Views
No views yet
An all-in-one sampler custom node for ComfyUI. Load a whole checkpoint or a split stack (diffusion model + dual CLIP + VAE), pick from many architectures (Flux / SD3 / Wan / Qwen-Image / Krea2 / Z-Image / LTXV / Hunyuan / Lumina2 …), set weight precision (default / fp8 / nvfp4 / bf16 …), wire LoRAs, feed upstream images for img2img — all from a single node.
🛡️ Binyuan采样器 V5.5,分类 Binyuan。
一体化采样器:自带模型/CLIP/VAE 加载、提示词、尺寸预设、采样参数、LoRA 列表、上游图像处理,并把图像、模型、CLIP、VAE、Latent、正负条件全部作为端口输出,方便串联下游。ComfyUI/custom_nodes/binyuan_sampler_plugin_v5.5/,重启 ComfyUI。Binyuan → 🛡️ Binyuan采样器 V5.5。加载模式:
整包Checkpoint:在 Checkpoint 下拉选一个 .safetensors 整包模型。分离式(Flux/SD3/扩散):分别选 扩散模型、CLIP_1/CLIP_2、VAE,并在 CLIP_类型 选对应架构(flux/sd3/wan/qwen_image/krea2…)。LORA_LIST 选 LoRA,或在 lora_json 里手写 JSON 列表。上游图像_1,按需设 重绘强度、上游图像处理、Latent输入源。图像 端口接 SaveImage 出图。| 参数 | 说明 |
|---|---|
| 加载模式 | 整包Checkpoint / 分离式 |
| Checkpoint / 扩散模型 / CLIP_1 / CLIP_2 / VAE | 模型文件下拉(自 |
外部模型、外部CLIP、外部VAE、外部Latent、外部正面条件、外部负面条件、上游图像_1/2/3。图像(IMAGE)、模型(MODEL)、CLIP、VAE、Latent(LATENT)、正面条件、负面条件(CONDITIONING)。CLIP_类型 必须与实际模型架构匹配,否则加载/编码会报错。权重精度 选 fp8/nvfp4 等需对应运行环境支持。trending_raw.json 是节点内"热门模型"列表的本地缓存数据,可保留或删除,不影响功能。

🛡️ Binyuan采样器 V5.5 under category Binyuan.
An all-in-one sampler: loads model/CLIP/VAE, encodes prompts, runs sampling,
decodes to image — and also exposes model/CLIP/VAE/Latent/conditioning as
output ports for chaining downstream nodes.ComfyUI/custom_nodes/binyuan_sampler_plugin_v5.5/, restart ComfyUI.Binyuan → add the node.加载模式 (load mode):
整包Checkpoint: choose one .safetensors checkpoint.分离式: pick diffusion model + CLIP_1/CLIP_2 + VAE separately, and set
CLIP_类型 to the matching architecture (flux/sd3/wan/qwen_image/krea2…).LORA_LIST or write a JSON list in lora_json.上游图像_1, tune 重绘强度,
上游图像处理, Latent输入源.图像 to SaveImage.图像(IMAGE), 模型(MODEL), CLIP, VAE, Latent(LATENT),
正面条件/负面条件(CONDITIONING).CLIP_类型 must match the real model architecture or loading/encoding fails.trending_raw.json is local cache data for the in-node "trending models"
list — safe to keep or delete.__init__.py — node BinyuanUltimateSamplerV9 (BinyuanUltimateSampler)js/chicun.js — frontend size-preset helpertrending_raw.json — trending-models cache dataREADME.md / LICENSE (MIT)Binyuan Sampler.git clone https://github.com/yjitiu/ComfyUI-Binyuan-Sampler.git binyuan_sampler_plugin_v5.5