<h1 id="nin_captioer-local-ai-captioning-krea-2-lora-train-and-test">NIN_Captioer — Local AI Captioning, Krea 2 LoRA Train &amp; Test</h1>NIN_Captioer is a Windows desktop tool for the full Caption → Train → Test loop. Start by preparing LoRA datasets with local Auto Caption and bilingual review, then train a native Krea 2 LoRA and verify results image.Scope: Training is Krea 2 only (train on Krea-2-Raw). Other bases (SD / SDXL / Flux, etc.) are not supported for training. Tested on: Windows x64 + NVIDIA RTX 4090 only. Other GPUs / OS setups are untested.<hr /><h3 id="core-features">Core Features</h3>Auto Natural Language Captioning<ul><li>Seamlessly connects to LM Studio or Ollama via local APIs</li><li>Uses local Vision-LLM models to generate detailed natural language captions suited for Flux / Krea2-style datasets</li><li>Optional WD14 Danbooru tags (ONNX) for tag-style captioning</li><li>Batch Auto Caption for images missing <code>.txt</code>, plus reCaption for the current image</li><li>Caption Analysis: coverage, LoRA Health Score, category distributions</li></ul>Real-Time Translation &amp; Side-by-Side Comparison<ul><li>Instant side-by-side bilingual translation next to English captions</li><li>Review, verify, and fine-tune AI captions without language barriers</li><li>Edits on either side stay in sync (default target: Traditional Chinese)</li></ul>Lora Train (Krea 2 only)<ul><li>Native Krea 2 LoRA trainer in-app (Raw train → Turbo apply / sample)</li><li>Multi-job Start / Stop, progress log, loss chart, sample previews</li><li>Loss spike markers and OOM detection surfaced in the UI and log</li><li>Resource monitor (CPU / RAM / GPU)</li><li>Low VRAM helpers: layer offload, quantization, gradient checkpointing</li><li>Hugging Face model Download / Update, Python path probe and install helper</li><li>LoRA export compatible with ComfyUI key format</li></ul>Lora Test (ComfyUI)<ul><li>Install / start / stop ComfyUI from the app</li><li>Generate with trained Krea 2 checkpoints, multi-prompt runs, and a result gallery</li><li>Smoke-test LoRAs right after training without switching tools</li></ul><hr /><h3 id="suggested-workflow">Suggested workflow</h3><ol><li>Add your image folder → connect LM Studio / Ollama</li><li>Auto Caption → review with translation → Analyze Health Score</li><li>Lora Train (Krea 2 Raw / Turbo) → Start Train</li><li>Lora Test → start ComfyUI → pick checkpoint → generate &amp; review gallery</li></ol><hr /><h3 id="links">Links</h3><ul><li>GitHub Repository: <a target="_blank" rel="ugc" href="https://github.com/BulbulLeung/NIN-Captioer">BulbulLeung/NIN-Captioer</a></li></ul>

Files with SHA256: 089029e70eb306e1fbe176bc57a917ad0644ab382f764ac2618bdfc198a9cddd