Sieve

As the year comes to a close, I’ve spent some time reflecting on our work at Sieve. We ended on a great note, signing some amazing new customers and seeing our existing customers expand into many new use cases on the platform. There are some use cases and teams I’d dreamt of supporting at some point, but couldn’t imagine to be supporting so soon. In my mind, there are two defining things that allowed us to make this progress: team and vision. We’ve built a great team who I am extremely excited and thankful to be working with everyday and have a vision of what we will tackle in the years to come that is clearer than ever. I’m excited to share more in the new year but if you’re interested in working on an early team building core infrastructure that touches some of the most forward-thinking industries in the world like media, communications, VR, gaming, robotics, and automotive — please reach out! We've hit an inflection point in growth and we’re hiring across research, engineering, sales, and marketing. 2025 will be incredible. Learn more about our work here: https://lnkd.in/gEVggFV9 You can email jobs [at] sievedata [dot] com, which will be forwarded to me directly.

Ball tracking is a capability that enables automated sports highlights, replays, and other cool statistics to be extracted from your favorite sports games. In this blog, we build a robust ball tracking system around Meta's SAM 2 model. We implement a combination of scene splitting, multi-frame prompting, prompt validation, and zero shot object detection to build a robust system that improves on common failure scenarios. Check out the blog here: https://lnkd.in/dbZqdpWA

Our CEO Mokshith Voodarla recently chatted with Steven Vonder Haar from IntelliVid Research about how Sieve is building a new set of primitives to understand, manipulate and create video. Today this is being applied by developers primarily in social and creative use cases, but we see these same primitives becoming key tools in powering the future of VR, gaming, and robotics as well. Watch the original video below or on YouTube: https://lnkd.in/egCH8X-t

Artificial Intelligence just may be enabling a whole new way to think about the concept of "video infrastructure." We typically think of hardware devices and network bandwidth as the building blocks that make digital video possible. We now may be entering an era where software development may be part of this infrastructure framework, as well. Start-up Sieve is in the business of developing Video AI features that other vendors can integrate into their own offerings. Don't want to develop your own green screen capabilities, plug-in an API from Sieve into your platform. Need better ways to handle translation and transcription, Sieve can have you covered there, as well. Sieve CEO Mokshith Voodarla visits the IntelliVid Research virtual studios to discuss his company's approach to developing API-friendly Video AI features and talk a little bit about the future of video infrastructure in the enterprise. #IntelligentVideo #SoftwareDevelopment #API #ArtificialIntelligence https://lnkd.in/gvfaWQ5i

Introducing the Sieve Moderation Suite: flexible, developer-friendly content moderation for text, image, audio, and video. - Support for 20+ text classes and 40+ visual classes - Single API call integration with self-serve access - Tune processing FPS and supply custom classes You can try it for free now and read the full blog (links in comments). Safety is critical to building AI applications, especially when dealing with rich media like images and video. We’ve built these pipelines in partnership with some of the largest video streaming, online shopping, and creative AI platforms in the world — all of whom wanted a more modern, tunable, cost-effective solution for moderation. Our solution also outperforms other leading providers on most categories. We’re excited to keep improving on this work as our customers continue deeply integrating AI into their products with safety in mind. Learn more in our blog and try it via the links below.

💨 You can now summarize YouTube videos in 72x realtime The ability to quickly summarize media content has become one of the most powerful use cases of LLMs over the past two years. In this blog, we discuss a few different ways to go about building a video summarizer and the various quality tradeoffs to be made while doing so. Learn more here: https://lnkd.in/e4dZ2P_5

🦃🍂 This Thanksgiving, we’re serving up something tasty: a demo pipeline that turns any YouTube video into a podcast-style AI conversation (like Google’s NotebookLM!). This video was fully AI-generated using that pipeline running on Sieve. Check out how we built it here: https://lnkd.in/gzBhN43S Check out the code here: https://lnkd.in/gpwqeFwh

YouTube is slowly starting to roll out automated dubbing directly into the platform, which changes the landscape for many marketing and creative video tools looking to offer similar functionality. More specifically, they'll need to focus on workflow or niche-specific features like more tunable language styles, granular editing, preferred words and more. Looking to build a more powerful version of dubbing over YouTube videos? Learn how to do just that in our new tutorial: https://lnkd.in/dGRTeVd2

A studio recording setup without a studio recording setup. This is what we find many video platforms trying to enable for their users, because it’s what makes the barrier to recording & sharing great video much lower. Some of the most common enhancements we see being applied include enhancing the audio, blurring the background, and fixing eye contact. So we built a simple demo of how developers can run all three of those processes in parallel on Sieve! The code is open-source and we have a live demo on Sieve you can try (both linked in comments).

Super excited to announce our partnership with ai|coustics to bring their leading audio enhance models directly into the Sieve ecosystem. ai|coustics offers two models both of which are now available via the `sieve/audio-enhance` pipeline: - Finch: an AI-powered subtractive model that effectively removes background noise, providing a clean audio signal ideal for noisy environments. - Lark: an AI-powered reconstructive model designed to repair and enhance audio by restoring lost frequencies and adding studio-quality clarity. This means: - No background noise - No room reverb - Clear and easy to understand voices - A reconstructive technology that preserves speaker identity - Easy customization for your level of speech enhancement You can try it for free now and read the full blog (links in comments).