Post "[ADVANCED WORKFLOWS]" by "Mickmumpitz" from patreon

Murathan Dogan

Hello, when i run it and get to 8% "samplercustomadvanced" so it says "reconnecting" and nothing happens after that. whats the issue?

Gabriel Abel

Hello Mickmumpitz I am try to run the SDXL Workflow, but i have this issue: DownloadAndLoadFlorence2Model Error(s) in loading state_dict for Embedding: size mismatch for weight: copying a param with shape torch.Size([50, 512]) from checkpoint, the shape in current model is torch.Size([50, 1024]). I used all your recomendet modules and all nodes. At the beginning i had an error with microsoft and huggingface, but i dont have the error logs, but it was because of de Florence2 Model too Thank you for you amazing help

Mick Mahler

Manually installing the node should fix this, but I recommend you check out the newest version of this workflow! It is much simpler and easier to install

Veysel Kılıç

Florence2Run 'NoneType' object has no attribute 'shape' I am getting the error. I am working in the sdxl_adv workflow.

Mick Mahler

Hmm it still works for me after updating, but I highly recommend you check out the newest version of the consistent character creator (newest post). I simplified a lot and it is now more versatile and easier to install!

Yury Goryaynov

The workflow is broken due to multiple package conflicts and missing nodes. Specific issues: streamdiffusion ⟂ diffusers → streamdiffusion requires diffusers ≈ 0.24, while the workflow requires ≥ 0.29 → unsolvable conflict. inference-cli 0.53.0 is incompatible with newer dependencies (aiohttp 3.12.15, opencv-python 4.12.0.88). torchaudio 2.8.0 requires torch 2.8.0, but the workflow installs a different version (2.1.1 → 2.3.1). sam-2 requires torch ≥ 2.5.1 and torchvision ≥ 0.20.1, but lower versions are installed. dctorch is incompatible with numpy 2.x (needs < 2.0.0). gradio and gradio-client break due to mismatched dependencies (aiofiles, markupsafe, pillow, websockets). grpcio-status requires protobuf ≥ 5.26.1, but the workflow uses protobuf 4.25.8. mediapipe is not supported on macOS/ARM at all. Inside the workflow: The DiffusersModelMakeup node fails → it tries to call init_custom_adapter from StableDiffusionXLPipeline, but that method does not exist there (only in MVAdapter…SDXLPipeline). Nodes like DiffusersSchedulerLoader / DiffusersVaeLoader / LdmPipelineLoader / DiffusersMVSampler / DiffusersModelMakeup are marked as UNKNOWN → because the ComfyUI-Diffusers extension (Limitex) is missing. Bottom line: The workflow cannot run “as is” on macOS (and very likely not on Windows/Linux either without fixes). It requires manual dependency patching, downgrading/upgrading packages, and installing missing nodes.

Mick Mahler

I am sorry to hear that. :( Can you post the error logs in the #need-help channel on discord?

Erns

Just joined and having a bad start , it completely derailed my comfy , I am getting so many errors and even Python is getting a pop up window message for python .exe error, none of the above seem to be able to fix it :(

Excelsiorian

Same issue here, any help available?

Ralphie Oakley

Hey, could you help with this error? Tried reinstalling transformer,s but not working unfortunately. Any suggestions are appreciated! Thanks DownloadAndLoadFlorence2Model 'Florence2ForConditionalGeneration' object has no attribute '_supports_sdpa'

Mick Mahler

Are you using v02 of the workflow? Oh and have you checked out the troubleshooting section in the guide? If you want can you post this in the #need-help channel on discord along with a screenshot? That helps a lot figuring things out!

Mick Mahler

Hey I am a bit confused, how did you find this post? I show every step in the YouTube video. I show and test the loras I trained with it. You can download the datasets I created using this workflow attached to this post. You can download and test the loras I trained using this workflow. I show how to train a lora for flux.

