Fizgig
Fizgig — LoRA & Fine-tune Studio for Klein 9B, Krea 2, MiniMax H3, Qwen Image 2.1, Anima & Z-Image Turbo and Ace-Step XL: train, fine-tune, profile, repair and extract LoRAs & LoKR for Image, Video & Music Models
- 785
- Python
- Apache-2.0
- 1 d ago
Fizgig — LoRA & Fine-tune Studio for Klein 9B, Krea 2, MiniMax H3, Qwen Image 2.1, Anima & Z-Image Turbo and Ace-Step XL: train, fine-tune, profile, repair and extract LoRAs & LoKR for Image, Video & Music Models
WeeLLM runs large diffusion models with as little as 4 GB of VRAM, without any quantization. It dynamically determines how many layers can fit within the available VRAM and streams the text encoder and transformer layers to the GPU layer by layer, enabling inference on hardware with limited VRAM. It supports both safetensors and GGUF models.
Native ComfyUI nodes for Krea 2 AnyPaint - arbitrary-mask inpainting, outpainting, and mixed image editing.
Krea 2 (Turbo) prompting skill with huge references library, including adult (the most tested section ;)
中文 9 段式人像提示词创作体系 · 可直接出图的模板/风格预设/千条级批量生成器 · 覆盖全身照与上半身中景照 · 适配 ComfyUI/Krea2/GPT Image 2 · 无强制约束,创作自由
514 Krea 2 Turbo prompts with generated examples and ComfyUI wildcards
18+ only|文生图提示词教学站:figure-photo / youth-seduction / erotic 三线导演台 · 维度体系 · 写法规则 · 组合图库。含露骨提示词文本,禁止未成年;不随仓库分发图片。Adult-oriented text-to-image prompt engineering guide. Explicit textual content. 18+ only.
A node designed specifically for image editing models—supporting tiled refinement, editing, and upscaling—compatible with FLUX.3/FLUX.2 klein, Qwen-Image 2.1 / Edit 2509 / 2511, Krea 2+lora, Moss +lora, MiniMax-H3-Image, while ensuring cross-tile consistency.
ComfyUI 节点:Krea2 白垩光长文提示词自动装配 · 553 母题 × 2432 角色 · 一个母题决定整张画面 · 纯本地零联网 · 只生成提示词不生成图片
A ComfyUI diffusers implementation of Krea2 for the highest possible quality and most options
ComfyUI custom nodes that guide Krea 2 with reference images: stack up to 12 guide cards using plain-language recipes - keep the subject, borrow style/palette/lighting/material, copy pose or shapes, avoid text/logos.
⚗ Calibrated sampler for Krea 2 Turbo in ComfyUI: NAG (negative prompt at cfg 1), two-phase prompts, composition previews, prompt enhancer. By AlchemicaMente.
Community-calibrated Krea-2 LoRA training workflow (v0.4.0) for face/digital-human/character LoRAs on Ostris AI-Toolkit. TRAIN on Raw, RUN on Turbo. WebUI job takeover, gradient accumulation, 12 pitfalls documented. 训练 Krea-2 人脸/数字人 LoRA 的社区校准工作流 v0.4.0
AetherCanvas AI: one local web app for visual AI on your own GPU: text-to-image, photo editing, inpainting, uncropping, video and face swap with Qwen-Image 2.1, Krea 2 and Wan 2.2 (GGUF), driving ComfyUI.
ComfyUI custom node for AI image generation with Krea 2, Z-Image, Anima, Qwen Image, Qwen Image 2.1 (Turbo) and MOSS-Image. Optional PixelDiT (PiD) 2x/4x upscaling, multi-LoRA and Korean/English UI.
Inference Krea-2, Qwen Image 2.x , MOSS-Image, Flux 3 Image Klein, P-H3-Image, Minimax-H3-Image with ComfyUI