facefusion for me doesnt work right, when i do a video it doesnt do all the frames done that in the
facefusion for me doesnt work right, when i do a video it doesnt do all the frames done that in the last couple of versions
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=40)
Donly does so many
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=16)
Ftry next version
Flatest one
DI will afer rope i remember i had to use rope which had no problem
Drope live v10 ?
Dk here we go
Dphew tensort cu12 done
Ddownloading all model files please wait patiently 
Darghhhh
Dits working! 
Dno notification in cmd about now running on local host
Dah fantastic now thats much better
Dmuch faster
Dloaded the folder within a few secs
Dits even cropped dogs faces
D[ERROR:0@296.711] global obsensor_uvc_stream_channel.cpp:158 cv::obsensor::getStreamChannelGroup Camera index out of range
Dspoke too soon
Fthis is why i dont trust kohya to resize my images : https://github.com/kohya-ss/sd-scripts/pull/1426#issuecomment-2334986320
GitHub
cv2.lanzous4 implement has some problems.
Fhe just fixed resize :d
Dstory of my life Doc lol
Danyway to bed i go
Dnight
Fnight i also need to go 
DDoes it make things better? How to set it up?
DSo you use <segment:yolo-face_yolov9c.pt,0.7> instead of <segment:face,0.5>?
DDo you need to download it separately or it downloads on its own on upgrading swarm UI?
DMan, I LOVE SWARM UI. I haven't gone back to comfy ever since it started working for me.
DIt's like A1111 on steroids lol
Fyuou nee dtgo download it
Fput into yolov8 folder
DI can't find that folder in the root. Do I create it somewhere?
Fytes create it
Fit is useful since segment masks animal face too
DWhere can I find this parameter 'Distilled CFG Scale'?
DIs this same as Flux Guidance Scale?
DGuys, try iPNDM sampler with flux guidance 3.5, it's really good.
MSomething is wrong with the Flux generation. I typed this prompt "a flaming heart on a black background" and got a lousy result, I don't know why. I use Forge. Whatever sampler I try, I always get a blurry picture
Mif I set a flux1-schnell-Q4 model, I get a good picture.
�set CFG to 1
Mcfg on 1, when I switch to flux, it automatically sets the forge
Mthere must be something with the euler sampler, because if I generate flux1-schnell with Euler and simple, the result is still lousy
Mif I generate the same with DPM++ 2M Karras:
MSo the Flux models are still interesting as I look at their possibilities.
�You are using the nf4 model I have not used it yet. I use dev fp16 model with unipc sampler and normal scheduler, CFG scale 1 and Flux guidance scale 3.5
MIllyasviel clearly recommended the NF4 model for generation, so I downloaded this version for Forge and it is really fast. But really, I haven't tried it with FP16 yet. I will try it!
https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/981
https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/981
GitHub
(Before we start, Forge now supports UI presets like this. Click this GIF to zoom in.) (Again, before we start, to the best of my knowledge, I am the first one who made the BitsandBytes low bit acc...
�what gpu you have? Are you on windows or cloud?
MI have RTX 3090 on windows
�I think you can use fp16 model but will be a bit slow