LegolasTheElf

are there any chance to see this work finished and useful? A) the output doesn't produce a set of images useful to seriously train any LoRA: to do that, you need at least a dozen of images (better 30) with the relative description in a txt file. we obtain a single image of a worksheet useful as a label, nothing more. no images of the same resolution or neither of the same proportions, no txt description files. B) No instruction or method given to train a LoRA with the output given, neither to use the LoRA to produce more images C) Nvidia 5000 series are very available at now, and a RTX 5070Ti 16 Gb GPU is the most affordable and available GPU for AI working at now. All scripts given doesn't consider that public availability, and gives headache to whom use it. D) as for what you affirmed into the other posts you have postad, all of your works is based on this object. But this object simply does't do what you say. it's , at least, unfinished. E) you don't ask for a low-price patreon, your monthly request is one of the highest around. i think that we can ask for a finished job for that price before you pass to do different things like animation. Please, give us a working script, not something we have to arrange to obtain something. At least to the patreon who pay you for this.

Bunny

Hey Mick! Just so you know, in the SDXL version, in 01_MULTI_VIEW section, masked pipeline output from Diffusers Model Makeup doesnt connect to pipeline input of Diffusers MV Sampler. And the workflow gets stucked there.

Mario Billy

Hey everyone, I'm getting this error trying to launch ComfyUI: AssertionError: Torch not compiled with CUDA enabled This is from a portable build (ComfyUI_windows_portable) with embedded Python. I didn’t touch the Python or PyTorch versions, just added some custom nodes from this tutorial. Anyone know how to fix this without breaking everything or having to reinstall all my models?

Scott Riemer

same issue.. cant run torch 2.6 because it nukes my venv every time.. so until they backport it to 2.1.1 this will never run because of the torch update

Tim Sullivan

A heads up…If you have a 50xx GPU and get a crash in Flash Attention with these workflows, the fastest solution is to disable xformers in ComfyUI via the command line. This is a known conflict with 50xx GPUs and FlashAttention and it’s been reported as a bug. The only other solution is to downgrade PyTorch to 2.7.0 but I need 2.7.1 for other custom nodes.

Bunny

I am getting this error when I open the files "The workflow you've loaded has corrupt linking data that may be able to be fixed."

Ben

Locate the top left node with the name "Fast Groups Muter (rgthree)". Mute every group but the image generator, queue the workflow. After the run has finished, unmute every group again. Queue the workflow again and it should run normally. At least that's what made it work for me. Good luck!

jomy nn

I get the error about Florence2Run module 'tensorflow' has no attribute 'random' Please help me.

Ninja Pumpkin

Hey anyone got this error? how do you fix it? Error(s) in loading state_dict for T2IAdapter: size mismatch for adapter.conv_in.weight: copying a param with shape torch.Size([320, 1536, 3, 3]) from checkpoint, the shape in current model is torch.Size([320, 384, 3, 3]).

Ninja Pumpkin

Hi just bought the workflow because I like supporting creators that provide great information! I imported the workflow but it seems like I am missing, the use input image nodes? the first part of the character creation is missing it seems, how do we adequately use this?

paul fedor

I would love to see a workflow...once character is generated.....moved into video workflow to add movement.

Pixel Pilfer

Hallo Mick, ich habe Ihnen gemäß dem deutschen Verbraucherschutzgesetz ausreichend Zeit gegeben, um auf meine Nachricht bzw. Reklamation eines mangelhaften Produkts zu reagieren – doch Sie haben sich entschieden, nicht zu antworten. Ich gebe Ihnen hiermit eine letzte Chance, den Betrag innerhalb von 24 Stunden zu erstatten. Andernfalls werde ich nicht nur einen Fall beim PayPal-Käuferschutz eröffnen, sondern die Angelegenheit auch an die Verbraucherzentrale eskalieren. Vielen Dank.

Navajo Demar

I'm seeing an odd error, "TypeError: graph.nodes is not iterable on loading the SDXL v2 workflow.

Jeffrey Allen

I came here for the same problem. Disconnected the pose sheet (v2), and got a realistic person. Not sure what other text in the character sheet drives it to cartoon / anime style. Any thoughts?

Manuel Nixon

Not sure why but, when using Load Image, the Multiview images come out like cartoons, whereas the face (input was real photo) of the person looks true to life.

