Lookup turns raw footage into structured answers, proof clips, and automations with just a few lines of code. From counting people to compliance checks, it makes video searchable and programmable—this will help any app see.
I’m Ajay, co-founder of Lookup, and today we’re launching.
The problem:
Most vision products are narrow. They optimize for one thing — safety, checkout, traffic — and stop there. But vision is only valuable when it’s inside another product. As a superpower.
If you’re a dev or PM, you don’t want a “vision startup.”
You want your app to see. So it can improve metrics and outcomes on its own.
What Lookup does:
👉 Upload or connect video
👉 Ask a question in plain English
👉 Get structured answers — counts, timelines, events — with proof clips you can trust.
We hide the heavy lifting (detection, tracking, reasoning). You just call the API.
How it works:
Lookup runs a cascade of models (motion → detection → video-language).
We store results in:
– Object DB: what + where
– Temporal DB: when + order
Together, video becomes queryable like a database. See once, remember forever.
Who it’s for:
⚡ Product / FDE / Dev teams — add “eyes” to your app, build your app workflow on top of vision.
Why now: Cameras are everywhere. Storage is cheap. What’s missing is a layer that makes video useful for any product. Lookup is that layer.
Who we are:
We’re second-time founders in video AI.
– At OurEye.ai (YC W22) we scaled to 1,500+ deployments.
– At AvairSense, our AI camera hit 1,000+ homes.
We learned one thing: computer vision only wins when it disappears into the workflow. Lookup is that lesson, turned into a platform.
Playground:
Try it yourself. Upload a video, ask questions, see answers. No training. No labeling. Just truth from video.
Video-to-structured-data sounds incredibly useful for security, retail, and compliance apps. The "few lines of code" promise is appealing for developers. Main questions: accuracy with complex scenes, processing speed, and cost per video hour analyzed.
A promising API for embedding “vision” into apps easily. Excellent for prototyping and smaller use cases; a bit early-stage for mission-critical deployments.
The exact thing I was waiting to see in my home CCTV footage application. I have pets at home and people visiting home when I am out. I generally have to go through 9 hours of footage to catch 30s clip which drains me otherwise. Looking forward to check it out.
Also, does it process only real world video or can it process screen recordings. Curious to know if it can be integrated to get insights on remotely recorded screen sessions.
hey amazing product I often use Gemini 2.5 pro video understanding for the same task, I dint get it is it open source? if commercial how your API pricing fares against other LLMs
This looks cool!! Can we run this over a live stream of video and get structured output for further processing?
Congrats on the launch! 🎉 Super cool — feels like there could be plenty of use cases for this.
What if the video data is too large to upload or query via your cloud API. I'd imagine this indexed data could get unwieldy real fast. What if the organization wasn't comfortable sharing their sensitive video footage to the cloud? Do you have a plan in the roadmap to do all of the processing and indexing at the customer's edge or customer-managed cloud and giving them full data sovereignty and visibility into their indexed data without leaking the data and metrics to the big boys? I'm sure there are all sorts of privacy and multi-tenancy laws to adhere to if this system is hosted in the cloud especially for enterprise live cameras. For example, Illinois has stringent facial recognition and associated biometrics laws in place to analyze footage without consent
Will it be possible to upload a YouTube link (or any kind of link from a third-party platform) instead of only uploading the file?
Great launch! How does Lookup handle different video formats? Excited to try it out
very cool! how long of videos do you support right now? also, does it provide timestamps for given events?
What would you build in your app if you could integrate vision within your product?
Awesome Idea , and a very nice implementation
Keep it up guys !! Long way to go
Wow! This is what I was looking for. Excited to leverage the vision capability in our product. Let me loop in our tech team
👋 Hey Product Hunt!
I’m Ajay, co-founder of Lookup, and today we’re launching.
The problem:
Most vision products are narrow. They optimize for one thing — safety, checkout, traffic — and stop there. But vision is only valuable when it’s inside another product. As a superpower.
If you’re a dev or PM, you don’t want a “vision startup.”
You want your app to see. So it can improve metrics and outcomes on its own.
What Lookup does:
👉 Upload or connect video
👉 Ask a question in plain English
👉 Get structured answers — counts, timelines, events — with proof clips you can trust.
We hide the heavy lifting (detection, tracking, reasoning). You just call the API.
How it works:
Lookup runs a cascade of models (motion → detection → video-language).
We store results in:
– Object DB: what + where
– Temporal DB: when + order
Together, video becomes queryable like a database. See once, remember forever.
Who it’s for:
⚡ Product / FDE / Dev teams — add “eyes” to your app, build your app workflow on top of vision.
Why now: Cameras are everywhere. Storage is cheap. What’s missing is a layer that makes video useful for any product. Lookup is that layer.
Who we are:
We’re second-time founders in video AI.
– At OurEye.ai (YC W22) we scaled to 1,500+ deployments.
– At AvairSense, our AI camera hit 1,000+ homes.
We learned one thing: computer vision only wins when it disappears into the workflow. Lookup is that lesson, turned into a platform.
Playground:
Try it yourself. Upload a video, ask questions, see answers. No training. No labeling. Just truth from video.
👉 If your app could see, what would you build?
We’d love to hear. 🚀
Shoutout to @thisiskp_ for hunting us. Also @zaczuo and @rohanrecommends for the help.