Customize, deploy, and scale models on Serverless GPUs with the world's first Python Cloud.
External link for fal
Remote First, US
💨
Real-time AI painting is here. Visit https://fal.ai/dynamic to test it out. This is fully hosted on fal (Features & Labels)'s GPU infrastructure at massive scale. We already have a ton of builders building apps on top of this infra. Cant wait to see the results.
Try Flux on fal.
The "Midjourney Killer" open-sourced text-to-image model by our friends at Black Forest Labs is live on fal. https://lnkd.in/evJpCiN6 The model is state-of-the-art when it comes to text generation, photorealism and can do 4K resolution out of the box. BFL is the original team that worked on Stable Diffusion. Excited to see them continue their mission. Congrats on the launch Robin & Patrick!
Announcing fal Research Grants which provides free compute resources to researchers and developers working on cutting-edge open source initiatives.
Open-source AI is in jeopardy. At fal, we believe in the power of open-source and collaboration to drive progress in tech. Most of the innovation in this wave of AI is thanks to open-source projects. You don't have to go too far to see this. Just look at Meta's Llama, Mistral AI, Stability AI's Stable Diffusion, LangChain or Georgi Gerganov's GGML. We owe this wave to these projects. Today, we are announcing fal Research Grants which provides free compute resources to researchers and developers working on cutting-edge open source initiatives. Learn more here: https://fal.ai/grants Here at fal, we have historically been busy making models run really really fast. But in the last 6 months that we spent on hacking on open-source models, we realized something important: With a humble amount of compute, you can build amazing models. And this is why we decided to launch this program to support researchers building open-source models. As first recipients of the fal Research Grant, we have already given tens of thousands of hours worth of H100 compute to prominent open source models such as moondream2 and birefnet. We are super excited to see all the great projects that will come out of this. Send a DM or email [email protected] to apply.
fal reposted this
fal has launched 'Wizper', their new Whisper v3 offering and it blows past previous speed factor records with ~250X real-time audio speed 🌬 Near-instant transcription opens-up new use cases for natural-feeling voice chatbots and fal's Wizper is the fastest option currently available. fal charges based on inference time and as a result, the speeds drive a price of ~$0.5/1000 audio minutes, 20X cheaper than OpenAI's Whisper v3 offer! We have independently validated that Wizper maintain's Whisper v3's word-error-rate of 8% indicating no quality degradation. fal has let us know this is driven by underlying improvements to their inference engine and no changes have been made to the Whisper large v3 model itself. Our Speed Factor is benchmarked on our test of a 10 minute audio file. Different audio file lengths and particularly very short clips may experience different speeds. See below the link to our Speech-to-Text leaderboard 👇
fal reposted this
Today, fal's inference engine breaks another record. This time for running OpenAI's open source speech-to-text model Whisper. fal is 3 times faster than the runner up while still being the most affordable. Check out the full report by Artificial Analysis who runs independent benchmarks for inference performance across vendors. https://lnkd.in/eqUZy4s9 Follow fal to stay up-to-date on our progress for becoming the fastest and the most reliable inference cloud.
fal reposted this
𝗛𝗲𝗿𝗲 𝗶𝘀 𝘀𝗼𝗺𝗲𝘁𝗵𝗶𝗻𝗴 𝗳𝘂𝗻 𝘆𝗼𝘂𝗿 𝘆𝗼𝘂𝗿 𝗹𝘂𝗻𝗰𝗵𝗯𝗿𝗲𝗮𝗸. 𝗣𝗿𝗼𝗺𝗽𝘁𝗶𝗻𝗴 𝗳𝗮𝘀𝘁𝗲𝗿 𝘁𝗵𝗮𝗻 𝘆𝗼𝘂 𝗰𝗮𝗻 𝘁𝘆𝗽𝗲 🚀 https://fastsdxl.ai/ Just start by exchanging a word from the original prompt and off you go 💯 I always love to run into demos like these. Helps to fight your AI fatigue and is just a fun way to play around. This is also a good GenAI example to show to your (grand)parents, kids or non-tech-savy friends how AI prompting works. No login, no subscription, no complicated interface. This little demo app is using the lightning fast SDXL API, done by the team from fal. https://fastsdxl.ai/
Build a Real-Time AI Image App with WebSockets, Excalidraw, Next.js, and Fal.ai by Nader Dabit https://lnkd.in/eaANR_v2
🧢 ☯️ 🤗
☯️ at Hugging Face (hats courtesy of fal (Features & Labels))
fal reposted this
Playing around with React Native Skia Canvas + fal (Features & Labels) Real Time API So much power in this combination of libraries + APIs! It's open sourced at https://lnkd.in/gf7yuSCr
⚡
Did you know that fal (Features & Labels) has the fastest non-distilled SDXL endpoint? If you can find a faster one, we guarantee to beat it within one week. https://lnkd.in/eTYGxMrU