Sabitlenmiş Tweet
Nate Raw
3.3K posts
Nate Raw
@_nateraw
machine learning hacker @splice. previously @huggingface @lightningai
Austin, TX Katılım Şubat 2017
1.6K Takip Edilen9.6K Takipçiler
After months of work, today I’m releasing Foundation-1.
A SOTA text-to-sample model built specifically for music production workflows.
It may also be the most advanced AI sample generator currently available - open or closed.
• ~7 GB VRAM
• Entirely local
• 100% free
😁
English
I dont think, as a student, you shouldnt pay 3$/hour to rent a gpu to study blackwell arch.
More like you should be able to submit small jobs remotely where 100 people share one b100.
for student projects, each job will probably take < 10 seconds to finish. Submission will probably cost 1 cent.
Who is fixing this?
English
@sarahookr Few spots for you - enjoy your trip :)
Paprika ATX
Discada
KG BBQ
Desnudo
Comadre Panadería
Saigon On 7th
English
Nate Raw retweetledi
Our latest paper, LatCHs (Latent-Control Heads), is out today!
Really happy we'll get to present it at ICASSP’26 in Barcelona :)
> Guidance-based controls
> Selective TFG
> Latent-Control Heads (LatCHs)
🗄️ arXiv: arxiv.org/abs/2603.04366
👾 blogpost: artintech.substack.com/p/latchs-expla…
English
@_lyraaaa_ What are you doing to the poor model to make it do this 😂😂
English
crazy to me how much ppl will spend doing architecture and training method search but avoid doing data work
English
My "Hello World"s of Music ML:
2015: Essentia w/ SVM on GTZAN
2017: madmom RNN on Ballroom/Beatles
2019: VGGish embeds on MagnaTagATune
2021: Demucs on MUSDB18
2023: DDSP timbre transfer on NSynth
2025: Stable Audio Open for TTA
English
Guy behind him is cooler 🤷♂️
hardmaru@hardmaru
@pmddomingos Spotted you at the Music Workshop 🎹
English
My favorite question from today's AI music workshop at NeurIPS:
"AI can help make music creation easier, but how can we use AI to make music creation even harder?" -@dadabots
English
Nate Raw retweetledi
this is what a koan sound track looks like to music AI
[umap of stable audio latents]
English
Nate Raw retweetledi
LoRA training works well for Stable Audio. Even just a few hours of training. Stoked to bring yall into this artform for 2026
Mat Dryhurst@matdryhurst
My strong feeling is that training models is an art form we are a short hardware leap away from being ubiquitous. I don't care in the slightest about universal or disney, generating pastiche movies or genre music, but I do want the freedom to work in the medium of our time to be available to everyone
English
🫡🪴
Logan Kilpatrick@OfficialLoganK
Thrilled to welcome @fofrAI to Google DeepMind and the AI Studio team as a Senior Prompt Engineer!! Our generative media models (Veo, Nano Banana, Lyria, etc) are reshaping the industry, such a special moment to make sure everyone can get the most out of them, and more!
ART