Nobody reads a 30-page report to the end. Five Mila Agent skills turn it into something people watch and listen to: a video overview, a “live frames” clip, an infographic, a two-voice mini-podcast and a flashcard deck. Every fact comes from the source you gave.
For an expert, a school or a business that needs content for a channel and for training.
Each skill takes your material — a report, a lesson, an article — and makes its own format out of it.
Voiced scenes, a pencil sketch or the brand style, 16:9 and 9:16 from one script, subtitles, YouTube chapters, a cover.
video-overview:5No filming: on every important word an icon, a number or a diagram appears in the frame.
live-frames:9-10A NotebookLM-style sketch, a series of pages in a PDF, Uzbek in Cyrillic and Latin. No image generation available — a free branded HTML fallback.
infographic:5 infographic:53-61A lively two-voice ElevenLabs conversation about your material, an mp3 of 1–4 minutes with a loudness check.
mini-podcast:5 mini-podcast:40-42Cards from a report, a book, a lesson or a law: a source on every card, spaced repetition, a print PDF, Anki and Telegram quizzes.
flashcards:5The comparison comes from the skills themselves: this is how they were written, with NotebookLM in view. video-overview:13-21 video-overview:49-52 flashcards:11-13
| NotebookLM | Mila |
|---|---|
| A frame is a picture: an edit means regenerating the whole video | The script is JSON: change a line and rebuild; finished pictures are not redrawn |
| Static slides | A slow zoom into the picture, points appear with the voice, chapter titles |
| No Uzbek; vertical only in short English | ru · uz (Cyrillic and Latin) · en; 16:9 and 9:16 from one script |
| No subtitles or chapters | Subtitles in the frame, subtitles.srt and chapters.txt for YouTube |
| No brand | The brand style or the client’s style, a cover, an ending with a call to action |
| Flashcards: no spaced repetition, no link to the source, no printing, no Telegram | A source on every card, repetition, a print PDF, Telegram quizzes |
Every skill leaves files you can open, print or upload. video-overview:57-59 flashcards:15-23
| Skill | Files |
|---|---|
| Video overview | <name>-16x9.mp4, a cover, subtitles.srt, chapters.txt; the same for 9:16 |
| Infographic | PNG pages and a PDF made from the series |
| Mini-podcast | An mp3; the script prints the length, the average loudness and the peak |
| Flashcards | PNG 1080×1350 front and back, an A4 print PDF with 8 cards per sheet, trainer.html for the phone, Anki files, Telegram quiz polls |
The five skills live in the skills/ folder of the open repository.
| File | What it does |
|---|---|
| video-overview/SKILL.md | video overview: script, style, voice, formats |
| video-overview/scripts/vo_build.py | a plan with no spending first, then voice, pictures and the render |
| live-frames/SKILL.md | a clip from a voice track: an element on every word |
| infographic/SKILL.md | a sketch in one picture, a PDF, a branded HTML fallback |
| infographic/scripts/build_pdf.py | a PDF from a series of pages |
| mini-podcast/scripts/podcast.py | two-voice mp3 with a loudness check |
| flashcards/scripts/deck.py | deck check, build, PNG and PDF, Telegram quizzes |
The set is open: skills, hooks, tools and templates. Every pull request is read by a human and by Mila Quality.
Facts on this page: github.com/shakhruz/mila-agent (main, e26b9f3, v1.7.1), checked on 2 November 2026. The chat in the phone is an illustration. “Mila”, “MILAGPT” and the logo are trademarks of MILAGPT.