This was great. I've used Reactor quite a bit in A1111 but never realised it was available in Comfy. This workflow is basically a dedicated face swapping solution without all the rest. I'll definitely be using it.
Hmm, this is frustrating. It keeps failing to import the node through the manager. I've tried several times and Comfy is up-to-date. I'll give it another look in the morning but any ideas I can try would be welcome. 🙂
@@matwin2034 Sorry I didn't but then I rarely get anything to work in Comfy, which is part of the reason I don't use it. Far, far too many hassles for it to be worth the trouble, at least for me when I can do almost all I want (including this) in Forge.
very nice tutorial, but un fortunaltly i couldnt install sucessfully, i got error this... Getting requirements to build wheel: finished with status 'error' error: subprocess-exited-with-error pls suggest what shall i do
@@BlenderBob this video is fairly old in AI terms, and that may be the issue. I know that reactor has had several updates since then. There’s probably a better way to get going with reactor these days. I might want to consider updating that video.
Quick and easy, thanks for sharing this workflow. Only thing missing is upscaling, but I guess that could be a separate workflow after you cherry-pick the best images first.
Oh, there is so much you could add to this. But for this, I was really just trying to keep it super simple. It’s very easy for me to start adding things and adding things and defeating the purpose of the demo. 😄
Understood. I've seen numerous workflows and the vast majority are overwhelming. The simplicity and speed of this is perfect, and I'll be using it quite a bit. I'll probably add an upscaler at the end that'll be bypassed by default, and then switch it on when it generates something I really like. Thanks again!
Hi! I followed your instructions but I am getting a black square showing up in preview image area. The black square is showing up where the face swap should be. It is black widows body and over her face it is suppose to be Loki face but instead there is a black square over the face area. Do you know how I can fix this? I am not sure if I can attach an image to show you what I have on my screen?
1:13 thank you for sharing🙏🏻 I have the issue that it says "loading fail" next to ReActor in the Manager and the whole ReActor column is red. Tried everything, uninstall, repair, update. Nothing works. Any idea why i can’t install ReActor?
I don't know if your nodes are out of date or whatever is going on with this node, as I have two extra options on my node "face_restore_visibility" which I set to 1 & "codeformer_weight" which I set to 0.5
The models are based on human faces. Some animal faces can switch with humans from time to time, but it’s more of a fluke. Same with other animals. There may be some models trained on animal faces, but I’m not aware of them.
121 Gourieff (IMPORT FAILED) ReActor Node for ComfyUI The Fast and Simple 'roop-like' Face Swap Extension Node for ComfyUI, based on ReActor (ex Roop-GE) SD-WebUI Face Swap Extension Getting this error message when I try to install
I get this error? Prompt outputs failed validation ReActorFaceSwap: - Failed to convert an input value to a FLOAT value: face_restore_visibility, None, float() argument must be a string or a real number, not 'NoneType' - Failed to convert an input value to a FLOAT value: codeformer_weight, no, could not convert string to float: 'no'
Swap the node from this workflow with a fresh one. You can find it in Add Node - ReActor. I had the same error and I think it's because Bob's ReActor is out of date and we've got newer versions of that node.
im getting this error now? Error occurred when executing ReActorFaceSwap: [ONNXRuntimeError] : 7 : INVALID_PROTOBUF : Load model from F:\AI\sdxl-comfyui-ultimate-workflow\ComfyUI_windows_portable\ComfyUI\models\insightface\inswapper_128.onnx failed:Protobuf parsing failed.@@Elwaves2925
I’ve been hearing that a lot recently. I wonder if something changed? Everything is still working for me, but I have an updated that note in a while either.
Had the same issue... only way I could get it to work was to install the portal version of ComfyUI and install it into that directory... it worked fine from there. Let me know if that helps.
Have you Bob or anyone experimented with swapping all the detected faces in the source image into the faces inside target image? It works on indexes right now, which makes it quite rigid.
