Award-winning filmmaker. The Last Waltz - Winner, Dreamina AI Animation Summit at Annecy 2026. Dark, strange, and occasionally funny films.
Joined April 2020
- Tweets2.9K
- Following281
- Followers538
- Likes30.9K
Pinned Tweet
The Last Waltz trailer is here.
A game of cat and murder.
It started with a single image: a ballroom at night, all shadow and silhouette. The image became a concept reel. The concept reel went to Annecy and came back with a win. Now the film has a trailer.
Three hundred guests. A fortune in roses. A bride dead by eleven.
Watch it below. And when the police arrive — remember that everyone was watching, and no one saw a thing.
The Last Waltz, an animated pink noir murder mystery, currently in production at WellerVision. Winner, Project Proposal Track, Dreamina AI International Animation Conference, Annecy 2026.
Love this. Love the color palette. Love the story line too. I do not love the sharp lines. But good on them for bringing a classic back.
Can we talk about how important this is? The obsession to build and create. To see ideas through. Sometimes you just wanna see your Frankenstein.
Not into luxury goods. But I am into Sylvain Chomet. Dior went all on a beautiful story told by one of the great animation directors.
On a snowy winter’s night in 1949, all is quiet at 30 Avenue Montaigne until the phone rings: Marlene Dietrich needs a new dress…
Created to celebrate the holidays, “Christian Dior’s Night in Color” is a new animated short film by award-winning French director Sylvain Chomet. Narrated by Robert Pattinson, it features the voices of, in order of appearance:
Josh O’Connor as The Chair
LaKeith Stanfield as Ron the Director
Louis Garrel as Christian Dior
Nine d’Urso, Nadia Tereszkiewicz and Greta Lee as The Bees
Rihanna as The Lucky Star
Camille Cottin as Marlene Dietrich
Just before the tech gets to where you can snap your fingers and AI does it all - I’m buying a spindle, dragging my new epigenetically 33-year-old body to the top of a mountain, and spending my nights looking at the stars and remote viewing my way across the universe.
I think about this sometimes when I'm chasing data points around the web and being sandblasted with info from llms.
Is there a back alley dealer giving out access to Fable 5.5 somewhere? And can someone put me in touch with them?
It’s happening.gif
Fable 5.5 its over for video and image models 🤯
THIS IS PURE CODE
one shotted this video holy shit , how this is more beautiful than ai video models
Two ways to check this yourself:
Ask: “Where did @karpathy go in 2026?” with web search and memory turned off. Only the new Fable knows the answer. Then give it the prompt.
Run the same prompt I used for the test (it’s in the screenshot) and see for yourself.
Don’t rely on Grok for basic reasoning like this. Grok doesn’t have up-to-date knowledge of niche stuff. Either join the community or run the test yourself.
Also, a knowledge cutoff doesn’t mean a model knows everything before that date. Sometimes a model doesn’t know about an event even if it happened before its cutoff. This test shows that too.
Run the same prompt on regular Opus 5.5 and the quality drops. On Fable 5.1 it’s worse.
Three.js merged a meshlet pipeline. It splits every 3D model into tiny patches and lets the graphics card decide, patch by patch, what to draw, skipping anything hidden or too far away to matter. So a scene can hold huge amounts of detail and still run smoothly.
Experimenting with making films in code using three.js, I've learned every code animation needs its own kit of assets. The kits are about to get bigger and sharper.
This is what I mean by planning for a leap every 3 to 6 months
I was wondering when Three.js was going to get something like Nanite.
And then I saw the PR for a meshlet rendering pipeline was merged <24 hours ago!
github.com/mrdoob/three.js/p…
Here's a demo scene rendering 1B triangles at 150FPS.
raw.githack.com/sunag/three.…
This is huge! 🤯
OMEN. A coded film.
Small, fast-moving organic shapes using a difficult aesthetic was the test. It was better than I expected with best results toward the end.
Started in Opus, finished in Astra. Two filters were used. A paper finish to strengthen the textured look. And an application of diffuseRoughness from the @threejs r187 build to enhance the lighting surface.
No way 😂. Moat? What moat?
I recreated subvocal in 9 minutes with one prompt for AirPods Pro 3.
Source: github.com/joeblau/subvox
Creating via code hits so very different. It's 100% fun and creativity. Video gen has come a long way and will continue to improve but bringing a story to life without prompting frustrations is just mmmm mmm good.
Not every coding animation project works out. The method here was to create the location using three.js, create the characters using image models and trace the motion from a clip rendered with a video model.
This method worked previously but in those examples the character did not speak and the clip was more than 720p. In this case keeping facial features distinct owing to the low resolution and fineness of the lines was a struggle.
It also struggled with the applying shadows and colors accurately for a small fast moving character.