perfectgf/lora-dataset-studio

Self-hosted, one-tab workbench for the whole training lifecycle: build Character/Concept/Style datasets (generate, scrape, or triage a big dump in the Image Bank), curate, caption, clean watermarks, then train LoRAs on five model families or a full Krea 2 model on a rented GPU, and rank checkpoints in a Test Studio. Free & local.

What it solves

LoRA Dataset Studio provides a unified browser-based interface for the entire lifecycle of creating image datasets for LoRA training. It eliminates the need to jump between multiple disparate tools for collecting, cleaning, captioning, and training images, streamlining the process of teaching AI models specific characters, concepts, or styles.

How it works

The software acts as a central hub that integrates several specialized AI tools. It runs locally on the user's machine and leverages ai-toolkit for the actual training process and ComfyUI for local image generation and checkpoint testing. It organizes the workflow into several stages:

  • Dataset Management: Importing images, organizing them into banks, and curating them through cropping, mirroring, and quality scoring.
  • Captioning: Using local vision models or JoyCaption to generate prose or tags for images.
  • Training: Providing guided recipes for various model families (such as SDXL, FLUX.1, and Qwen-Image 2.1) and managing training queues.
  • Review: Comparing different checkpoints using fixed seeds and prompt batches to evaluate the quality of the trained LoRA.

Who it’s for

It is designed for AI artists and developers who want to create high-quality LoRAs for image generation models without having to manage complex command-line tools or manual file organization.

Highlights

  • End-to-End Workflow: Covers everything from image bank curation and text detection to training and final testing.
  • Extensible Plugin System: Offers optional plugins for cloud training (via vast.ai), web scraping, and publishing to Civitai or Hugging Face.
  • Multi-Model Support: Supports a wide range of training families including FLUX.1, SDXL, and Krea 2.
  • Integrated Testing: Includes a "Studio" for comparing checkpoints and strengths with lineage and ranking tools.

Related

  • Project
    no8d/ComfyUI-NO8D-controls

    ComfyUI‑NO8D‑controls is a set of custom ComfyUI nodes that streamline prompt handling, LoRA stacking, image loading, generation, compositing, and dataset saving. It includes a node that calls an external LLM (OpenAI‑compatible or Ollama) to generate prompts from text or images, plus visual tools for managing prompt libraries and building image grids.

  • Project
    egotoadcitadel/palette-from-image-ai

    Palette‑From‑Image‑AI is a portable Windows desktop app that offline‑extracts region‑aware, perceptually clustered colour palettes from any image. It uses an on‑device vision model, CIEDE2000 clustering, role tagging, harmony modes, and offers 12 export formats (HEX, CSS, ASE, JSON, PNG sheet, etc.). Features include batch processing, palette diff, swatch locking, GPU acceleration, CLI, and watch‑folder automation. No installer, no telemetry, MIT‑licensed.

  • Project
    nihui/qwenimage-ncnn-vulkan

    A portable ncnn Vulkan implementation of Qwen-Image-2.1 for high-performance text-to-image generation and image editing across various GPUs and OSs.

  • Project
    six-nut/PocketMen-with-you

    A local generative tool that transforms two or more photos into consistent, animated character sprites for Codex companions using open-weight image models.