Video with AILesson 58 of 66

Contents

Video with AI

Lesson 58 of 66. Sign in to keep your place.

Module 6 · From generations to a finished piece

The whole pipeline on a phone or an old laptop

9 min read

The honest situation

DaVinci Resolve is free and it is heavy. It wants a reasonable GPU and 16GB of RAM, and on a machine with integrated graphics and 8GB it will install, open, and then stutter through playback in a way that makes editing unpleasant.

That is a real barrier and talking around it does not help anybody. So here is what actually works on what.

The tool ladder

Resolve free tier, if the machine can run it. A professional editor, colourist and audio suite in one, free permanently, no watermark. Real limits in the free version: some effects and the full noise reduction are withheld, and there are restrictions around certain high-resolution and codec outputs.

Kdenlive and Shotcut, open source, much lighter, happy on older Windows and Linux machines that Resolve refuses. For cuts, titles, audio and basic colour they are not worse. For colour work they are clearly behind: no shot match, weaker scopes, no tracked secondaries.

Olive and Avidemux for very constrained machines, where Avidemux is a trimmer rather than an editor.

CapCut on a phone. The whole assembly is achievable: multiple tracks, audio, titles, subtitles, filters, speed. For vertical short-form it is often faster than a desktop. Its free tier and watermarking on certain templates have moved around over time, so check before delivering something commercial.

FFmpeg on anything at all, including a phone through Termux. Not an editor, and it does conform, trim, concatenate, retime, denoise, scale, mix audio and export. Every operation in the preceding lessons except the creative choices has an FFmpeg equivalent.

Proxies, which change everything

The single most effective intervention on a slow machine. Edit with low-resolution copies, swap in the originals at export.

Bash
mkdir -p proxy
for f in footage/*.mp4; do
  ffmpeg -i "$f" -vf scale=640:-2 -c:v libx264 -preset veryfast -crf 28 \
    -c:a aac "proxy/$(basename "$f")"
done

Resolve and Kdenlive both support proxy workflows natively, where the editor manages the swap. Doing it manually works too: edit the proxies, then relink to the originals before export.

A 640-pixel proxy plays smoothly on almost anything, and every editing decision except colour and fine detail can be made on it.

Working in stages

If the machine cannot hold a whole project, do not make it. Split the work:

  1. Conform and trim every clip with FFmpeg first, so the editor receives clean, matched material.
  2. Edit picture only, with proxies, no effects.
  3. Export that cut, then do the grade as a second pass on the flattened export.
  4. Do sound as a third pass in Audacity against the exported picture, then mux the finished audio back in.
Bash
ffmpeg -i picture.mp4 -i final_mix.wav -c:v copy -c:a aac -b:a 320k -shortest out.mp4

This is how editing worked before machines could play everything at once, and it still works. It is slower and it removes the memory pressure entirely.

What actually needs a good machine

Honestly, three things:

  • Real-time playback of many layers. Solved by proxies and by flattening between stages.
  • Noise reduction and heavy effects. FFmpeg’s denoisers cover most of the need at the cost of a render wait.
  • Generating the video in the first place. This is the real constraint, and it is why hosted tools and free GPU sessions exist. Editing is not the expensive part.

Notice that none of them is the editing itself. A sequence of cuts, a grade, titles and a sound mix are not computationally demanding tasks; they became demanding because editors render everything live.

The phone-only workflow, stated as a real option

Generate through a browser on a hosted free tier or a Hugging Face Space. Repair stills in Photopea in a mobile browser or Snapseed. Assemble in CapCut. Record voice and ambience on the phone’s own microphone. Clean audio in a mobile audio editor, or accept the phone recording as it is. Add subtitles with the editor’s own transcription. Export and publish.

That is a complete pipeline. It is not a lesser version of the work with a caveat attached: short-form vertical video made this way is the dominant format on the internet, and the constraints of the phone match the constraints of the format.


Today: make proxies for your current project with the loop above and edit the proxies. Time how much less you wait.

Every lesson here is open. Your progress is recorded once you pass the Editing Video exam.

Check your understanding

An editor on an 8GB laptop finds playback unusable in a free professional editor. What single change most improves the situation without changing tools?

Pick the one you would defend. Your answer stays private.

Next lesson · 9 minAn afternoon of curiosity can cost more than a day of work

Stuck on something?Ask Chiamaka, your AI tutor

How to cite: Addaly (2026) ‘The whole pipeline on a phone or an old laptop’, in Video with AI. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-video/editing-without-a-good-machine (Accessed: 13 October 2026).