@@BobDoyleMedia I eventually succeeded. But then I get this error Prompt output validation failed ReActorFaceSwap: - Failed to convert input value to FLOAT value: face_restore_visibility, None, float() argument must be a string or real number, not 'NoneType'. - Failed to convert input value to FLOAT value: Codeformer_weight, None, float() argument must be a string or real number, not 'NoneType'.
Yes, they've updated the node. You'll need to put a value in from the face_restore_visibility (like .8 or something) and also a weight value. You'll also need to choose a value for gender detection and maybe on other - can't remember off hand, maybe source and input values? I know someone shared it exactly in another comment, but I'm unable to find it.@@lilillllii246
If you're asking why a very high resolution source face image doesn't translate to that resolution in the final image, I'm personally not 100% sure why. It must have something to do with the resolution it can sample the image, and there may be a way to adjust that. I just have to rely on the GAN to sharpen the faces, but I totally agree that ideally it would be better informed by the resolution of the original face. One day, I'm sure...
it is because the model being used (by reactor) has been restricted. It will always be low quality. You need to run it through a ksampler with segs or something similar.
Too bad all of the face swappers (which are all basically branded versions of insight_face) only do front on or near-front on face angles. If the head is more than 45 degrees in any direction then it just looks bad. Hopefully the geniuses of the world are working on new tech that adds ability to do side and rotated angles.
That simply isn’t true across the board anymore. I’ve done tons of face swaps from various angles that look amazing. It’s come along way in the past couple of years. I’m not doubting that that’s your experience, but I have found great success with a lot of models at various angles.
@@BobDoyleMedia Hmm, that's good to hear, though I've yet to see an example on any of the videos or sites where a face looks good from significant angle. Do you have a link to any per chance?
I've followed all instructions on installing Pinokio. Downloaded ComfyUi. But i continue to get the same error message "ENOENT: No such file or directory, stat 'C:\Users\Comp\pinokio\api\comfyui.pinokio.git\{{input.event[0]}}' Very frustrating because i can't find a answer anywhere.
Wow, I really wish I could help you! I’ve only had my own experience with this, and I have not seen that error. I really hope someone reading this can help you!
this is crazy how great the results are, ty so much!
When trying to install the missing custom nodes I get an error saying - import failed. How can I solve this? Thanks!
Hey Bob, This is what I was looking for for a month ... thanks a lot..... Thanks for the workflow as well, It's a gift....
That ending was awesome!
This was great. I've used Reactor quite a bit in A1111 but never realised it was available in Comfy. This workflow is basically a dedicated face swapping solution without all the rest. I'll definitely be using it.
Hmm, this is frustrating. It keeps failing to import the node through the manager. I've tried several times and Comfy is up-to-date. I'll give it another look in the morning but any ideas I can try would be welcome. 🙂
@@Elwaves2925 yep same cant get the ReActor node to load/install... Did you end up getting it set up?
@@matwin2034 Sorry I didn't but then I rarely get anything to work in Comfy, which is part of the reason I don't use it. Far, far too many hassles for it to be worth the trouble, at least for me when I can do almost all I want (including this) in Forge.
That last part was the best :)))
It was awesome! What was the workflow used for the faceswap (not the voice) to create this clip? Elevenlabs for voice
very nice tutorial, but un fortunaltly i couldnt install sucessfully, i got error this...
Getting requirements to build wheel: finished with status 'error'
error: subprocess-exited-with-error
pls suggest what shall i do
and also i subscribed cause i like your tuts!
Thanks for the clip! I can't get reactor installed. In the manager I see Import Failed with a Try Fix button but it doesn't fix anything. Any ideas?
@@BlenderBob this video is fairly old in AI terms, and that may be the issue. I know that reactor has had several updates since then. There’s probably a better way to get going with reactor these days. I might want to consider updating that video.
@@BobDoyleMediaI fixed it. It was the Python version that needed to be updated. Thanks. :-)
@@BlenderBob Great! Thanks for letting me know!
this is amazing! Thanks for sharing. I agree with the warning in description. :) So much fun!!!
