Internal AI Product · Transcript Infrastructure
Flick
Flick turns a YouTube URL into editable source text and natural localized outputs. The product keeps the human in control before translation and protects the workflow from brittle transcript provider failures.
01 · The problem
Why it exists
A transcript product is only useful when retrieval survives provider changes and the source text can be corrected before errors multiply across languages. Flick was designed around that failure path instead of treating it as an edge case.
02 · Ownership
What I built
I built the product interface, transcript flow, localization experience and guarded recovery layer as an internal company tool.
03 · System
How it works
The workspace retrieves a transcript, lets the operator edit the source, then produces up to four parallel localized outputs. A five-stage recovery path keeps the workflow understandable when the primary retrieval route fails.
04 · Evidence
What is proven
- 52 supported languages
- Up to four parallel outputs
- Editable transcript before localization
- A five-stage guarded recovery path
- Internal use with no public product link