Manuel Nixon

Thank you Mick so creating this very timely workflow!

ice beer

Got the same problem, It is because it wants to multiply with a string instead of a int. but i dont know how to fix it

Pixel Pilfer

Hi there Mick. I was following your tutorials and was having issues. Having thought it was because I am not using your advance (subscription only) .json workflow, I paid for your Patreon. However, we are getting so many errors in the following areas despite repeated updated through the manager and even downloading and adding them to the relevant folders in the ComfyUI folder - ImpactControlBridge - SetUnionControlNetType - (VAEEncode, VAEDecode - SamplerCustomAdvanced - UltimateSDUpscale, CLIPTextEncode and so many more DESPITE repeated updated through the manager and even downloading and adding them to the relevant folders in the ComfyUI folder. - Can you please help.. this is URGENT (the reason I came to pay here on Patreon) ALTERNATIVELY, I am ready to pay you a fixed fee to do these installs for me either through TeamViewer or any possible platform. Melde Dich bitte! Danke und Grüße aus Berlin PxP //will also send comments on YT in case your attention here is limited. Thanks again

SOC_3D

Hello! I'm stuck with the advanced SDXL workflow in the emotions part. it says that "can't multiply sequence by non-int of type "numpy.float32". The error occurs within the expression editor, any Ideas?

Sérgio Hime

Hello. Just bought the Patreon exclusive. i am trying to use the workflow, i have installed everything since last week but i cant manage to use Florence2. I am running Comfyui on Pinokio. Any thoughts??

Arash

I can't import the workflow for the following reason: - **Exception Message:** TypeError: Cannot read properties of undefined (reading '0') Any idea what the issue is?

Carl Piaf

Try a clean ComfyUI install in C: One poster in the Discord help mentions this. Sadly, it's never mentioned in the videos. I found using a clean install works and also noticed at one point in a video that he uses multiple installs. Diffuser is a big problem if ComfyUI is not installed in C: (something with the ComfyUI Manager install function, I have been led to believe).

Carl Piaf

Yeah, I had this last night. Going to look into it today and see what I can come up with.

Carl Piaf

Try using a clean install of ComfyUI in the root C: In his videos I saw he has multiple installs labeled with the different dates of these workflows. Custom node incompatibilities are a major issue that can only be solved with clean installs it seems.

Alwyn Cruz

hello, i joined your community and downloaded the advance workflow. although my diffuser and DDUF i think is conflicting. diffuservaeloader is missing and i can't seem to install it. can you please help me. i messaged you in the community too.

Tokio Studio srl IT10640050968

Hi! I'm trying to use your workflow, but I can't get it to work. Here are the main issues I'm encountering: Error with ImpactControlBridge: missing required input value. Error with SetUnionControlNetType: missing input control_net. Several other nodes (VAEEncode, VAEDecode, SamplerCustomAdvanced, UltimateSDUpscale, CLIPTextEncode, etc.) are missing critical inputs (vae, sampler, clip, upscale_model, florence2_model). Some required files are missing from the input folder (example: Amber_00_CHARACTER_GENERATION_00001_.png). Also, I often get errors like missing positional argument. Can you please help me understand what’s missing or how I can make it work properly? Thank you so much!

CLAUDE BARAKAT

AutoCropFaces is not compatible with my nvidia 5090. any recomendatios please

Benjamin Lebron

I used claude 3.7 to troubleshoot by pasting error messages into, it was the only way Icould get it to work

Nicholas Boyer

what fixed it for me was I installed a local version that wasnt prepackaged. The prepackaged comfyui downlaod from comfy site has a version of python that is dated for this workflow and you can't easily install or upgrade python using pip installation. I'd suggest uninstalling comfyui if you installed it through a packaged file download a zip from github and install that instead. Then you can upgrade pip and it should recognize custom nodes.