Quick and easy, thanks for sharing this workflow. Only thing missing is upscaling, but I guess that could be a separate workflow after you cherry-pick the best images first.
Oh, there is so much you could add to this. But for this, I was really just trying to keep it super simple. It’s very easy for me to start adding things and adding things and defeating the purpose of the demo. 😄
Understood. I've seen numerous workflows and the vast majority are overwhelming. The simplicity and speed of this is perfect, and I'll be using it quite a bit. I'll probably add an upscaler at the end that'll be bypassed by default, and then switch it on when it generates something I really like. Thanks again!
Hi! I followed your instructions but I am getting a black square showing up in preview image area. The black square is showing up where the face swap should be. It is black widows body and over her face it is suppose to be Loki face but instead there is a black square over the face area. Do you know how I can fix this? I am not sure if I can attach an image to show you what I have on my screen?
1:13 thank you for sharing🙏🏻
I have the issue that it says "loading fail" next to ReActor in the Manager and the whole ReActor column is red. Tried everything, uninstall, repair, update. Nothing works. Any idea why i can’t install ReActor?
Wow!! D see Ed easiest best face swap I see now
Hi
I did some images that metods, but in finnaly image just face is little blurry or pixels, is there any fix for that ?
Did you use any face restoration? Be sure to choose "Face Enhanced" and then play with the values. I set it to around 40%.
cool stuff
how can i do it with video, it never works for me
nice and simple
Hmm... What if we wanted to retain more of the hair + facial hair of the input face?
ipadapter plus face
Great Stuff
I don't know if your nodes are out of date or whatever is going on with this node, as I have two extra options on my node "face_restore_visibility" which I set to 1 & "codeformer_weight" which I set to 0.5
I want that node! I’ve been trying to figure the best way to adjust the restore, visibility, and I couldn’t find that. Thanks for the tip!
Thanks!!!
can you do face swap with animals? if so does it look like a real animals face or does it turn it into a weird hybrid??
The models are based on human faces. Some animal faces can switch with humans from time to time, but it’s more of a fluke. Same with other animals. There may be some models trained on animal faces, but I’m not aware of them.
Sadly the results for me are extremely pixelated. Is there a way to fix that?
Yes, use the Face Enhancement. GfpGAN 1,4
121 Gourieff (IMPORT FAILED) ReActor Node for ComfyUI The Fast and Simple 'roop-like' Face Swap Extension Node for ComfyUI, based on ReActor (ex Roop-GE) SD-WebUI Face Swap Extension Getting this error message when I try to install
And you’re not able to resolve this in the ComfyUI manager by installing any missing nodes?
Hmm, I left you a reply with something of a solution for your problem but now I can't see it.
Did you get it?
what is the solution?
@@Elwaves2925
@@Elwaves2925if you used the name of a particular code sharing site, RUclips hides all of those messages. It's ridiculous.
It happened to me as well..import failed...what is the solution you replied before, please
End of the video got a sub, so good lol
Thanks!
thanks for the excellent video
I get this error?
Prompt outputs failed validation
ReActorFaceSwap:
- Failed to convert an input value to a FLOAT value: face_restore_visibility, None, float() argument must be a string or a real number, not 'NoneType'
- Failed to convert an input value to a FLOAT value: codeformer_weight, no, could not convert string to float: 'no'
Swap the node from this workflow with a fresh one. You can find it in Add Node - ReActor.
I had the same error and I think it's because Bob's ReActor is out of date and we've got newer versions of that node.
im getting this error now?
Error occurred when executing ReActorFaceSwap:
[ONNXRuntimeError] : 7 : INVALID_PROTOBUF : Load model from F:\AI\sdxl-comfyui-ultimate-workflow\ComfyUI_windows_portable\ComfyUI\models\insightface\inswapper_128.onnx failed:Protobuf parsing failed.@@Elwaves2925
Never mind fix it, had to replace my inswapper_128.onnx so its working so far@@Elwaves2925
Cant get reactornode to install for my comfyUI... no matter if i try with the manager or do it manually from github or with the zipfile :(
I’ve been hearing that a lot recently. I wonder if something changed? Everything is still working for me, but I have an updated that note in a while either.
