Dall-E, Midjourney, Stable Diffusion (etc) - who's playing, and with which?
- Started
- Last post
- 851 Responses
- PhanLo0
New phone based vfx system seems to use iphone metadata to build a HDRI for realtime lighting in the scene.
-
https://simulon.com
https://twitter.com/Simulon
-
- PhanLo2
text to 3d model, kind of wonky
-
https://www.meshy.ai
-- wonky but kewlutopian
- I think populating a world with bad models would be fun. :-)
The figures it was making were pretty messed up, but buildings and cars were a bit betterPhanLo - Made a rusty mecha https://i.imgur.com/…PhanLo
- I just can't get my head around learning yet another AI app. This is madness. I don't know how you have time to learning every last AI app.utopian
- I find it pretty overwhelming, but also quite exciting too. Made a wee seamless loop as a base tonight https://i.imgur.com/… need to draw it againPhanLo
- Often I'll abandon things once a better thing comes along, but it's happening almost week to week, big changes.PhanLo
- PhanLo2
RunwayML has just mad a video to pointcloud thing you can animate in the browser. Had a play with the demo models and they work really smooth.
-
https://twitter.com/nickfloats/s…
-
Quite liked this video:
- i_monk0
Trying to remember a Google Colab AI that sampled uploaded images and generated new images based on them. Sound familiar? Or any new site that uses multiple images as seeds?
- PhanLo0
street photos into videos
-
- PhanLo0
Finally something a bit more than Mixamo for generative motion in Blender and Unreal: https://www.motorica.ai
-
- PhanLo0
- wutNairn
- They're like newer versions of NERF's pretty cool tech https://twitter.com/…PhanLo
- epicsted
- i_monk1
Answering my own question above: Looking Glass (ruDALL-E)
v1.5 is the latest I can find: https://colab.research.google.co…
- PhanLo1
Midjourney to Unreal (head explodes gif)
-- Fuckin Bananas Bro!utopian
- That is pretty wild, it's been in development for ages though. I hope it's real and coming out soon :-)PhanLo
- pretty sick effects form a single image.utopian
- Defo! We're working on a project just now that it would be perfect for, but I think we've ended up getting game guys to build the world to fly through.PhanLo
- Folk were joking about generating immersive worlds in 10 years but it seems to be moving so fast now.PhanLo
- There's a cool example here https://twitter.com/…PhanLo
- Jeezusscarabin
- maikel-1
So, for you guys/girls who are actually intelligent... have you found any way to connect (directly via API ideally) a small sample input (say 1-4 selfies) to create a entirely new person based on the ones you uploaded? I remember some very early Dall-E models from OpenAI that allowed that kind of stuff.
- not MJ but training an embedding for SD is fairly easy. training is built in to SD tools like invoke.ai but there are sites like https://www.astria.a… thatkingsteven
- do it all and let you download the embedding as ckpt. runway has training tools too.kingsteven
- thanks man!maikel