Diego Ramos

I give up, Spending all day just by ...installing missing nodes, and VaEs etc...yes I did " install missin nodes" bottom ...but still ...doent work..error everywhere!!..a ahhhhhh!!..Im gettin crazy

Paint Punisher

This is great but i cant get any of these to work properly.. even on discord. I wish there was some kind of personalized care we could maybe pay for extra to help us get going.

siyan liu

Does anyone know why the download of flux gym from pinokio always fails? It seems that cuda cannot be downloaded.

Mint Copenhagen

Hi, we are testing out AI for creating pitch material. Our biggest hurdle is consistent characters. We'd be interested in more content on using multiple consistent characters. Especially for generating visual story beats.

Imamiablo

I would even pay more than 10$ if it were more beginner friendly. Because this looks marvelous, but ComfyUI is a pain in the ass itself and even for an average user like myself, who has previously used NKMD, Fooocus, Automatic1111 and Forge and some basic youtube knowledge of C++ and python, this is really hard and really spend lots of time on the side to figure out many things ... The author delivers really good information, but people are also angry for a reason. Clearly one has to be much more competent to fully understand what author says depsite it actually seeming easier than it look at the first glimpse.

Jose M. Aranda

Everything works perfectly. Many times what happens is that novice people need to be trained a little bit before getting into these workflows. If you know where to place everything works perfectly. It's not the workflow's fault, it's yours...

Giraffart

Paying $10 a month shouldn't give the right or encourage people to come here and fill the comments with rude language. Easy people, easy! Indeed, I think $10 is very cheap for this information and that's why some of you won't appreciate how big is the work and you are actually luckier than you think. Thanks very much for all the contents and support on discord! Please keep doing what you are doing!

Charawit Charawit

you need to switch mvadapter version

JINU M JOHN

why do the nodes in the worflow look messed up

No More

Wow - how are you allowed to sell your subs here - there is literally none of this workflow that works

gongora ia

why i got this error, DiffusersMVSampler Allocation on device

Alexander Holovko

Hi, I have an issue with DiffusersMVSchedulerLoader and DiffusersMVModelMakeup that are needed for this flow: 250226_MICKMUMPITZ_CCC_v02_SDXL_ADV.json Trying to run everything on Ubuntu 22.04 Could someone please advise how to solve this?

F

Hi, what should be the order for running those JSONs? Should I run Lux first and then sdxl? There are so many. Please provide more information.

Mick Mahler

Okay I tried uploading the .zip again and I think it should work now. If you need any other files or help please ask on my discord. I am on there almost every day helping out wherever I can.

Mick Mahler

sorry, this is the second time Patreon killed a file like this. I'll upload the stuff to drive and then add the link to the post.

Tom Rubery

Where is the additional workflow for realistic images? Why is the patreon.zip file empty? For the amount patreons this guy has, he doesnt really seem to pay much attention to them

Roberto Sanz

Correct, Patreon.zip doesn't exist

rajesh j

Patreon.zip file doesnt exist media file not found for downloading

Selina Druifmans

For anyone that is reading this: He DOES answers, but he prefers discord. He literally fixed my problems in a few minutes when i asked him on discord... So hop on discord if you have issues or questions!

Mick Mahler

Are you on our discord? I see stuff much faster on there! Also I am currently reworking the MV Adapter part as they released a new update and I hope it will work then, releasing the new workflows later today!

Иван Меер

😅✌️

rhb30001

I need help this doesnt work .. definitely not recommending this guy to anyone

上进王

When I imported the workflow into ComfyUI, the ComfyUI interface appeared completely blank - literally the entire page was white with no nodes visible at all. However, I have already installed all the missing nodes. When I click the "Fit View" button, it throws an error: "Cannot read private member #boundingRect from an object whose class did not declare it". Could you please provide assistance? Thank you.

T

When I try to get the workflow work, I need to install all missing nodes, also "ComfyUI-Diffusers" and "ComfyUI-DDUF" but ComfyUI tells me that both nodes are in conflict. When I ignore that and install "ComfyUI-Diffusers", it damages my ComfyUI system and I can't access until I reinstall torch and / or torchvision and then let ComfyUI Update it before it will let me use it again. How can a workflow need something that is impossible to use? I just paid my very first money on patron for a workflow that is not working at all, does anybody experience the same? Any solutions?

g balaji

getting following error DiffuserMVSampler with redline on the box on both SDXL and FLUX workflow

Goqor Grigoryan

he is not anwering even to personal messages

Magali Grande

Very angry to have paid and not to have support, I do not recommend you.

