Imo people are sleeping on the in context stuff from #nanobanana pro, i put my whole dataset in for a lora i was training on qwen image edit, and it outperforms the lora one shot.

We’re hiring 3 researchers for the Stability AI 3D team — and with our new EA partnership, this is an absolutely massive opportunity. If you’re passionate about 3D, graphics, AI models, or VLMs consider applying below:

@StabilityAI @jampani_varun @esx2ve @zqevans arxiv link: https://t.co/L1jDDfx6JQ

@StabilityAI You can find more comparisons and the paper on the project page: https://t.co/68yckLzNhR Thanks to @jampani_varun Simon Donné @esx2ve Julian Parker and @zqevans

@StabilityAI We find competitive results compared to concurrent work on the text-video-to-audio task on a range of metrics.

@StabilityAI Architecturally, we insert cross-attention blocks to modify the frozen base Stable Audio DiT, alongside pooled vjepa-2 embeddings for the video condition.

@StabilityAI Our model works with both video-to-audio and text-and-video-to-audio

Introducing our new work, "Foley Control: Aligning a Frozen Latent Text-to-Audio Model to Video" It's a control model for Stable Audio to generate aligned audio from an input video. Project Page: https://t.co/68yckLzNhR Paper: https://t.co/MK3HFl10CY 🧵 @StabilityAI

Today we’re launching Stable Audio 2.5: The first audio model built for enterprise-grade sound production 🔊 Audio influences brand engagement by 86%, but few enterprises are leveraging audio as an extension of their brand, making customized sound an untapped differentiator.

Sudden realisation that my expertise in both ml research and game engines puts me in a dangerously relevant skill set to embodied ai interfaces.

How today felt #o3pro

I’m currently exploring new job opportunities🧑‍🔬 My work revolves around text-to-3D with Geometry Images, generative UV mapping, multi-view models for texturing, and other genAI applications in graphics. I’d love to discuss how I can contribute to your research efforts!

The arxiv link will be coming in a day or two. What next? The model works amazingly although my limited budget as of recently means this is very much proof of concept and massive improvements will be coming very soon hopefully including much wider range of styles to work with.

I have set up a demo here to give it a try: https://t.co/TSXHiWSlEK

We find excellent results on a wide range of digital art style images, working best with in distribution style images for the training data.

The key idea was to jointly train an autoencoder, encoding input image embeddings down to a small latent and a control model on synthetic style data.

Introducing my new work: Stylecodes! An open source sref implementation for sd! Paper here: https://t.co/XHSkU4eDfB Code here: https://t.co/n2QVbsdmDS Stylistic control using only an easily sharable 20 digit base64 code! #stablediffusion #midjourney

https://t.co/bC5KcOoxKg

Also, come to poster #309 in the afternoon poster session and chat about the PBR texturing of objects.

#ECCV2024 hours, with my poster :D