~20 min
Output
narrated lecture video
tool · 2025 · client unnamed
A PDF of study material in, a narrated twenty-minute lecture video out.
~20 min
Output
narrated lecture video
AR / EN
Languages
None
Database
stateless by design
The problem
Turning written study material into a narrated lecture is hours of manual work per topic: writing a script, recording narration, building slides, assembling video. The inputs and outputs are both structured enough to automate.
The approach
A staged media pipeline with no database, because the job has no state worth persisting between runs. Text extraction from the PDF, a language model writing a slide-by-slide teaching script, speech synthesis for narration, slide rendering, and assembly into a finished video. Each stage produces a checkable artifact, so a failure is localised rather than requiring the whole run again.
Architecture
Staged pipeline with artifacts between stages and no persistent store. Stateless by design, since the only durable output is the video itself.
Each with what it cost
A decision without a stated tradeoff is a preference.
01
Why
A pipeline whose input is a file and whose output is a file has no state to keep. Adding a database would create an operational surface for no benefit.
What it costs
No run history or resumability, the correct trade for a job measured in minutes.
Outcome
A finished narrated lecture video of roughly twenty minutes generated from a source PDF, in Arabic or English.