@Dr. Furkan Gözükara whats the best lip syncing method out there ?
@Dr. Furkan Gözükara whats the best lip syncing method out there ?
Mhas anyone here experimented with flow edit?
Mim basically trying to recreat the severance john wick video and i have gotten pretty close
VDid you see this .it's so mind-blowing....
M@Dr. Furkan Gözükara Replying to Ur Reddit - since i commented ill share with you the Flow Edit
MVid2Vid_flowedit.json24.28KB
MReddit
Explore this post and more from the comfyui community
Mbasically im trying to recreate this..
Mthe settings are weird to control
MSkip Step the higher u go the closer to the original input video - but you need to match is with scheduler steps at least 20 steps to gen.. anything below breaks it...
Mdrift steps also break it if its too high..
Mbalancing is really weird but plausable.. and gen time on my 3090 takes me 1min for 1080vid
Myou can resize to smaller and make it faster but you lose some of the detrails from this input vid
Mwhats annoying is there no inbetween
Mso far
Mhey @Dr. Furkan Gözükara, how can finetne a dfm model on my own face so that whenever i want to swap my face with other faces, it does it with more accuracy?
Fi dont know yet
Fvery likely
F5000 series is a scam
Fi will buy 5090 only since it has 32 gb vram
Fi shared video of it
Fbut not useable yet
FYouTubeSECourses
OmniHuman will shake the AI world forever if get published and works as they claimed. This is literally the ultimate video model announced so far. It is not published yet but I will be following it very closely. I hope that they will publish models publicly so we can use locally hopefully. As soon as published I plan to make APP for this model h...
Fthanks a lot
Fdeep face labs model?
Mi just found out prompting helps a lot
Fnice
Fdoing research for a client to train flux lora in 5 minute
T5090 would be 17% faster at fp16 flux dev inference compared to 4090 according to current reports
Not a deal breaker but nice incremental upgrade, plus 32GB vram for faster prompt switch
Not a deal breaker but nice incremental upgrade, plus 32GB vram for faster prompt switch
Ji run a 5090 but most ai apps do not yet support Blackwell architecture, new CUDA and pytorch version ect. but i am sure it will not take long
Fwell until i can test myself hard to tell 
Fexperimenting with 5 min training
Fbad so far :d
F
Fok i got some decent results
RTraining for 5 minutes?
D@Dr. Furkan Gözükara I had tried 'couple trainings' with images of them together in the input dataset. I captioned images using joycaption. Tbh, the model always learnt one person better than the other. What could be the reason for this? I used the same settings as your one -character dreambooths.
Fyes
Finteresting not sure why
RTargetting a few layers?
RWhat's the size of the LoRA file?
Fnope fully layers
F128 rank
Ffp16 saving : 1.27 GB
RThat's a small file.
RResemblance is good?
ROh, and how fast is the H200?
Freally fast
Fnot too bad but lower quality here
F
