So you think AI isn't going to take your job?

[cristhiancordoba]

@randy

The links above are for Mac only

Not really. I have it running on RL9.3 flame. Give it a try.

[aaronneitz]

Welp, we have 2 prominent clients who insist on using AI to do some VFX on their spots. Here we go!

[RufusBlackwell]

Good luck! It’s fun, but it’ll take longer than you think. I done tons of it, any issues, hit me up, happy to help out.

[cnoellert]

@randy

…ia hardware, at the end of a very specific image manufacturing pipeline, and happen to be the best problem solvers and creative minds in post.It’s like what….3 lines of code to install Comfy? 2 lines in a terminal to setup a venv? 7 clicks to download and move Flux Dev into /models?Fuck, even the workflows under the Browse Templates is enough to move the needle forward as a compositor’s companion.

Just an anecdote, I’m in a hotel room in Chicago. I’m running Flame and Blender on my laptop and Safari’s logged into linux hardware back in LA. Currently it’s running 80 iterations of a Flux in-paint Comfy setup at 20 samples which I’ll prune-down to 5 selects.

The Comfy output directory is linked back into Lucid which means as each iteration completes I immediately have access to access to them here in the hotel room. Once I’m back from Lollapalooza with my daughter (Bleachers and The Dare yesterday!!!) I’ll throw my 5 selects back for higher samples (the output pngs contain the setup and seed for for the job) and then into Runway or something else to breathe some life into them, comp them back into the plate, post and sleep.

I’m not a pro like @RufusBlackwell but I get it and ultimately you can’t ignore the wave we’re being forced to ride. It’s a hideous dialectical feeling of being both emboldened and empowered while simultaneously feeling fucking awful and truly scared for whateverthefuck is going to happen tomorrow.

[BrittCiampa]

I have the same feelings @cnoellert Ive been figuring my way around generating depth and normals because those are just cool things and I love the idea of generating aovs from 2d footage. Or building meshes out for either proxy geometry or geo for projections just from a few frames. Very helpful in terms of quality, not necessarily quantity.

In terms of complicated paint there’s a lot of stuff that seems really time intensive for stuff that can just get sent to India and have someone do a great job and just not worry about it. The outsourced work is no less an ethical quandary though. Some of those places are hellscapes. Shadow misery mechanical Turk kind of stuff. So we’ve just shut eyes and ears to that because of ‘necessity’ and the nasty world of ever-shrinking margins. The AI robot doesn’t eliminate that suffering though. Just transposes it. Even if it’s not someone working in a roto and paint sweatshop, there’s someone in Kenya helping to train and scale models by looking at the most disturbing imagery imaginable and trying to decide if it’s a kid or not in there.

The neoliberal nightmare never really diminishes, no matter what hurdles it may appear to be stumbling on, it just changes shape and form and ends up in the shadows where it’s nice and shady and cool. We’re all just trying to get our shots out and pay the rent. Hard to know where lines are in a fuzzy fuzzy world.

[allklier]

Good read:

https://www.wsj.com/business/media/disney-ai-hollywood-movies-5982a925?st=5pvJRv&reflink=article_copyURL_share

[GPM]

Thanks for sharing this article, Jan. Disney is my biggest client and so far the rule from King Mickey is “no AI.” I don’t expect that ruling to remain for long.

[RufusBlackwell]

Interesting AI related test. We had a AI generated BG plate that we had to comp some CG in to. It’s tricky to estimate the lens info from a photo, but even harder from a AI generated image. What I did was take the AI image in to Kling, prompt it to dolly in, so we get movement and parallax in the scene. Then clean it up in Topaz and camera track in Syntheyes. Syntheyes I have found to be very forgiving, and to be good at estimating lens info. Then take the camera in the exported FBX, bin the movement info, but keep the camera lens info, and use that for the CG comp. Worked quite well.

[allklier]

Excellent solve for this problem.

[cnoellert]

My fave solve for this is Fspy.

Depends on the frame but it can get you extremely close. The Blender add-on imports it directly and then you can push an fbx to Flame.

[andymilkis]

@RufusBlackwell

…r and makeup do you? You want a proper stylist offering ideas and that being fed in to the eventual end product.That central role — managing and integrating all those creative inputs — is the new online artist.I’ve currently got three major AI/VFX projects on the go, and the work is some of the most creatively rich and technically interesting stuff I’ve done. Honestly you should be getting stuck in.

I love this take, Rufus. Thank you, and thank you for sharing!!

[BrittCiampa]

This is cool! I’ve been using Blender a lot lately in my day to day, another reason to have it in the dock. Love me a good free software that works well!

[Maurypb]

Google interactive promptable worlds

Genie 3: A New Frontier for World Models

[philm]

@Maurypb - Hey Genie3, Make America Great Again… Take your time…

[ytf]

Just droppin’n this here:

48K views · 856 reactions | Now playing on social media… “MAGA BALLROOM 2028...

[philm]

I don’t have a facebook account but i’m sure it’s great

[snacks]

Phil, you can cancel the login page and view the video. Worth the watch.

[philm]

@snacks - thanks brother, i’m knee deep in something which has equally negligible value to the planet as a whole. i’m sort of selfish that way.

[AamirQ]

Please forgive me if this has already been posted here:

With Google Earth + Weavy + Veo3 | Rory Flynn | 128 comments

[pixelmixer]

@ytf

Just droppin’n this here:

Outstanding. Chills down my neck.