Askable

You can Control-F a document. You can't Control-F a video. When the answer you need is buried in a two-hour recording, you drag the scrubber and hope — and then you quote from memory. Askable makes any video askable. Paste a YouTube URL or upload a video file, and it becomes a source you can interrogate in plain language, where every answer cites the exact second it came from as a link that jumps you straight there. How it works: Bright Data fetches transcripts for YouTube videos, Speechmatics transcribes uploaded files, so a video works with or without captions. The transcript is chunked with its timings, embedded through AI/ML API, and stored in Postgres with pgvector on Supabase. Questions are answered by two Deno Edge Functions — no third-party key ever reaches the browser. One thing ordinary RAG gets wrong is counting. "How many times do they say government?" cannot be answered from the eight closest chunks, because the model never saw the rest of the transcript. Askable recognises those questions and scans every segment literally instead, returning an exact count and a jump link for each hit. If the literal search misses, it falls back to semantic search rather than wrongly answering "never mentioned". Ingestion never blocks the app: transcribe one video while chatting with another, and conversations survive navigating away. It is for anyone who has to quote a video correctly — journalists, researchers, analysts, students, support teams. The entire product, frontend to edge functions to schema, was built with Native.builder.