Viridian's AI Experiments

[Viridian]

Some interesting Runway results.

[Acidtester]

They all cool, but I really love the Vicky ones!

[KTG_42]

Amazing what AI can accomplish vs just a year ago. Nice work.

[Viridian]

RunwayML had an update. Now, it's possible to add image WITH a description, so that Runway doesn't randomly do stuff with it. It's still touch and go, but the results are far less random and closer to what I ask for. It's a huge step forward, closer to key frame technology, though you only get to choose the starting frame. What it does fill out, when it gets it "right", is genuinely amazing. This is perhaps the closest to bringing still images to life yet, especially DerMudGecko's drawing.

These are just sample animations. I can incorporate my current ability to generate AI voice and lip sync if I could put a sequence together.

[sixgunzloaded]

Yep, definitely a big improvement. It looks to be distorting faces a little less too. It's getting there! Nicely done.

[Viridian]

Some of the better outputs. The animations of the mud interaction with clothing and body parts is mindbogglingly good.

[JSample]

"Susan Desert" looks especially impressive/immersive!

[Viridian]

An animation of my Tomb Raider 3 India scene. We're starting to get pretty close to good results.

[sixgunzloaded]

Some of the better outputs. The animations of the mud interaction with clothing and body parts is mindbogglingly good.

It really is! Very fluid. And it held the faces even longer in this batch!
"Susan Desert" was really exceptional with how much she visibly sank and the way her hair started to fan out near the end.
And it's way cool to hear Lara's British accent and understated snark. lol
Sweet! Great batch!

[MDLambert]

I gave up after a while, so much nightmare fuel lol