zh-CN/tutorials/flux/flux-1-uso

[giscus[bot]]

zh-CN/tutorials/flux/flux-1-uso

使用字节跳动 USO 模型实现统一风格和主体驱动生成

https://docs.comfy.org/zh-CN/tutorials/flux/flux-1-uso

[chwlili]

节点去哪下载啊？？

[comfyui-wiki]

更新到最新的提交，这个是原生的节点

[Manish-cel]

one node is showing missing in the workflow, and when i go to install missing node then it's showing me the following error-

Failed to find the following ComfyRegistry list.
The cache may be outdated, or the nodes may have been removed from ComfyRegistry.
comfy-core

[comfyui-wiki]

It's only supported in Nightly. If you are using Stable or Desktop, it's not supported yet. I use some subgraph nodes in this workflow, so the missing nodes are inside the subgraph.

[Manish-cel]

🙏🏻 Thankyou.

[sipie800]

I did pull the lastest commit till the posting time, at least two nodes are missing

[comfyui-wiki]

? Can you find the USO template?

[sipie800]

why is node type a hash code ??😂
{
"id": 89,
"type": "d31e1095-65ee-4ba3-b4d0-a21e493dd0bd",
"pos": [
-630,
2160
],
"size": [
274.080078125,
120
],
"flags": {},
"order": 24,
"mode": 4,
"inputs": [
{
"name": "clip_vision",
"type": "CLIP_VISION",
"link": 142
},
{
"name": "image",
"type": "IMAGE",
"link": 143
},
{
"name": "model",
"type": "MODEL",
"link": 144
},
{
"name": "model_patch",
"type": "MODEL_PATCH",
"link": 145
}
],
"outputs": [
{
"name": "MODEL",
"type": "MODEL",
"links": [
137
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.56"
},
"widgets_values": [
"center"
]
},

[comfyui-wiki]

It's a subgraph not a new node, check this: https://docs.comfy.org/interface/features/subgraph

[gitmeAnV]

Comfyui-workflow-templates==0.1.75
ComfyUI version: 0.3.57
ComfyUI frontend version: 1.25.11

I see ByteDanceUSO in templates
and
Downloaded example workflow also works as expected.
Thank You!

[Hermit6202]

CLIPTextEncode

ERROR: clip input is invalid: None

If the clip is from a checkpoint loader node your checkpoint does not contain a valid clip or text encoder model.

[BigZampano]

Same problem here..

[BigZampano]

Edit: It works if you use the DualCLIPLoader node + Load VAE node and connect them to the nodes the checkpoint was connected to...

[comfyui-wiki]

Make sure you are using the FP8 checkpoint version, It's an all-in-one model : https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors

[aitalus]

I'm trying to get it working on MimicPc with ComfyUI v0.3.57 and it doesn't work, even though they announced that it is supported. Specifically, I have created the "model_patches" folder in the given path (/models/model_patches/use-flux1-projector-v1.safetensors) and loaded the correct file, but the ModelPatch Loader node does not display it. Does anyone have any suggestions?

[comfyui-wiki]

Can you leave a screenshot here?

[min-star]

comfyui version is 0.3.57,but error broken!

Missing Node Types
When loading the graph, the following node types were not found

d31e1095-65ee-4ba3-b4d0-a21e493dd0bd
805a5f96-6fdd-45a4-a1f3-623234fa734f
08624421-a41c-413b-9de7-d68b0b60b667
74d27e51-9780-451f-9dde-8bf303d00011
581ab310-5783-4e50-b220-4d94035eb469

[min-star]

solved!used workflow which is uso official,changed it to comfyui org is okay

[mangurian]

How much VRAM is needed?

[comfyui-wiki]

It’s just similar to Flux.1's workflow. But I haven’t fully tested it.

[sweihub]

How to do you use multiple subjects?

[comfyui-wiki]

Here is the example

[MrFries1111]

extra hands ,arms and body stuctures are normally having problems. even changing the paramters it always give this errors. Hopefully can be fixed and updated. thank you

[comfyui-wiki]

Is it a ComfyUI issue? Could leave a screenshot?

[sc-cheng17]

这个工作流生成的图片全是噪点，图片与提示词内容无关

[Ling-APE]

Is it possible to add nunchucku support? Just tested with it and gets an error :

• Exception Type: KeyError
• Exception Message: 'single_transformer_blocks.0.proj_mlp.lora_A.weight'

The way I change the workflow is changing the checkpoint loader to nunchucku's dit loader and using the nunchucku's LoRA loader to load the USO LoRA.
Does this means we need a different LoRA version support or is this nunchucku's support issue?
The original workflow in this blog post works fine for me. Thanks!