For anyone interested here is the link: lipdub.ai
The way their model works is you have to upload 1min or less footage of your character speaking (Doesn't need to have audio - can be synthetic)
It then trains a mini model on your character to do the lipsync
LipDub AI Test
Undeniably best in class for lipdub / lipsync
..... but who the hell is paying these prices?! Each gen costs ~10-15 creds so for $150 USD a month you can make 15 videos.
Audio was txt2speech from Hailuo on a custom voice I trained
Random question:
Of the content I pose (media / highlights tab) - what outputs would you be most interested in generating yourself?
Is there a specific thing I make that most of you would want to make?
Meme_template.exe loading….
Back to my roots to end out the year 😂
$WARPER
BlueSky is most def having scaling issues....
Pushing a video from my X post straight through their API fails ~75% of the time
how is
BlueSky is most def having scaling issues....
......... it quite literally works maybe 30% of the time through my X -> Post over there
how is
BlueSky is most def having scaling issues....
......... it quite literally works maybe 30% of the time through my X -> Post over there
how is
FLUX Redux with my RUBBERHOSE Lora
Works pretty well! (link in first reply)
A new era is upon us!
Redux (REMIX)
Depth / Canny guidance
Inpainting
blog.comfy.org/day-1-suppor...
blackforestlabs.ai/flux-1-tools...
Black Forest Labs FLUX released "Tools"
Fill = Inpainting / Outpainting ✔️
Depth / Canny ✔️
Redux (think Midjourney's Remix..... but better) ✔️
All FULLY supported in ComfyUI on Day 1
Links and examples below 👇
Happy Thursday 🧸
Don't ask.... you wouldn't get it
Scratch that - got video working too
And y’all thought we’d have to pay Runway for stable vid 2 vid
This is fully local
Fully open source
Fully FREE TO USE
(Video created by 👑 Kijai)
If I can confirm it works without a pile of memory leaks and vulnerabilities yes for sure
Update on my X -> Bluesky automated Chrome extensionNow works with video, images, and text!https://bsky.app/profile/aiwarper.bsky.social/post/3lbft56hk3w2o
Mint it
Followed
Some early testing of this opensourced "Act-One"
There are so many knobs to turn and test but it's incredibly promising
I can get it better still this is just a couple hours of tickering - I want 1:1 expression transfer.
Stay tuned!
The good news is - I wrote a custom Chrome extension to push all of my tweets to Bluesky going forward 🙂The bad news is - Bluesky API doesn't accept video uploads from their API ☹️
The 🐐Kijai with another incredible, fully open sourced tool than you can run locally 🥵
This is a creative use case for these tools that I haven’t seen anyone do (or at least do this well) yet.
PJ converted a hand drawn anime into a live action trailer nearly shot for shot!
OpenFlux "De-Distillation" Schnell model from Ostris
Kijai compiled it into a single file fp8 model here
Here is the original for context
Just me asking @aakashsastry and the @HotshotSupport team to kindly release the SDXL v2 motion model weights that are likely sitting on a shelf collecting dust never to be used again
The open source community could put this to good use
Pika 1.5 is out today with some new "effects"
Here I test one of their new effects called Crush It.
Maybe it's frozen? viscous? 🤣
Reached out to the community manager at Runway yesterday to inquire about my application(s) for a creative partnership
My brother in Christ it's been a literal YEAR since I applied for this
Three hundred and sixty five days - but at least they're making their way through 🤡
