Talk
AI Video Editing
By method rather than by model: when to let AI generate a video, when to let it transform your own footage, and when to have it write the video in code.
Generate what may be imagined. Transform what may change. Code what must be exact.
Tools you can use
Where to go, and what each one is for. Checked on each maker's own page, 25 September 2026. Plans and prices change often.
01Transcribe, with a time on every word
- Gemini 3.5 Transcribe Transcript, speakers and a time on every word. Egyptian Arabic is on its list. Free tier in Google AI Studio (preview)
- ElevenLabs Scribe Upload audio or video, get a time on every word, then export. Free: about 30 minutes a month
- Whisper, in the Buzz app Runs on your own computer, nothing uploaded. Switch on Word-Level Timings. Free and open source
- Gemini 3.1 Flash-Lite What I used: close to word for word on my Egyptian lecture. No word times, only a time per sentence if you ask. Free tier in Google AI Studio
Ask for word times, then read the Arabic yourself. Mixed Arabic and English trips every one of these tools.
02Generate or transform video
- Gemini app › Create video Type a sentence, or add your clip and say what to change. Runs on Gemini Omni. Needs Google AI Plus or Pro: EGP 249.99 or 939.99 a month in Egypt
- Google Flow Scenes and edits with Omni or Veo 3.1. Upload your clip and brief the change. 50 free credits a day: about three 10-second clips, or one edit
- Google Vids Create AI videos with Omni 1.1, in 1080p. Free for any Google account since 23 September 2026
- Google AI Studio The same models from code, for developers (the Gemini API). Paid only: Omni about $0.10 a second of video
Start free in Flow or Vids. A clip of your own must be 10 seconds or less.
03Write the video in code
- Remotion Agent Skills The official skills. They work with Claude Code, Codex, Kimi Code or Cursor. Free
- Remotion Video written as code, rendered on your own computer. Needs Node.js. Free for individuals and companies of up to 3 people; larger companies from $25 a month per seat
- Remotion Studio Opens with npx remotion studio: play the video, drag things, change the timing. Free, part of Remotion
- Claude Code The coding agent that writes and edits the code for you. In Claude Pro, $20 a month; not in the free plan
In a terminal, inside your video project
npx skills add remotion-dev/skillsOne line teaches your AI Remotion. Then you check every frame in Studio.
04Give it a voice
- ElevenLabs Text to speech in 70+ languages, Arabic included. Clones your voice from 1 to 3 minutes of audio. Free: 10,000 credits a month. Cloning from $6 a month
- Fish Audio Text to speech in 80+ languages, Arabic included, with tone tags like [whisper]. Clones a voice from 10 to 30 seconds of audio. Free plan. Plus $11 a month
Clone only your own voice, and listen to every Arabic line before you publish.
05One account, many models
One bill, not a studio: most of it is used from code, and ElevenLabs and Gemini Omni are not on it.
06Cut a long talk
Let it cut the silences. Check the Arabic on screen yourself.
Share this lecture
About this lecture
AI Video Editing is a talk by Dr. Mohamed Maslooh, given on 27 September 2026 for doctors with no coding background, that sorts AI video tools by method rather than by model name. It covers three methods in depth: generating a video from a prompt, transforming your own footage, and having an AI agent write the video in code with Remotion. Using his own filmed lecture as the test, it shows where generated video invents details, what each method costs, and how to choose. It closes with three lines: generate what may be imagined, transform what may change, code what must be exact. Slides in English, delivered in Egyptian Arabic.
- AI touches video editing in five ways: generate (the model draws every pixel), transform (it redraws your footage), assist (it helps you edit), operate (it edits for you) and code (it writes the video).
- Generated and transformed video invents details: in a Google Flow test, the numbers on the speaker's own slide came out scrambled, and the same character requested twice came out different each time.
- With code, an AI agent transcribes the speech, gives every word a time, cuts at the silences and adds graphics, and the same code renders the same video every time, with faces and numbers exactly as given.
- Costs differ sharply: generating through Google's API costs about $6 a minute for every try, while the coded video in the talk cost about $1 in total on a $20 Claude plan, by the speaker's own estimate.
- The deciding question before picking a tool is whether the frame must be true: if it must, code it or use real footage; if it need not, generate it or transform your footage.
Questions
Can AI edit my videos for me?
Yes, in several ways. It can generate new clips, redraw your own footage, or, through an AI agent writing code with Remotion, transcribe your speech, time every word, cut the silences and add captions and graphics. In the talk, 27.4 seconds of raw lecture footage became a 16 second clip this way.
What is Remotion and how do I start using it with AI?
Remotion is a tool for writing video as code and rendering it on your own computer, and it needs Node.js. To start, open an AI agent such as Claude Code or Codex, install the official Remotion skill with the command npx skills add remotion-dev/skills, give it your clip or start from scratch, and design the look before you render.
Why does AI-generated video get text and numbers wrong?
A video model does not find a picture, it draws one from noise, redrawing every pixel. That is why exact words and numbers come out garbled, the same character changes between tries, and sizes and positions drift. When a frame must be exact, such as a slide, a chart or a face, the talk recommends code or real footage instead.
Who gave the AI Video Editing talk, and where are the slides and tools?
Dr. Mohamed Maslooh, a medical intern at Kasr Al-Ainy, gave it on 27 September 2026. The slides open free in any browser at maslooh.com/lectures/ai-video-editing, and the same page lists every tool used, with links, prices and the Remotion install line.
Is it cheaper to generate AI video or to make it with code?
Code was far cheaper in the talk's comparison. Generating costs about $6 a minute through Google's API, roughly $3 a try on a $20 Google plan, and every new try costs again. The whole coded video cost about $1 on a $20 Claude plan, by the speaker's estimate, and re-renders are free on your own computer.