Workflow Venice AI Workflows and Models (DON'T POST REAL PEOPLE!!)

wan 3 is incredible with reference video. i dare say it's better than peak grok times (except for grok making images, nothing comes close to that yet). what I do is i give it a reference video and an image of the woman. then e.g. for a blowjob video i just say "the exact woman in @image1 but she is giving a blowjob like in @ video1" - some results were mind blowing. i always use 480p as 720p is not worth the price personally. sometimes it struggles with clothing, but it's usually easily fixed in prompt e.g. "the woman from @image1 is wearing something similar like in @ video1".

we are so back. now all we need is an image model as good as peak grok
 
what about the audio? How do you make it better? Or do you just generate without any audio?
 
I have good results using Wan 3 as an image editor. just set Wan 3 R2V to only 2 seconds and then grab a frame from that tiny clip. at the moment it's the best way to get good pussy when nudifying
 
yes. It has a much higher success rate in replacing women in porn vids. And it does it with less extensive prompting
 
If you want your gens to be cinematic or professionally shot, then Seedance 2.5 is the way to go. Wan 3 is more consistent with less dangling penises. I'd say Wan 3 is better because it costs less credits, and it produces some good shit
 
any tips?, it always cut and a new scene from an image and start a new scene or sometime start with the image
 
As already explained twice... the videos need to be made in 480p (at least the ones without added video in the references); if I do them in 720p, it turns into a mess...
 
I've noticed that Wan 3 R2V doesn't handle drastic cuts in scenes that well. I suggest replacing women in individual scenes and then splicing the cuts together later. I just put in a woman I know into the entire striptease scene from True Lies. And in HD no less
 
Has anyone figured out how to get deep penetration with WAN? Or how to get a deepthroat BJ, but with the woman controlling the action?
 
For Wan3.0, abit frustrating that I can't use a longer reference to generate a longer video (only 15 sec of reference for 15 Seconds of video).

I tried using 10 seconds of reference, to make 20 seconds of video, but I realise the 11-20 seconds the video would be largely doing nothing.

Anyone tried this and actually got the 11-20 seconds to be a realistic extension of the reference video actions?
 
since most of the discussion seems to be around video, do we all consider image generation to be 'solved' by Seedream v5 Pro now? Is anyone using something else thats better? I'm too broke for video gen lol
 
Seedream is weeker when it comes to pure editing, but if you want to recreate a porn scene from a reference video its fantastic
 
has anyone else noticed seedream becoming less permissive? It was allowing everything a few days ago
 
Can anyone who has used Dzine share an opinion how it is compared to VeniceAI? How is VeniceAI better, except maybe that video generations seem cheaper?
 
How long is your reference video? For some reason, it doesn't work for me most of the time. I tried a 5s video and it worked once. Could it have to do with my shitty upload speed?
 
You must log in or register to reply here.
Back
Top