“No faces found” In MacOS

[ZZZJX0912]

First, confirm

• I have read the instruction carefully
• I have searched the existing issues
• I have updated the extension to the latest version

What happened?

I downloaded ComfyUI on my MacBook and was able to run the ReActor workflow successfully. However, the face swapping process failed. No other error messages were displayed; the only prompt was "face not found".

Steps to reproduce the problem

There are no special steps, only the simplest face-swapping nodes.

Sysinfo

MacBook Pro with Apple M1 Pro

Relevant console log

got prompt
[ReActor] 02:15:55 - STATUS - Working: source face index [0], target face index [0]
[ReActor] 02:15:55 - STATUS - Analyzing Source Image...
[ReActor] 02:16:02 - STATUS - Trying to halve 'det_size' parameter
FETCH ComfyRegistry Data: 10/107
[ReActor] 02:16:09 - ERROR - No faces found
[ReActor] 02:16:09 - STATUS - Analyzing Target Image...
[ReActor] 02:16:16 - STATUS - Trying to halve 'det_size' parameter
FETCH ComfyRegistry Data: 15/107
[ReActor] 02:16:16 - ERROR - No faces found
[ReActor] 02:16:16 - STATUS - No source face(s) in the provided Index
[ReActor] 02:16:16 - STATUS - --Done!--
[ReActor] 02:16:18 - STATUS - Restoring with GFPGANv1.4.pth | Face Size is set to 512

Additional information

No response

[ZZZJX0912]

pip list

onnx                       1.19.1
onnxruntime                1.23.2

[Buumcode]

[ReActor] 02:16:16 - STATUS - No source face(s) in the provided Index

Reactor can't find face on this image.

Did Reactor works with other Source image?

[ZZZJX0912]

This is really confusing for me. Face swapping works when I use a different image, so what could be the reason for this?

[ZZZJX0912]

The same workflow was able to work normally on the GPU cloud server before.

[Buumcode]

This is really confusing for me. Face swapping works when I use a different image, so what could be the reason for this?

This face is too close to camera - takes all territory of frame (and don't have enough space AROUND face) - so program detect this image as bunch of different objects.

Make more space AROUND face.

[wiredhamster]

Same issue for me. Happens on ComfyUI running on Windows with a RTX 5060 Ti (16GB).

It's the same for every source / target image that I have tried. It has never detected a face, regardless of the images or their resolutions.

[Buumcode]

It's the same for every source / target image that I have tried.

Try this ZIP (this is copy of Reactor I've use myself, 100% working Reactor):

comfyui-reactor-node_Reactor_v0.6.1-b2+Hyperswap+Restore-filter+VHS-fix+Float-fix_GITHUB.zip

Delete your Reactor folder from custom_nodes, and unzip this (as example):
C:\ComfyUI\custom_nodes\comfyui-reactor-node

Restart your ComfyUI.

[Buumcode]

Trying to halve 'det_size' parameter

"Trying to halve 'det_size' parameter" — Program can't find any face, and trying to reduce 'face size' in half - for trying to find very small face.

Wtf is going on...

Did you try other Facedetection options? YOLOv5l?

[wiredhamster]

Try this ZIP (this is copy of Reactor I've use myself, 100% working Reactor):

comfyui-reactor-node_Reactor_v0.6.1-b2+Hyperswap+Restore-filter+VHS-fix+Float-fix_GITHUB.zip

Thanks for your reply. I tried installing from your ZIP, but it seems to be missing the files from the root dir, i.e. init.py, install.py etc. So it didn't show up in Comfy.

Wtf is going on...

Did you try other Facedetection options? YOLOv5l?

I get the same result selecting all face_detection models:
retinaface_resnet50
retinaface_mobile0.25
YOLOv5l
YOLOv5n

But - I get the impression that in the background it is always using buffalo_l.

[Buumcode]

I tried installing from your ZIP, but it seems to be missing the files from the root dir, i.e. init.py, install.py etc. So it didn't show up in Comfy.

Here is screenshot from this ZIP. It have all init.py, install.py and all other files. It's full working copy of node:

But - I get the impression that in the background it is always using buffalo_l.

Of course, because Buffalo models maded by Insightface uses for face detection and other works. May be you need to re-download this models files (your file is corrupt)? Try re-download buffalo_l from here: https://huggingface.co/vladmandic/insightface-faceanalysis/tree/main

[wiredhamster]

Here is screenshot from this ZIP. It have all init.py, install.py and all other files. It's full working copy of node:

Thanks for the reply, and you are quite right. I'm not sure how my original download got corrupted - but all files are present in the zip.

I tried again with the zip you provided and redownloaded buffalo_l. Same result unfortunately.

Next steps would be to try deleting all other custom nodes. If that fails then reinstall Comfy - although I am not keen on this as it takes a lot of work to get it up and running again.