Universal-3 Pro by AssemblyAI — Speech-to-text that finally understands context
Speech-to-text that finally understands context
Promoted
Maker
📌
Hey Hunters! 👋
I’m Snehal, builder of Distill.
We all know the drill: You launch a product, and suddenly your inbox, Discord, and Slack are flooded with feedback.
User A: "The login button is broken."
User B: "I can't sign in."
User C: "Auth error 500."
To a human, these are the same thing. To a database, they are three different rows.
Existing tools (like Canny or Upvoty) try to solve this by forcing users to "Vote" on a public board. But public boards create bias, invite toxicity, and force you to expose your roadmap before you're ready.
We took a different approach: The Invisible Intelligence Layer.
Distill is a headless backend that ingests raw feedback from anywhere (your own widget, API, or email) and uses Vector Search to cluster it automatically.
How it works:
Ingest: You send us raw text via a simple API call.
Cluster: Our semantic engine groups "Login broken" and "Auth error" into the same bucket instantly.
Signal: You get a dashboard showing Issues by Velocity, not just a list of emails.
Under the hood:
We are running on a heavy stack of Convex (Vector DB) + Gemini (Embeddings) to make sure the clustering is "self-healing." As new feedback comes in, the clusters evolve.
We are launching today in Beta with a generous Free Tier for indie hackers. I’d love for you to throw your messiest, unstructured feedback at it and see if it breaks.
Roast my design in the comments! 👇 And leave feedback in the app itself so I can also test it is working perfectly HAHA.
– Snehal
