Open-Source LoRA Studio: Fully Automated Training from a Single Photo
Ill-Ant-9489 · reddit · 2026-08-02
A developer has released lora-dataset-studio, an open-source local tool designed to simplify the tedious LoRA training pipeline into a one-stop workflow.
- Dataset Generation: Using just a single reference photo, it generates identity-locked multi-angle variations locally via models like Flux-2 Klein or via API.
- Smart Curation: Automatically calculates face-similarity scores, detects and cleans watermarks, blur, and duplicates, supporting libraries of up to 200k images.
- Auto Captioning: Pairs JoyCaption with a local uncensored Ollama vision model for accurate image descriptions and custom trigger words.
- Cloud Training: Offers one-click vast.ai GPU rentals with live monitoring of upload/boot status; automatically stops billing if the machine fails.
- Checkpoint Eval: Built-in fixed-seed grid generation and face scoring; supports combining multiple LoRAs with different weights for side-by-side comparison.
The tool features a one-click Docker install aimed at beginners but allows advanced users to override all parameters.
More from coding & agent
- shadcn Launches Copper: A Mac Scratchpad App Built for AI Workflows — shadcn · 2026-08-02
- SpineCodex: Extending AI Coding Assistant Effective Context to 2.5M Tokens — LingmingZhang · 2026-08-02
- ArkhamMirror: An Open-Source Local AI Platform for Investigative Journalism — tom_doerr · 2026-08-02
- Test: AI Models Hallucinate Missing Tool Parameters Like Emails — eliobldr · 2026-08-02
- Experiment: Independent AI Coding Agents Collaborate in Shared Codebase — Psychological_Arm645 · 2026-08-02
- Replacing Complex Graph Orchestration with OOP: The PyBotchi Intent-Driven Agent Framework — madolid511 · 2026-08-02