Kang Hua

You upgraded Comfyui first. Then take a look at the installation. Additionally, ComfyUI Diffusers does not require installation, Installing the mvadapter is enough. You can try it

Kang Hua

Is the training workspace created using the above workflow? What about creating a basic image? Is this training set feasible? Do you need to add other picturesSample FLUX Training Datasets --

Maria Bartzoka

There are custom nodes missing ComfyUI-Easy-Use ComfyUI-MVAdapter ComfyUI-Diffusers ComfyUI-DDUF I have tried everything, plus the "ERROR 1 - Failed to install custom nodes" instructions but nothing worked.

Kang Hua

May I ask if this training lora dataset created from a single image, with only one set of clothes remaining unchanged, will affect the generality of lora?

Hu$h

If you're new to image generation, and want to watch a creator who actually troubleshoots with his audience, even free members. Go check out this dude. His channel is new and he is free often to help. https://youtu.be/9UrLQbdESjk Discord: https://discord.gg/crMY3zHzZh Would love to see more people get involved who are excited to learn.

Nick S

this is my biggest challenge: output is a masked pipeline and input is a pipeline . . . now sure how to fix that one.

ArtJunkie

Been wondering the same thing

Asyouwere

Why don't you ever response to the questions @mickmumpitz?

Magali Grande

Hi I've a issue, please can you help me : # ComfyUI Error Report ## Error Details - **Node ID:** 651 - **Node Type:** CLIPTextEncode - **Exception Type:** AttributeError - **Exception Message:** 'NoneType' object has no attribute 'float' ## Stack Trace ``` File "C:\Users\mag\Documents\SD3\ComfyUI_windows_portable\ComfyUI\execution.py", line 323, in execute output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)

Luis Estrada

I was able to install it, I also had to do some extra stuff, but now I'm stuck at a point where I Queue, and I get an error saying this: DiffusersModelMakeup - Allocation on device. I totally lost in this part.

foxi

so i put my \input folder link into KJ. by defaut : width = 0 height = 0 keep_aspect_ratio = crop image_load_cap = 0 start_index = 0 include_subfolders = false i leaved all text boxes by default. when i press "Queue" i got this error --> "LoadImagesFromFolderKJ division by zero"

foxi

Which image i have to put on the \input ? The MVAdapter in the custom nodes in comfyui, did you customize it or is it ready to use? Sorry for the questions, I'm just starting out and discovering at the same time. In fact, I understood that we can generate an image or start from a reference image, but in "Load Images From Folder (KJ)" are the 6 images generated by the MV-Adapter on hugginface put in a folder that must be loaded? Or I have to assemble theses 6 images in an image editor into a single image then load the image. The problem is that I don't see the complete process, in the video Mick has already set up his workflow, but I have to update some points that are not explained :s

Pete

I couldn't download it via the Custom Node manager, I ended up getting it from GitHub https://github.com/ssitu/ComfyUI_UltimateSDUpscale

Pete

I got the MVAdapter in the Custom Nodes in ComfyUI. The load images come from the \input directory under your ComfyUI install

foxi

First of all thx. I get your adv flowflow, but im stuck. Well, your workflow in your video seems different to the one i downloaded from patreon, i dont have the node "1 USE GEN / 2 USE INPUT IMG", also Is there a way to install a workflow version of MV-Adapter [Image-to-Multi-View] directly in my ComfyUI to not depend on hugginface's? ? And also, where im supposed to load folder from "Load Images From Folder (KJ) ? I got an error there Im pretty stuck, i had to reinstall ComfyUI cause conflits after i switch ADV for low VRAM to the normal ADV one. thx for your content and work @Mickmumpitz

Luis Estrada

Could you solve the issue?

