Sieve is the cloud for video & audio AI. Leading product teams use Sieve's APIs, tools, and infrastructure to ship AI-powered capabilities faster, together.
🔎 Today we’re launching a redesign to the Sieve explore page!
- Faster, more accurate search
- Filter by function tags (understand, video, generate, audio)
- Function card improvements
Find exactly what you need to understand, manipulate, and generate video at scale. On sievedata.com/explore!
And that's a wrap to launch week 🚀
🌍 Sieve Dubbing 2.0 is here. Since launch, we’ve partnered with top video platforms who’ve helped refine the pipeline. Today, see why 2.0 is the highest-quality automated dubbing solution available.
Most AI dubbing still suffers from issues like awkward translations, monotone speech, and unnatural timing. These problems arise because dubbing involves transcribing, translating, cloning voices, and syncing speech, with small errors compounding at each step.
The biggest issues developers face with other dubbing tools:
1. hallucinations, where TTS models produce gibberish
2. speaking speed, as matching timing across languages often results in speech that's too fast or slow
With Ahmed Hanzala's lead on these efforts, we solved both! Learn more about the issues and see side-by-side comparisons in our blog (in comments).
Day 4 of launch week ✅
We are excited to announce our partnership with Kaiber as they launch Superstudio, an AI-native platform redefining how creatives interact with generative AI.
Kaiber partnered with Sieve as their AI infrastructure provider in April 2023 amid rapid growth. Sieve enabled Kaiber to run AI video workloads at scale while minimizing R&D time and reducing infrastructure spend by over 50%. So today, it’s with great excitement that we officially announce our partnership as they launch Superstudio.
Superstudio offers creators a highly curated selection of foundational models for image and video creation through an intuitive Canvas interface. Sieve’s flexible GPU infrastructure, pipeline orchestration, and pre-built components for video-focused use cases made it the perfect choice.
Check out their launch and learn more about the partnership (in comments section).
👀 LAUNCH WEEK DAY 2: The first generally available API for eye contact correction!
- Redirect eye gaze to focus on the camera automatically
- Costs $0.10 / min of video processed
- Process videos up to 8x realtime
Over a year ago, there were some impressive demos of this, but many required complex software or were too slow. We took inspiration from NVIDIA's Broadcast and LivePortrait's facial expression control to build this pipeline on Sieve.
Try it for free now and learn more from our blog (both links in comments).
🚀 LAUNCH WEEK DAY 1: A new playground interface for Sieve functions!
We’re kicking of launch week with a refresh to our function UI
- Easier to navigate layout
- Compare job outputs
- Tweak job inputs
…all on the same page
It’s going to be easier than ever for developers to find the right function parameters for their use case!
P.S. This was Sourabh Mane's first ship since joining Sieve. Our interfaces are only getting more slick from here on out 😎
Sieve is kicking of its first ever Launch Week next Monday where we’ll be announcing something new everyday for five days straight 🚀
It’ll include exciting new video AI pipelines, partnerships, and product updates.
Stay tuned!
We were already the fastest way to run SAM2 (2x faster than the next best in-fact). Now we support the newly released 2.1 checkpoints!
Meta frontend + Sieve backend soon?
🔥 OpenAI just released whisper-v3-turbo ahead of DevDay 2024. It's 8x faster than large-v3 with minimal quality degradation, and you can now run it on Sieve!
Try it via `sieve/speech_transcriber`: https://lnkd.in/gVmNREQ4
Access Whisper directly: https://lnkd.in/garUvi6x
It’s incredible that you can transcribe + summarize videos 50x faster than realtime 💨
Our out-of-the-box pipeline for this on Sieve just got a major update where we processed a ~19 min video is 22 seconds. Transcribed, summarized, tagged, and chapterized. It’s a simple, table stakes use case but it’s cool to see how new model optimizations get it to really blaze.