goyo

@hugothomel Wowow😯

you can now fight dragons in a world model! mixing cutscenes and real-time video generation makes for a real fun experience try stormcaller here 👇

New blog from our community! @aktasbatuhann, @Swan_aes, and @promptlyconfuse introduce us to the Kids Multilingual Benchmark, a 2,312-prompt, 23-language benchmark for child–AI conversations that evaluates 4 production models & validates the LLM-as-judge pipeline with 5 independent judges. 🚸

@ltx_model lipsync workflow with external audio based on lipdub lora, here are the sample results

(7/7) Built something with LTX-2.3? Drop it in the replies. Best one gets featured next week.

(5/7) World building 🤝 LTX-2.3

been testing @LTXStudio lipdub lora for lipsync with external audio and i've to say, im quite impressed with the initial results i got after testing in @ComfyUI

still new to e2e inference optimization but quite happy with the progress i made with @bfl_ml klein4B distilled model. on an rtx6000 pro, i could bring the latency down from ~324ms to ~128ms (2.5x speedup) for a 576x384 image at 4 steps. I started with the compile side wins for DIT and VAE, then worked on caching and using faster attention. Also, i tested nvfp4 weight + activation quant along with a lighter vae at the end and these helped in bringing latency down a lot without any noticeable quality loss. btw all testing was done on @modal gpus

@demian_ai @nebiusai Hey dylan, not able to DM you.

We are aggressively hiring at @nebiusai. If you see a job and you think you are a match, my DMs are open careers.nebius.com

@_alyxya hey can you pls share the repo?

first time trying to run nanogpt speedrun and it costs ~$12 in modal credits, I don't think I'm going to be able to afford many experiments here

@MokadyRon been experimenting with FiboEdit + character consistency lora - really good results so far. i've put together a @modal deployment for it: github.com/Yuvrajxms09/Fi…

Fibo-Edit is live 🔥

@aktasbatuhann @bria_ai_ @modal haha thanks, i think you might like this one too github.com/Yuvrajxms09/mo…

@Swan_aes @bria_ai_ @modal another must try cookbook, great work!

Fibo Edit — Character Consistency just deployed FiboEdit by @bria_ai_ with lora(from bria too) on @modal to get consistent characters. repo - github.com/Yuvrajxms09/Fi… Input PC - @tatyana-doloman-728740365/" target="_blank" rel="nofollow noopener">pexels.com/@tatyana-dolom…

@_alyxya great read and visuals are very clean btw

new blog post on the first principles of ML alyxya.com/blog/ml-is-opt…

really amazed by how far we’ve come in a year! last year I fine tuned SmolLM-135M on WinoGrande; yesterday I did the same with Monad-56M with the same setup and got almost identical results at half the size. Great work 👏👏 @pleiasfr @Dorialexander