🔍 Read the full analysis: 5 Hard-Learned Rules For Turning An AI Idea Into A Production Video on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
Thorsten Meyer says a hand-drawn AI video style moved from an evening experiment to a live production format, but the first full video was slow and visually weak. His account emphasizes independent code review, human approval, measurable quality checks and monitoring that reaches a person; it is a personal case study, not independent evidence that the approach works in every production setting.
Thorsten Meyer says he turned an idea for a hand-drawn AI video format into a live production workflow within a day, but the first full video made with it was, in his words, slow and boring. In a November 6 account on ThorstenMeyerAI.com, he describes five practices behind the revised version: defining original visual rules, separating code creation from review, requiring human approval, measuring quality and turning those measures into checks that can stop a build.
Meyer said the initial idea came from seeing another creator’s whiteboard-style explainer. To avoid copying it, he wrote down a limited set of general traits, including a hand-drawn line and a continuous visual world, and says he used no drawings, assets, layouts or code from the original. Claude built the engine, according to Meyer, while OpenAI’s Codex reviewed the code without contributing to its construction. Meyer reported that Codex identified five issues in the first review; he says each was fixed and paired with a test that failed against the earlier code.
A roughly one-minute test video received Meyer’s approval and went live that evening within an article about local votes against data centers. The next morning, however, the first full video in the style did not meet his expectations. Meyer identified excessive white space, icons that did not support the narration and images that stopped moving after they were drawn. He says the original video remains online as an example of what the team wanted to improve.
Instead of relying only on subjective tweaks, Meyer says he measured the share of the frame occupied by content and the duration of still images. Across the first three scenes of the rebuilt video, he reports that frame coverage rose from 15% to 66%, still time fell from 29 seconds to zero, and camera moves increased from 3 to 27. He says these measures became production gates: checks can stop scenes that are too empty, while a build can flag an object that does not illustrate the narration. The improved-look video was published a few hours after his critique.
Infographic 01 · The new video look
One evening, one boring first video, one measured fix
Speed was not the point. The steps in between were: a second AI that reviews, a human who says yes, and numbers that became gates.
Share of the frame filled with content on the first three scenes after the fix. Before: 15%.
From an evening idea to a measured fix
Relative times, this week
- EveningA video I likedClean room: only a list of style traits, no drawings, assets or layouts
- Same eveningOwn engine, first reviewClaude builds; OpenAI’s Codex reviews; 5 findings fixed, each with a test that fails on the old code
- Same eveningA human says yesOne-minute test video approved and published
- Next morningFirst full video: slow and boringToo much white space, icons that did not fit, a picture that stood still
- Hours laterMeasured fix, then gatesThe first video in the improved look goes live
Source: my own build notes for the new video look, this week (engine history, review round, approval, critique, before/after measurement of the first three scenes).
Powered by Thorsten Meyer AI
From Video Metrics to Production Gates
The account’s practical point is that getting a generative system to produce an output does not establish that the output is useful, accurate or ready to publish. Meyer’s process adds checks at several stages: a separate code reviewer, tests designed to detect specific defects, a person responsible for approving the result, and visual criteria that can block later renders.
That distinction matters for teams using AI in published media. A prototype can work while still producing dull or poorly paced material, and a passing test suite can miss problems the tests do not cover. Meyer’s numbers show how one creator translated a subjective criticism into trackable checks in this project. They do not establish that the same thresholds will suit other formats or audiences.
Popular right now
De’Longhi Classic Matte Black Espresso Machine with Milk Frother
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
Ninja CREAMi Scoop & Swirl 13-in-1 Ice Cream and Soft Serve Maker (NC701)
As an affiliate, we earn on qualifying purchases.
How the First Cut Fell Short
Meyer presents the video work as a personal production account, with figures drawn from his own build notes, incident notes and experiments published on his site. The described workflow assigns different roles to two AI systems: Claude builds the engine, while OpenAI’s Codex reviews it. A human decision remains part of the release process rather than being replaced by either system.
The source also describes a separate experiment involving edits to a model’s weights for legal-question answering. Meyer says the model completed 12 of 15 answers after the edit, compared with 4 of 15 before, but its legal score remained zero. The result is a caution against treating task completion as proof of quality; it is a separate experiment, not a test of the video workflow.
“A demo that runs is not a decision.”
— Thorsten Meyer
What the Case Study Cannot Establish
The account does not provide independent evaluation of the videos, audience response, viewing data or a comparison with other production methods. The reported measurements apply to the first three scenes of one rebuilt video, and Meyer does not specify the thresholds used for every production gate. It is also unclear how the checks perform across different subjects, video lengths or visual styles.
The five rules are Meyer’s takeaways from his own work, not results from a controlled study. His report also mentions an earlier monitoring failure in August, when a guard generated about 1,300 warnings in a log that nobody read and an alert email did not go out. The account does not give further details on how that monitoring issue was corrected.
How the Workflow Will Be Tested
Meyer says the improved video look and its checks are now being used in subsequent videos, including the explainer linked with his account. The next evidence readers can assess is whether the gates continue to catch empty scenes and mismatched visuals in later productions, and whether human review changes or rejects outputs that pass the automated checks.
No independent assessment or further release schedule is provided in the source material. For now, the production results and the proposed rules remain a documented account of one creator’s workflow, rather than a verified template for AI video production more broadly.
Key Questions
What happened to the AI video project?
Thorsten Meyer says he developed a hand-drawn video engine from an initial idea and published a short test video. After the first full video fell short, he measured visual issues, revised the engine and added checks for later builds.
What are the five rules Meyer describes?
His account emphasizes defining original visual traits rather than copying a reference, using a separate AI for code review, requiring human approval, measuring quality instead of relying only on taste, and turning those measures into checks that can flag or stop production.
How did the rebuilt video differ on the reported measures?
For the first three scenes, Meyer reports that frame coverage rose from 15% to 66%, still time dropped from 29 seconds to zero, and camera moves increased from 3 to 27. These are his project measurements, not independently verified results.
Does the account prove these rules work for every AI video project?
No. It is a personal case study of one workflow. The source does not provide an independent evaluation, audience metrics or evidence that its measurements and thresholds apply to other teams or formats.
Source: ThorstenMeyerAI.com
