Skip to content
malekverse
All missions

AI pipeline2026

Edah Studio

Turns a written lecture into a narrated Arabic and English video, without letting the model invent facts.

Role
Solo: design and engineering. Pilot for an education platform.
Year
2026
Stack
Python, FastAPI, React 19, three.js, edge-tts, Playwright, FFmpeg
Edah Studio cover: the live product on screen, above a dotted planet

The problem

Teachers have written lectures, and students want video. Generating the video is the easy part. Making sure it doesn't add a date, a statistic or a whole scene that was never in the lecture is the hard part.

What I built

  • A studio in Arabic, right to left: paste or upload a lecture, choose the identity and voice, then review the result scene by scene.
  • A 3D lip-synced presenter, with 10 Arabic voices (2 of them Libyan) and 8 English voices.
  • 19 types of animated visuals, synced to the narration, with bilingual subtitles.
  • Frame-accurate MP4 export: every animation is a pure function of time, so the preview and the export match frame by frame.
  • Per-scene human review: approve, comment or re-voice before export. Plus a question bank generated from the same source.

How it earns trust

Numbers must exist in the source

A quality gate deletes chart and statistic values that don't appear in the original lecture. Plausible isn't good enough.

Scenes are re-anchored by content

Each scene is matched back to its source paragraph by what it says. The model's own paragraph numbers are never trusted.

Self-correction, then a verbatim fallback

A scene that fails the checks is regenerated. If it still fails, the source text is used verbatim, and a coverage check flags any paragraph that was skipped.

Tested with the worst model available

A small local model (Llama 3.2 3B) invented whole scenes. Every one was removed or reverted, ending with zero flagged scenes.

By the numbers

layers of faithfulness checks
5layers of faithfulness checks
animated visual types
19animated visual types
neural voices, AR and EN
18neural voices, AR and EN
flagged scenes after the adversarial run
0flagged scenes after the adversarial run

How it fits together

  1. LectureLLMScene writer
  2. Scene writerdrop unsourced numbersNumber gate
  3. Number gatematch to sourceRe-anchoring
  4. Re-anchoringapprove per sceneHuman review
  5. Human reviewframe-accurateMP4 export

01

The studio

Arabic and right to left: paste a lecture, review every scene, export the video.

Edah Studio in Arabic: the 3D presenter over a title scene, the scene list, and the script, visual and source tabs with an approve-scene button
Each scene keeps its source text next to it and needs approval before export.
Edah Studio new course page: paste or upload the lecture text
A new course starts from the lecture itself: paste it or upload a Word, PDF or text file.
Edah Studio evaluation report in Arabic comparing the pipeline with a traditional video production
The pilot's report, comparing the pipeline with a traditional shoot.

02

What it makes

A lip-synced presenter, live subtitles and one of 19 visual types per scene.

Edah Studio player in English: the presenter beside a title slide, with the course chapters listed
The same demo lecture in English, with its chapters.
Edah lesson scene: a bar chart comparing solar panel efficiencies
A chart, with only numbers the lecture actually gives.
Edah lesson scene: a mind map of a home solar system
A mind map.
Edah lesson scene: the photovoltaic effect in four steps
A process in four steps.
Edah lesson scene: a check-your-understanding question
A check-your-understanding question.

What I'd do next

Turn the adversarial run into a standing benchmark, so every prompt or model change is scored before it ships.

malekverse

© 2026 Malek Maghraoui. Built with Next.js, three.js and GSAP.

Every color on this site is a star temperature, from 3,200 K to 30,000 K. There are no green stars.