Shawn Harris

Is there anyway you could do a tutorial on consistent characters that are more illustrations than 3D. I haven't see that anywhere and it seems like your comfyui tutorial makes them 3D unfortunately

Raf Duran

Same problem @admin

Luis Estrada

I decided to install the Flux version. I followed the guide, but I'm getting an error. It says UltimateSDUUpscale is missing. How can I solve that?

distant melodies

Initially I completely redid my env and comfyui install, so this is on an updated version of comfy as well.

distant melodies

To those of you having issues, I'm not sure how your comfyui is set up, but I use anaconda navigator to create an anaconda environment for comfy. I was getting issues with 4 or 5 nodes not installing, reading back through the cmd prompt, I noticed it was an issue with hugging face, upon googling i found I just needed to update it. I activated my environment, went to: "C:\Users\x\anaconda3\envs\ComfyUI\Lib\site-packages\huggingface_hub" and ran " pip install -U huggingface_hub". Now it's all installed.

Goqor Grigoryan

he will wait, unit everyone pay one more $10, then will update something

Goqor Grigoryan

Yes. He even don't answer to personal messages

Aaron

Ya it's only going to be $10 for me. If you charge, the workflows need to work proper.

Agency Nilton

So I just gave 10$ to someone who doesn't answer to anyone here? I'm going to try to install what I need, hoping to solve the problems I'm going to encounter ...

Goqor Grigoryan

it is okay, workflows are wrong. And he will never anser. You have to debug by yourself

Chris Hatfield

Don't have that data in front of me right now, but Have you tried running the SDXL version of the JSOn files Mickmumpitz released. I got it working on the updated version of Comfy UI ( don't have manger version in front of me) will respond when I get home. I'm running on Colab T4 GPU. ( Didn't do anything special other load some of the missing files Control net union etc, as didn't have them installed on my instance of Comfy )

kid poker

Thanks a lot for answering both my original question and the updated one. Can you please share your comfy ui manager version? Actually I'm running this on an EC2 linux instance but constantly getting differnet errors due to version mismatch of comfyui (the latest guide and your answer sorted that part) and **comfyui manager**.

Chris Hatfield

@Kid poker. I see you've edited your orignal question. I'll include full question for better context for anyone reading. @ Kid poker QUESTION: " How do you know he uses that particular version?" If you don't mind, could you share your pip freeze of the python environment through in which you made this work? It will be a great help for everyone in the comments." ANSWER: Information came from documentation released to patreon subscribers. Not an iteration of a Pip Freeze. Hope that answers your question with full transparency and clarity.

Chris Hatfield

@Kid poker. If i'm making an implicit statement. That would imply that I've made the determination through a documented source ;). Instead of framing the question to challenge the statement. Maybe try asking direct questions related to your core intent. " How do you know he uses that particular version?" Lets breakdown. examine you intent behind the question. Why you want to know how, I how I know that he uses that particular version? Now, internally examining your intent ask direct questions more closely aligned to your core intent.

Chris Hatfield

How do you know to use your fingers on the keyboard and press enter? How do you know, how to right click while using the mouse. How do you know how to use your current operating system? Now, read the questions above interpret and think internally of your first reaction to being asked those questions ( no response needed ). That should answer your initial question. ;)

kid poker

If you don't mind, could you share your pip freeze of the python environment through in which you made this work? It will be a great help for everyone in the comments.

Jeremia

same for me. would be great for him to upload his comfyui folder withouth the big models, even if its a 100gb download

Aaron

You do great work, but I cannot get 4 of the custom nodes to work even after the hugging face upgrade. Just joined for the flux workflow today.

Chris Hatfield

Can you provide any guidance on the specific issues of why work flow will not work with ComfyUI above 0.2.3. This may help users who want to modify the JSON work flow, to work with the most recent version of Comfy UI. Any information would help the community, and thanks for releasing.

