🔍 Read the full analysis: Turning An AI Idea Into A Video: My One-Evening Experiment on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Thorsten Meyer says he built a new hand-drawn video engine in one evening and published a short test video after reviewing it himself. The first full video the next morning felt slow and empty, so he measured scene composition and motion, revised the engine and turned those measures into build checks. The account is a personal experiment, not an independent evaluation of the tool or workflow.
Thorsten Meyer says he built a hand-drawn video engine in one evening, then revised it the next day after its first full video struck him as slow and boring. In an account published November 6 on ThorstenMeyerAI.com, Meyer describes using one AI tool to build the engine, another to review its code, and measured checks to guide changes before later videos went live.
Meyer writes that the idea came from seeing a whiteboard-style explainer he liked. To avoid copying it, he says he listed only broad visual traits: hand-drawn strokes, a continuous world rather than slides, and large colored numbers. He says the new engine used no drawings, assets, layouts or drawing code from the video that inspired him. Its colors came from his own design system, and its handwriting font was openly licensed.
According to Meyer, Claude built the engine, while OpenAI’s Codex reviewed the code but did not build it. He says Codex identified five findings in its first review, all fixed that evening, and that each fix came with a test that failed against the old code. Later that evening, Meyer watched a roughly one-minute test video and approved it for publication. He says it appeared in an article about local votes against data centers.
The first full video in the new style, made the following morning, did not meet his expectations. Meyer says it had too much white space, icons that did not support the narration, and images that stopped moving once drawn. He kept that video online unchanged as an example of the shortcomings. For the revised version, he measured frame coverage and still time across the first three scenes: filled frame rose from 15% to 66%, still time fell from 29 seconds to zero, and camera moves rose from three to 27. He says these measures became checks that can stop a scene from rendering if it is too empty, while another check flags objects that do not illustrate the narration.
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 Build Checks
The account’s main point is not simply that a video was produced quickly. Meyer describes a workflow in which human approval and automated review sit between an AI-generated tool and publication. The distinction matters for readers considering similar systems: a demonstration that runs does not by itself establish that its output is useful, accurate or ready to publish.
His example also shows how subjective criticism can be turned into operational tests. Meyer’s response to the first full video—too much empty space and too little movement—led to measurements and then to gates in the production process. Those gates could catch specific visual problems before rendering, though the measures do not establish whether a video is clear, factually sound or engaging to viewers. The account does not report audience testing or independent verification of the results.
Products Worth Considering
Dell XPS 13 9345 Laptop, 13.4-Inch FHD+ 120Hz, Snapdragon X Elite, 32GB RAM, 1TB SSD
Apple MacBook Pro MPHF3LL/A 14.2-Inch Laptop | M2 Pro 12-Core Chip 16GB Unified Memory 1TB SSD | 19-Core GPU Liquid Retina XDR Display Space
Dell UltraSharp 27 Thunderbolt Hub Monitor (U2726DER), QHD
As an affiliate, we earn on qualifying purchases.
The First Video Setbacks
Meyer presents the video experiment as part of a wider approach to putting AI-generated work into production. His article argues that green tests are not proof: in a separate review round he describes, all 12 tests written by an author passed, while four of six checks from a reviewer failed. He also recalls an August monitoring incident in which a guard produced about 1,300 warnings in a log that nobody read, while an alert email did not go out.
Another experiment in the same week illustrates a different kind of limit. Meyer says a model edited for a legal-answer task completed 12 of 15 answers, compared with four before, but its legal score remained at zero. These are examples from his own build notes and experiments, not independently reported benchmark results. Together, they frame his video account as a discussion of validation and oversight as much as a production diary.
“And honestly, it was slow and boring.”
— Thorsten Meyer, in his November 6 account
What the Metrics Cannot Show
The figures in the account are Meyer’s own measurements; the article does not provide an independent audit, audience study or full technical description of how frame coverage and still time were calculated. It is also unclear how well the gates perform across different topics, video lengths or visual styles, or how often they block acceptable scenes.
Meyer says the improved video about the Nasdaq’s worst day in 19 months went live a few hours after his critique, but the account does not offer comparative viewer data or establish that the revisions improved comprehension. The source also does not specify the exact publication time of that video or provide evidence that the workflow’s checks can catch factual errors in narration.
How the Workflow Will Be Tested
Meyer’s article does not announce a formal next milestone, release date or plan for outside testing. For now, he says the scene-level thresholds and checks are part of the engine and that the earlier, less successful video remains online as a comparison. Further evidence would require results from additional videos and information about whether the checks consistently identify problems without rejecting sound work.
Readers can watch the roughly four-and-a-half-minute explainer linked in Meyer’s article, which covers the initial idea, the dull first full video, the measured revision and the production rules he describes. Meyer says its narration uses an AI voice clone of himself. Whether the workflow can be reproduced by others, and how much human review it needs as production scales, remains unanswered.
Key Questions
What did Meyer build?
He says he built a hand-drawn video engine in one evening, inspired by broad traits of a whiteboard-style explainer. He says the engine used original drawings and layouts rather than assets from the video that inspired it.
Which AI tools did he use?
Meyer says Claude built the engine and OpenAI’s Codex reviewed its code. In his described setup, Codex did not build the tool; it checked for problems and reported findings.
What changed after the first full video?
Meyer says he measured frame coverage, still time and camera movement after finding the first full video too empty and static. Across the first three scenes of the rebuilt video, he reports coverage rising from 15% to 66%, still time falling from 29 seconds to zero, and camera moves increasing from three to 27.
Are the reported improvements independently verified?
No independent verification is described in the source. The figures are Meyer’s own measurements, and the account does not report audience testing or a third-party review of the results.
Source: ThorstenMeyerAI.com
