官网:https://blackforestlabs.ai/
GitHub:https://github.com/black-forest-labs/flux
forge github:https://github.com/lllyasviel/stable-diffusion-webui-forge
Hugging Face:
https://huggingface.co/black-forest-labs
https://huggingface.co/lllyasviel
https://huggingface.co/comfyanonymous
Flux.1-Dev Hyper NF4:
Source: https://huggingface.co/ZhenyaYang/flux_1_dev_hyper_8steps_nf4/tree/main from ZhenyaYang converted Hyper-SD to NF4 (8 steps)
Flux.1-Dev BNB NF4 (v1 & v2):
Source: https://huggingface.co/lllyasviel/flux1-dev-bnb-nf4/tree/main from lllyasviel
Flux.1-Schnell BNB NF4:
Source: https://huggingface.co/silveroxides/flux1-nf4-weights/tree/main from silveroxides
ComfyUI: https://github.com/comfyanonymous/ComfyUI_bitsandbytes_NF4
Forge: https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/981
Forge支持的Flux基础模型一般有这几种:
- flux1-dev-bnb-nf4-v2.safetensors 张吕敏大佬搞的unet量化模型。
- flux1-dev-fp8_unet.safetensors 只包含unet模型。
- flux1-dev-fp8.safetensors 包含unet、vae和text encoder模型。
这几个模型需要放到 Forge 模型目录的基础模型子目录下。
Forge需要的VAE和Text Encoder模型包括:
- ae.safetensors Flux使用的VAE模型;
- clip_l.safetensors 相对传统的文本提示词编码器模型;
- t5xxl_fp16.safetensors 比较新的文本提示词编码器模型,使用自然语言。
这几个模型需要放到 Forge 模型目录的 vae 子目录下。
值得一提的是:FLUX的跑图参数是比较固定的,不建议用户随意修改采样器、调度;CFG等参数。
推荐参数:
Steps: 20,
Sampler: Euler
Schedule type: Simple,
CFG scale: 1,
Distilled CFG Scale: 3.5,
Model: flux1-dev-bnb-nf4-v2
- 提示词反推插件:
https://github.com/richadlee/stable-diffusion-webui-wd14-tagger.git
- 视频插件安装:
https://github.com/licyk/sd_forge_hypertile_svd_z123
参考: