Comfyui Lora Training
Walks you through the full pipeline for training character LoRAs: prepping 15-30 image datasets with proper captioning, configuring training tools like AI-Toolkit for FLUX or Kohya_ss for SDXL, monitoring checkpoints, and evaluating results. The guidance is opinionated where it matters, like preprocessing 3D renders with style variations to avoid baking in that aesthetic, or using unique trigger words instead of generic descriptors. It covers both high-VRAM and low-VRAM setups, includes a testing protocol to catch overfitting, and shows how to combine trained LoRAs with zero-shot methods like PuLID for the best consistency. Use this when you need repeatable character generation across a series rather than one-off shots.
- skills.sh installs
- 587
- GitHub stars
- 94
- last updated
- Jun 3, 2026
Install
$ npx -y skills add mckruz/comfyui-expert --skill comfyui-lora-training --agent claude-codeInstalls into .claude/skills of the current project.
Files
ComfyUI LoRA Training
Guide the user through dataset preparation, training configuration, and evaluation for character LoRAs.
When to Train vs Zero-Shot
| Scenario | Recommendation |
|---|---|
| Need absolute consistency across many images | Train LoRA |
| Building a character series or ongoing project | Train LoRA |
| Quick one-off generation | Use zero-shot (InstantID/PuLID) |
| Limited references (1-5 images) | Use zero-shot |
| Testing concepts | Use zero-shot first, train if committing |
Training Pipeline
Sponsors
Agent CaffeineSponsoredClose your lid, come back to finished work
SnugSponsoredThe clipboard manager for the AI eraPromote your product hereReach 125k+ visitors monthly
WaxSponsoredBlock anything with supervision, without erasing your iPhonePromote your product hereReach 125k+ visitors monthly
knobsSponsoredSee your site on any phone, any day, in any languageDetails
- Repository
- github.com/mckruz/comfyui-expert
- Category
- Generative Media
- First seen
- Jun 3, 2026
More from mckruz/comfyui-expert
All 10Recommended
More Generative Media
jianying-video-gencclank/lanshu-waytovideo使用剪映(Jianying/小云雀)的 Seedance 2.0 模型自动生成AI视频。支持文生视频(T2V)、图生视频(I2V)、参考视频生成(V2V)和向后延伸(Extend)四种模式。当用户需要生成AI视频、使用Seedance模型创作短片、基于参考图像/视频进行风格转换，或对已有结果继续延长时使用此技能。需要预556skills.sh installs
text-to-image-prompt-optimizermanzxiao/text-to-image-prompt-optimizertext to image prompt optimizer526skills.sh installs
tuzi-image-gentuziapi/tuzi-skills通过兔子API（nano-banana 模型）、Google、OpenAI、DashScope 和 Replicate 进行 AI 图片生成。支持文生图、参考图片、宽高比、模型选择，以及 OpenAI 兼容网关方言切换。当用户要求生成、创建或绘制图片时使用。208skills.sh installs