Pedro Gengo 🐉 @Predogl

I was fascinated when I first heard about kernel fusion from @cHHillee's blog post "Making Deep Learning Go Brrrr From First Principles" (yes, I am a big fan of this post). I am still a novice in kernel programming so could not build a fused kernel myself. The `kernels` project from @huggingface came to the rescue. I could choose from the 100s of kernels on the Hub and profile it. In the blog, we use the GeGLU FFN fused Liger kernel, and the profile is beautiful. 🤩

It is profiling time! ⏰ In Part 2 we cover: > trace a Linear layer > talk about mul + add vs linear > gemm epilogues (my favourite part) > trace MLP > torch compile vs liger kernels > use hugging face kernels Bookmark this for a fun weekend read! 😎

Link to post huggingface.co/blog/torch-mlp… Thanks to @remi_or_ @SergioPaniego @RisingSayak @pcuenq for their vital reviews, and also @NoeFlandre and @Predogl for their reviews in the initial draft. Thanks for being critical and supportive of the effort. 🤗

Last chance to RSVP for our annual CVPR Friends Dinner in Denver on June 4th for conversation, cocktails and bites. RSVP at the link below.

"Otterly Lost" is a short film 100% generated by AI, based on prompting by me, Fabrício Carraro. This project is an experiment in pushing the boundaries of generative storytelling by leveraging modern AI workflows, which tells the story of Ollie, a river otter who's lost his favorite rock, and Pippa, his unamused otter wife. Every element of this short was generated using the following tools: - Video & Voice Generation: Produced entirely on @runwayml, using the Seedance 2.0 video model. - Background music: Generated on @ElevenLabs. - Sony Vegas for editing the generated chunks and music together.

Aleph 2.0 is here. Now you can edit a single frame in your video, preview the change and then Aleph 2.0 carries that edit across the rest of your video. Try it now in the new Edit Studio on web at the link below.

@ariG23498 @runwayml 👀👀

Today we're announcing Runway's expansion into Japan: we're opening a Tokyo office and investing an initial $40M in the country, as we build on what is already one of our largest and fastest-growing markets. We've tripled our enterprise customer base in the country in the last 12 months, with Japan becoming our third largest market globally. Companies like Yamaha, NHN and SoftBank are adopting Runway across marketing, advertising and creative content, and this expansion brings us closer to Japan's leading companies in gaming, media and robotics. Learn more about the initiative and the roles we're hiring for at the link below.

ICYMI: Over the past two weeks, we've published a series of in-depth looks into how we deploy frontier video and world models into production, across engineering, research, and trust and safety. From how we optimized GWM-1 to serve Characters (which I believe is one of the largest deployments of real-time autoregressive diffusion today), to reducing cold start times by 60x for our inference workers via GPU-to-GPU transfer of model weights, to open-sourcing our configuration system for training and evaluation configs, and finally, the continuous work we do to ensure child safety across our platforms. There are a lot of novel challenges we face as Runway increasingly becomes the world's infrastructure for generating pixels, media, and worlds, and we'll start sharing more of the behind-the-scenes in the weeks and months ahead.

Runway added more than $40M in net new ARR so far this quarter, and we're less than halfway through. The biggest growth period in the history of the company. Generative video has hit its inflection point. Enterprises like Amazon and Robinhood are using Runway daily to reshape how they tell stories and reach customers, on our frontier video and world models. Thank you to our team and our customers. The best is yet to come.

runway characters represents an important step forward in research: real time video gen that feels smooth, natural, and yet still consistent. Super proud of the team and this amazing accomplishment!

Runway @runwayml

4 months ago

Introducing, Runway Characters. Real-time intelligent avatars that turn the internet into a conversation. Deployable anywhere via the Runway API, Runway Characters can be customized in any way across every style. All with the ability to embed bespoke knowledge banks, custom

19 78 410 56K 298

Introducing, Runway Characters. Real-time intelligent avatars that turn the internet into a conversation. Deployable anywhere via the Runway API, Runway Characters can be customized in any way across every style. All with the ability to embed bespoke knowledge banks, custom voices and instructions. Start integrating Runway Characters directly into your apps, websites, products and services today. Available now at the link below.

We showed 1,000 people a series of videos. Some real and some generated with Gen-4.5. Over 90% of people couldn't tell the difference. Can you? Take the test at the link below.

Introducing Image to Video for Gen-4.5, the world's best video model. Built for longer stories. Precise camera control. Coherent narratives. And characters that stay consistent. Gen-4.5 Image to Video is available now for all paid plans.

Introducing our new frontier video model, Runway Gen-4.5. Previously known as Whisper Thunder (aka) David. Gen-4.5 is state-of-the-art and sets a new standard for video generation motion quality, prompt adherence and visual fidelity. Learn more below.

Reframe is here. Outpaint and resize any uploaded video, image, or #DreamMachine creation into every format you need. Customize and reposition freely on the canvas to make every version of your story fit perfectly. Goodbye cropping. Hello creative freedom.

@ariG23498 @karpathy @lusxvr @andimarafioti Amazing (as always)!!!

Introducing A Branding New ComfyUI Look and Native API Nodes! Seamlessly integrate 11 amazing online visual AI models into ComfyUI, including @bfl_ml Flux, @Hailuo_AI, @minimax_ai, @Kling_ai, @LumaLabsAI, @PixVerse_, @recraftai, @StabilityAI, @Google Veo2, @ideogram_ai, @OpenAI GPT image-1, and @pika_labs! 🔹One platform, 11 models, 65 nodes, all at once 🔹Chain the APIs with open models and workflows 🔹No hardware limitation!

Introducing #Ray2 Camera Motion Concepts in #DreamMachine — 20+ precision-tuned camera motions designed for smooth cinematic control and great reliability. Concepts compose with each other making hundreds of impossible new camera moves possible. Available now.

Ray2 Flash is now available in API.