A workflow is the same five stages in the same order for every video: idea, script, voice, visuals, edit. AI can speed up each stage, and it works best when you give it fixed instructions and keep the decisions yourself. Make the voiceover before the visuals, so every image can be timed to what is said. Keep a human approval step between idea and script, and between script and upload.
This lesson connects the earlier modules into one process. The next lesson automates parts of it.
Plays from YouTube when you press play. Watch on YouTube
The stages: idea, script, voice, visuals, edit
| Stage | What AI does | What you decide |
|---|---|---|
| Idea | Collects candidates, proposes angles | Which idea is worth a video |
| Script | Drafts from your outline and instructions | Structure, facts, tone |
| Voice | Generates the read | Which voice, which take |
| Visuals | Writes prompts, generates images | Style, what is shown when |
| Edit | Captions, rough assembly | Pacing, what to cut |
The simplest version uses free tools: a chatbot writes the script from the idea, you collect one image or clip per line from stock sites, and a text-to-speech tool reads it. Money Groot deliberately does not name a voice. Pick or create your own, so the channel does not sound like every other one. Money Groot 3:00
Voice-driven image prompts by timestamp
E'Calm's order is voice first, pictures second:
- Generate the voiceover and transcribe it into subtitles with timestamps.
- Give the timestamps to the AI with a reference video for the visual style. He takes the reference from another niche, to stand out in his own.
- It returns one image prompt per timestamp. His team edits the prompts by hand before generating.
- Generate the images in batches, each file named after its timestamp.
In the editor the images go over the voiceover in timestamp order and are trimmed at each time. The timestamps are rounded, so he checks every cut by ear and moves it half a second where needed. E'Calm 21:35
A free-tools explainer pipeline
For the "every X explained" format the whole chain runs on free plans: a chatbot for the script (ask it to extend the script for a longer video), free stock sites for one visual per line, and a free text-to-speech tool for the voice. Money Groot 3:00
Free plans often limit commercial use. Check the terms of the voice and the stock footage before you monetize.
From a competitor's best title to a finished video
Romayroh takes the most-viewed title of a channel he uses as a model, adapts it to his own subject, and has Claude write the script from a prepared set of instructions. An AI editing tool turns the script into a narrated video with a thumbnail in about 40 minutes. He notes that the tool was not publicly available at the time, and that a freelancer plus a voice tool can do the same job. Romayroh 9:55
Two cautions. He shows only the opening of the result, so judge such tools on a full video of your own. And adapt the title to a subject that is really yours; a near copy of title and topic is what the policy lesson warns about.
An agent pipeline with a human approval step
Tim Danilov runs two AI agents:
- Agent one, daily: collects new videos from competitors in a spreadsheet, judges whether each could be adapted to his channel, and proposes five ideas per video.
- He ticks the ideas he likes.
- Agent two, the next morning: writes scripts for the ticked ideas from the channel's script guide, saves them as documents and adds a card to the team's task board.
His own work is approving ideas and scripts. The approval step is what keeps the channel from publishing whatever the machine proposes. Tim Danilov 23:50
Checklist
Your ticks are saved in this browser only.
Income and cost figures mentioned in these videos (for example "replaced a script writing team costing $3,000 to $4,000 a month") are the creators' own claims and are not verified.