Had the same issue... only way I could get it to work was to install the portal version of ComfyUI and install it into that directory... it worked fine from there. Let me know if that helps.
Have you Bob or anyone experimented with swapping all the detected faces in the source image into the faces inside target image? It works on indexes right now, which makes it quite rigid.
I have not tried that. I just do multiple passes. Doesn't really take all that long, but something more advanced would be nice!
Thanks. Your json file( workflow file) won't open in comfyui. All other workflows open.
Hmm. I wonder if something changed. No one else has told me this, but I can go back and see if that json works for me!
@@BobDoyleMedia I tried google colab. is it impossible?
@@BobDoyleMedia I eventually succeeded. But then I get this error Prompt output validation failed
ReActorFaceSwap:
- Failed to convert input value to FLOAT value: face_restore_visibility, None, float() argument must be a string or real number, not 'NoneType'.
- Failed to convert input value to FLOAT value: Codeformer_weight, None, float() argument must be a string or real number, not 'NoneType'.
Yes, they've updated the node. You'll need to put a value in from the face_restore_visibility (like .8 or something) and also a weight value. You'll also need to choose a value for gender detection and maybe on other - can't remember off hand, maybe source and input values? I know someone shared it exactly in another comment, but I'm unable to find it.@@lilillllii246
I also tried replacing ace_restore_visibility,codeformer_weight with numbers, and no with no, and the result is the same.
why egzactly even if you put 2 great resolution pictures quality is so bad??
If you're asking why a very high resolution source face image doesn't translate to that resolution in the final image, I'm personally not 100% sure why. It must have something to do with the resolution it can sample the image, and there may be a way to adjust that. I just have to rely on the GAN to sharpen the faces, but I totally agree that ideally it would be better informed by the resolution of the original face. One day, I'm sure...
it is because the model being used (by reactor) has been restricted. It will always be low quality. You need to run it through a ksampler with segs or something similar.
facefusion interface is sucks!!!!! thank you for your quality content 👑👑👑
I don't have ReActor installed on win10
You should update your ReActor, it now features a builtin Fidelity and Visibility sliders ✌ Sometimes i feel CF with 1.0 Fidelity works better!
Yes, I’m hearing from a lot of people that I’m behind on my reactor version. Excited to update!
I see no difference and sometimes Gfan is better. Maybe I am doing it wrong?
Can anyone help me installing reactor?
I always get issues and really troubleshooted with a lot of people and forums already.
I can cashapp! :D
Too bad all of the face swappers (which are all basically branded versions of insight_face) only do front on or near-front on face angles. If the head is more than 45 degrees in any direction then it just looks bad. Hopefully the geniuses of the world are working on new tech that adds ability to do side and rotated angles.
That simply isn’t true across the board anymore. I’ve done tons of face swaps from various angles that look amazing. It’s come along way in the past couple of years. I’m not doubting that that’s your experience, but I have found great success with a lot of models at various angles.
@@BobDoyleMedia Hmm, that's good to hear, though I've yet to see an example on any of the videos or sites where a face looks good from significant angle. Do you have a link to any per chance?
Can you suggest a few that's in ComfyUI? sorry to be that guy
@@MangaGamified For ComfyUI just install the ReActor Face Swap nodes... very easy and works as well as anything out there.
Cool would be to use only code!
I've followed all instructions on installing Pinokio. Downloaded ComfyUi. But i continue to get the same error message "ENOENT: No such file or directory, stat 'C:\Users\Comp\pinokio\api\comfyui.pinokio.git\{{input.event[0]}}'
Very frustrating because i can't find a answer anywhere.
Wow, I really wish I could help you! I’ve only had my own experience with this, and I have not seen that error. I really hope someone reading this can help you!
@@BobDoyleMedia I appreciate the reply. Thank you.