McLovin

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

SDXL and SD1.5 are mixed

Goqor Grigoryan

why you have SDXL and AD1.5 at same time? This makes conflicts. And why you don't answer? Bro i will spam you if you continue like this))

Goqor Grigoryan

it is because he uses SDXL and SD1.5 at same time and that makes conflicts. I dont know why he does this. May be it is time to unfollow

Goqor Grigoryan

No, this guy never answers

Bold

same error here :(

Petr Šalša

hello on flux andvance i have error flux checkpoint doest found but i have it on checkpoint source

Focco

Hello, I don't have errors with the FLUX and SDXL, but these emotions are not generated at all. I selected the image number, but there are red bars around the four emotions options, so the end character sheet is also not generated. But the rest is working without any issues.

gopherturtlerunning-3d@yahoo.com

Yes, the FLUX version does not function. Errors are with "DiffusersMVSampler" and with "LdmPipelineLoader" It also resulted in corrupting the previous FLUX work flow, but my SDXL seems to be working well.

Mac

There's SDXL stuff in the Flux workflow. Juggernaut etc. Pls update.

McLovin

dont bother banning me, I unsubbed already downloaded everything you offered. Save and sound on my drive :) - advanced and the free workflow are the same. Keep turning yourself into a wizzard for the LBTQ community

Mikael Tyrsen

Change crop_factor form NaN to 1.5

Mikael Tyrsen

same error

GRAZDA Michel

Hi everyone, just by curiosity, what type of GPU do you use with ComfyUI?

Alper Demirağ

Could u solve it?

kid poker

same error

Goqor Grigoryan

IPAdapter not found for SDXL. Who could run this?????

Goqor Grigoryan

Bro, generally, you have copied wrong things from SDXL to FLUX workflow. Fix this ASAP!. PLease

Goqor Grigoryan

'NoneType' object has no attribute 'float' on advanced workflow. Also Models are wrong (From SDXL)

귀연 김

same error..

Jasem Ali

I'm getting this error @ KSampler @ group 5 (lighting) mat1 and mat2 shapes cannot be multiplied (131072x320 and 768x320)

Lois González Alonso

I have the same problem... "Failed to convert an input value to a FLOAT value: crop_factor, OnlyExpression, could not convert string to float: 'OnlyExpression'"

Daniel

Try to change the node to this: IPAdapter Unified Loader FaceID and use a portrait FaceID file

Martin Christensen

I also get this error, even though all the models are there.

Amit Jangra

I tried this. It is working. Only issue I am facing is PHM expression editor. Those remain red colored although evrything installed corretly.

knifeandcj

i got this error [Failed to validate prompt for output 975: * DiffusersMVSampler 1322: - Return type mismatch between linked nodes: pipeline, received_type(MAKED_PIPELINE) mismatch input_type(PIPELINE)]

Nils Reith

I am new to this, the boxes with the modifiers for the expression editor have a red border and the workflow crashes at this point. I have installed everything like described in the instructions except ComfyUI-Diffusions because I can't get it to work. Installation via the Manager fails and the instructions from the github repo fail with "ModuleNotFoundError: No module named 'torch'". Someone knows hot to fix that? :)

Marcin Wilczyński

Gr8 work, mate. Keep going with it. We need you. I made it work on Ubuntu using an AMD card.

Lois González Alonso

Which nodes?

Christian Schülling

Im trying the whole day but I cant get the missing nodes properly installed with Comfy over Pinokio. I deinstalled everything and started from scratch but everytime amy other node is missing. How to fix that?

Ranjeet Sathe

Please share bonny image with workflow

Petr Šalša

me too:/

ric

Are there any options / alternatives for 'ComfyUI-Florence2' and 'comfyui-tensorop'? I've tried everything I could to install those two but nothing worked.

Daniel

installed everything but still got the message, even a clean comfyui installation with all the nodes was not possible. But i found my workarround

Lorus

You need to install the ipadapter models like he is showing in the video starting at 22:30 https://youtu.be/grtmiWbmvv0?si=u-qkmy0eRuDk_Zed&t=1370

Lorus

Could you please share the workflow for making a T-Pose from an input image with only a face?

