Replicate

Just published the last little things changelog round up of 2024. Lots of bug fixes, tweaks, and ongoing improvements to the platform. Take a look at what we've been up to the last couple weeks at https://lnkd.in/dfVQVHR7

AI video is having its Stable Diffusion moment. OpenAI's Sora made everyone realize what is possible, and now there are loads of models that are just as good. Some are even open-source, so you can tinker on them, fine-tune them, and build upon them. Read more in 🧵 Sora reset expectations about what a video model could be. The output was high resolution, smooth, and coherent. The examples looked like real video. It felt like we’d jumped into the future. The problem was, nobody could use it! It was just a preview. This was like when OpenAI announced the DALL-E image generation model back in 2021. It was one of the most extraordinary pieces of software that had been seen for years, but nobody could use it. This created all of this pent-up demand that led to Stable Diffusion. Now the same thing is happening with video. There are tons of models from Hailuo, Tencent, Runway, Luma, Haiper, Genmo, Lightricks, and more. Some focus on realism, some focus on style and creativity. Some are designed to be really fast. Some are even open source. The community is modifying them, optimizing them, and building upon them. You can fine-tune them with new styles, objects and characters, and more. Read more on our blog: https://lnkd.in/dNiFXYgA

You can now monitor how much GPU memory your deployments are using so you can make sure you’re using your GPUs efficiently. View your deployments to see the latest data on your GPU memory usage. https://lnkd.in/ghhGmM3k

Two more models from @recraftai are now up on Replicate. These 20B models are faster and cheaper than their flagship V3 model. Try them out: https://lnkd.in/dcU4RTJT ($0.022 per image) https://lnkd.in/deS5SH9N ($0.044) You can use the API to add them to your app or website: https://lnkd.in/d3dH_XmB https://lnkd.in/dvFzt9zw We really like the vector outputs.

Black Forest Labs just dropped Flux Tools, for pro and open-source dev models: - Fill: Inpainting and outpainting - Redux: For image variations - Canny and depth controlnets They're all really good and you can run them on Replicate: https://lnkd.in/gXUrdpr8 More details in 🧵 Flux Fill's text inpainting is 🔥 Just mask the text you want to change, add a prompt, and it matches the original typography perfectly! https://lnkd.in/gsPAjmwk Flux Redux is also exceptionally good. You can use it to create variations of existing images, or use a prompt to restyle them. If the original contained text, Redux will usually preserve it excellently. https://lnkd.in/gj_BtkwZ Here's a variation on the "This is fine" dog. Canny and depth models are also best in class for Flux https://lnkd.in/g4jWQYSu https://lnkd.in/gkUp-9VU Here's a video game screenshot with photorealistic render using the Canny model

From Docker to diffusion models: This was a fun discussion with Ben Firshman and Matt Bornstein about what has / has not changed in developer tools over the past decade or so. Ben built Docker Compose and helped drive mega user growth there, and how has firsthand insights into how developers are building with generative AI models via Replicate. https://lnkd.in/gPrMQS5r

Hi, do you want to join us at Replicate to help make AI easy and accessible for all developers? Our Platform team builds the infrastructure that thousands of developers rely on to create and serve their AI models on. Do you have previous platform experience and are excited about how to build one to handle the specific needs of Generative AI? Do you want to dig into how to scale heterogeneous GPU workloads across multiple cluster? How to maximize utilization and uptime? How to fairly and efficiently distribute work? If our paths have crossed before, it would be great to catch up! If not, and you're interested in working with us, we'd love to see your application through the jobs portal below.

Train your very own custom AI models 🎨✨ 🧙♀️ Using Flux on Replicate - No coding required Then, use the model to generate 👉 Avatars, clothing, product shots, and more! Reply with a 👋 and we’ll send you the full BuildShip workflow!

Ben Firshman, founder and CEO of Replicate, joined Dylan Fox for our 3rd installment of Assembly Required. In building Replicate, Ben’s had a front-row seat to how researchers and developers are building with AI and the changes and trends emerging in this space. In this talk, he shares: - how unlocking AI innovation and user adoption starts with unlocking barriers to building with AI - why the stage of a company is a critical factor in how teams build with AI - the evolution of fine-tuning models and which use cases are best suited for customization Watch their full conversation here: https://lnkd.in/eAzXWSC9

San Francisco! Our very own Miguel Lara will be speaking at Replicate's AI Art Show tomorrow, October 1, at the Replicate hackspace. The hangout starts at 5pm. RSVP here for location details: https://lnkd.in/gBKZ2AMv