🔍 Read the full analysis: The Hard Lessons Behind My First AI Production Video on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Thorsten Meyer says a hand-drawn AI video style moved from idea to a live test in one evening, but the first full video was slow and visually dull. He describes how code review, human approval and measurable scene checks helped improve the production process, while noting that passing tests alone is not proof of quality.
Thorsten Meyer says he developed a hand-drawn video engine in one evening and published a short test that same night, but the first full video produced the next morning was slow and visually dull. In a November 6 account on ThorstenMeyerAI.com, Meyer describes how he used AI code review, human approval and scene-level measurements to revise the workflow before publishing a later video.
Meyer says the idea came from a hand-drawn explainer video whose visual style he admired. To avoid copying it, he says he recorded only broad traits—such as hand-drawn strokes and a continuous illustrated world—and used no drawings, assets or layouts from the reference. Claude built the engine, according to Meyer, using his own design-system colours and an openly licensed handwriting font.
Meyer says OpenAI’s Codex reviewed the code but did not build the system. It identified five issues in the first review round; he says each was fixed with a test that failed on the old code. He also describes watching a roughly one-minute test video himself and approving it before it was published that evening within an article about local votes against data centers.
The next morning’s full video did not meet his expectations. Meyer said it had too much empty space, icons that did not support the narration, and finished drawings that remained still. After measuring the first three scenes, he reports that filled-frame coverage rose from 15% to 66%, still-image time fell from 29 seconds to zero, and camera moves increased from three to 27 in the rebuilt version. He says those measures became checks that can stop a scene from rendering if it is too empty.
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
Turning Visual Critique Into Checks
Meyer’s account illustrates a production problem that can be obscured by demonstrations: generating a video is not the same as deciding that it is ready to publish. The first output ran, but he found it unengaging. His response was to translate subjective criticism—too much whitespace and too little movement—into measurable criteria that could be checked again on later scenes.
The distinction matters for teams using AI in creative production. A model can help build an engine or produce an output, while review, editorial judgment and release approval remain separate responsibilities. Meyer’s workflow assigns different roles to Claude, Codex and a human reviewer. The described checks address visual composition and movement; they do not, by themselves, establish that a video is accurate, fair or otherwise ready for publication.
His figures are specific to his own build notes and the first three scenes, rather than an independent evaluation of AI video production. Still, the account offers a practical example of how a team can turn feedback into repeatable checks instead of relying only on repeated subjective adjustments.
Products Worth Considering
Samsung Galaxy Tab A11+ 11-Inch Tablet, 6GB RAM, 128GB Storage, Expandable, Gray
Samsung Galaxy Tab A11 Wi-Fi Tablet, 8.7-Inch, 64GB, 4GB RAM, Gray (SM-X133, International Version)
Bose QuietComfort Ultra Earbuds (2nd Gen) – Wireless Noise‑Cancelling Earbuds with Mic, Immersive Audio, USB‑C Charging, Up to 6 Hours Batte
As an affiliate, we earn on qualifying purchases.
From One-Minute Test to Full Video
The sequence began with a short test made on the evening Meyer developed the engine. He says he watched the roughly one-minute video and decided it could go live. That test appeared in an article about local votes against data centers. The following morning, the first full-length video using the look exposed problems that were less apparent in the brief test.
Meyer kept that first full video online unchanged as an example of what the system had done poorly. He then rebuilt the scene approach around frame coverage, movement and whether each object illustrated something being said at that moment. A later video about the Nasdaq’s worst day in 19 months was published in the improved look, according to his account. The video accompanying his article is described as a roughly four-and-a-half-minute hand-drawn explainer with an AI voice clone of Meyer.
The account also links the video work to a separate experiment that week: a model edited another model’s weights in an attempt to improve legal-answer performance. Meyer says it completed 12 of 15 answers in a legal benchmark, compared with four before, but its legal score remained zero. He presents that result as a parallel lesson: more completed answers did not mean the model had improved on the measure that mattered.
“The idea is the easy part.”
— Thorsten Meyer, writing on ThorstenMeyerAI.com
Limits of the Reported Results
The figures come from Meyer’s own build notes and measurements; the account does not provide an independent review, a detailed measurement method or results across a larger set of videos. The reported frame coverage, still time and camera-move counts apply to the first three scenes of the rebuilt video, not necessarily the entire production.
It is also unclear how consistently the automated checks identify whether an object genuinely supports narration, or how often a person must override them. Meyer’s report describes a working process and its reported outcomes, but does not establish that the same setup will produce comparable results for other creators or content types.
More Tests of the Workflow
Meyer says the visual measurements now act as gates in the production process: scenes that are too empty can be stopped before rendering, and objects that do not match the spoken content prompt a build warning. The account does not set out a future release schedule or a formal evaluation plan.
The next evidence readers would need is how these checks perform across additional videos, including whether they reduce the problems Meyer identified without introducing new ones. The separate legal-model experiment also remains an example of why output volume and task performance must be evaluated independently: according to Meyer, the model completed more benchmark answers, while its legal score stayed at zero.
Key Questions
What did Thorsten Meyer publish?
He published a first-person account on November 6, 2026, describing how he developed and revised a hand-drawn AI video workflow.
What was wrong with the first full video?
Meyer said it had too much white space, icons that did not support the narration, and drawings that stopped moving once complete.
How did he measure the revised video?
For the first three scenes, Meyer reported frame coverage rising from 15% to 66%, still time falling from 29 seconds to zero, and camera moves increasing from three to 27.
Who reviewed the video engine’s code?
Meyer says Claude built the engine and OpenAI’s Codex reviewed the code. He also watched and approved the short test video before publication.
Do the reported numbers prove the workflow works for other creators?
No. The numbers are from Meyer’s own notes and a limited set of scenes; the account does not include independent testing or broader results.
Source: ThorstenMeyerAI.com