Lois González Alonso

¡FIXED! I had a conflict with Diffusers and HuggingFace Hub, so I did the following: !pip uninstall -y huggingface_hub !pip install huggingface_hub==0.27.1 and !pip uninstall -y diffusers !pip install diffusers==0.32.2 I am also having problems importing some nodes. This is an error i get when I charge my files from Google Drive to my Colab session: Restoring [64/66]: /content/drive/MyDrive/ComfyUI/custom_nodes/ComfyUI-Diffusers ........................................ ....................................... DEPRECATION: git+https://github.com/cumulo-autumn/StreamDiffusion.git@main#egg=streamdiffusion[tensorrt] contains an egg fragment with a non-PEP 508 name pip 25.0 will enforce this behaviour change. A possible replacement is to use the req @ url syntax, and remove the egg fragment. Discussion can be found at https://github.com/pypa/pip/issues/11617 Running command git clone --filter=blob:none --quiet https://github.com/cumulo-autumn/StreamDiffusion.git /tmp/pip-install-6i0mbjz4/streamdiffusion_c5a1344c8b2449b7830232f05d7dfff2 Collecting streamdiffusion (from streamdiffusion) Cloning https://github.com/cumulo-autumn/StreamDiffusion.git (to revision main) to /tmp/pip-install-6i0mbjz4/streamdiffusion_c5a1344c8b2449b7830232f05d7dfff2 Resolved https://github.com/cumulo-autumn/StreamDiffusion.git to commit b623251dc055e1fd858d53509aa43e09dfc5cdc0 Installing build dependencies: started Installing build dependencies: finished with status 'done' Getting requirements to build wheel: started Getting requirements to build wheel: finished with status 'done' Preparing metadata (pyproject.toml): started Preparing metadata (pyproject.toml): finished with status 'done' ..Collecting diffusers==0.24.0 (from streamdiffusion->streamdiffusion) Using cached diffusers-0.24.0-py3-none-any.whl.metadata (18 kB) ....................................................Using cached diffusers-0.24.0-py3-none-any.whl (1.8 MB) Installing collected packages: diffusers Attempting uninstall: diffusers Found existing installation: diffusers 0.32.2 Uninstalling diffusers-0.32.2: Successfully uninstalled diffusers-0.32.2 Successfully installed diffusers-0.24.0 After that when I charge the nodes appears: (IMPORT FAILED): /content/drive/MyDrive/ComfyUI/custom_nodes/ComfyUI-DDUF (IMPORT FAILED): /content/drive/MyDrive/ComfyUI/custom_nodes/ComfyUI-Diffusers (IMPORT FAILED): /content/drive/MyDrive/ComfyUI/custom_nodes/ComfyUI-eesahesNodes Does somebody know how to fix it? I tried but I could not yet fix it. :(

Ivan Vetrov

Hey! Great work!

Data Swami

is there a way to use part 1 to create the t pose from an existing image as using the image in step 2 gives odd outputs

Daniel

fixed it with a "IPAdapter Unified Loader FaceID" Node

Christoph Struber

Maybe it works with the fixes that are mentioned int the simple verion. Have to try it later

larry leng

我也遇到了整个问题1358#节点出错

Christoph Struber

Hi, new member here. I tried the Workflow on a new comfy instance, sadly I still cannot install all nodes.

Daniel

got an error IPAdapterUnifiedLoader IPAdapter model not found. everything is installed, PLUS FACE (Portraits) is selected and in the list.

Florian Scherer

Thank you very much! Unfortunately, I just can‘t make the custom nodes work as I still receive the import fail error. I do not use the portable Version of comfyui. Could anybody provide some info on how to solve this?

Arbaaz Nasser

Does these workflows work on comfyui.exe. I have been facing end of road issues with all previous versions so I am tinstalling comfyui.exe again(fresh installation) then linking the models/custom nodes to the old directory,

Roid Mommy

Thank you very much Mick! Much appreciated!

Alyndiar

New member. Thx Mick. I want to explore how to create more poses and emotions from a single image, with consistent face AND costume. And I need help with creating loras from these images. Especially the tagging/caption and parameters for Kohya_ss. From your videos, I think you're pretty good at all of this.