🔍 Read the full analysis: From AI Concept To Production: The Lessons Behind My First 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 video style went from concept to a live one-minute test video in one evening, but the first full video the next morning was slow and visually sparse. He reports rebuilding the video engine around measured scene checks and adding gates to catch problems before rendering. The account is a first-person report, and its results come from Meyer’s own build notes and experiments.
Thorsten Meyer says he turned an idea for a hand-drawn explainer video into a working video engine in one evening, then rebuilt it after the first full-length result proved slow and visually sparse. In an account published November 6, 2026, on ThorstenMeyerAI.com, he describes adding human approval, independent AI code review and measurable production gates to help catch the same problems before later videos are rendered.
Meyer writes that the initial idea came from watching a video whose whiteboard style he liked. To avoid copying it, he says he recorded a limited set of general traits—a hand-drawn stroke, a continuous visual world and large numbers used as color accents—and did not reuse its drawings, assets or layouts. He says Claude built a new engine from scratch, using his own design system and an openly licensed handwriting font.
He also says OpenAI’s Codex reviewed the code but did not build the engine. Codex identified five issues in its first review round, according to Meyer, and each fix was accompanied by a test designed to fail against the old code. Meyer describes this as a way to check that a test can detect the original problem, rather than merely pass after a change.
A roughly one-minute test video went live that evening after Meyer watched and approved it. The first full video, produced the following morning, drew a different response: he found it boring, with too much white space, icons that did not support the narration and completed drawings that remained still. Meyer says he left that video online unchanged as an example of what the process should avoid.
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 Visual Critique to Build Checks
The account’s central point is that generating a video is not the same as producing one that is ready to publish. Meyer describes a workflow in which a person reviews the output, measurements turn subjective criticism into specific problems, and automated gates stop some of those problems from reaching a rendered scene.
For the first three scenes of the rebuilt video, Meyer reports that the share of the frame filled with content rose from 15% to 66%, while time spent completely still fell from 29 seconds to zero. Camera moves increased from three to 27. These are figures from his own project notes, not results from an independent evaluation, and the article does not establish how well the checks perform across other videos.
The approach is relevant to creators and teams using generative tools because it separates several jobs that are often treated as one: making a draft, reviewing the implementation, judging the finished work and checking it against defined requirements. Meyer’s account argues for treating each as a distinct production step rather than assuming a successful demo is ready for release.
Products Worth Considering
Samsung Galaxy Tab A11+ 11-Inch Tablet, 6GB RAM, 128GB Storage, Expandable, Gray
Samsung Galaxy Tab A7 Lite 8.7-Inch 32GB Wi-Fi Tablet, Gray
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.
How the First Video Changed
Meyer says the improved engine was rebuilt after he identified three problems in the first full video: empty space, weak links between icons and narration, and static images after drawing completed. The revised rules required each shot to fill more of the frame, the camera to move with each sentence, finished objects to retain slight movement and each object to illustrate something being said at that moment.
He says the measurements became build gates: a check can stop a scene judged too empty, while the build can flag an object that does not illustrate the narration. The first video using the improved look went live a few hours after his critique, in an article about what he described as the Nasdaq’s worst day in 19 months. The accompanying explainer is about four and a half minutes long and uses an AI voice clone of Meyer, according to the source.
Meyer places the video work alongside a separate experiment involving a local model and legal questions. He reports that after a model’s weights were edited, it completed 12 of 15 legal-benchmark answers, compared with four beforehand, while its legal score remained zero. The account supplies no further benchmark detail in the provided material, so the result does not establish improved legal performance.
““A demo that runs is not a decision.””
— Thorsten Meyer
Limits of the Reported Results
The metrics and workflow details are Meyer’s own account, based on his build notes, incident notes and experiments. The source does not provide independent verification, a detailed evaluation method for the video measurements, or results across a larger set of scenes and projects. It is also unclear how the automated checks determine whether an object genuinely supports the narration, or how often they flag acceptable scenes.
The article’s legal-model example is separate from the video system. Although Meyer reports more completed answers after a weight edit, he says the legal score remained at zero. The supplied material does not specify the benchmark’s scoring method, model details or sample conditions, so the answer-count change should not be read as evidence of better legal accuracy.
Checks for Future Video Builds
Meyer says the new scene measurements now act as production gates, and that the video linked with his account uses the rebuilt approach. The next practical test is whether those checks keep identifying sparse, static or poorly matched scenes as he makes more videos. The article does not give a timetable for further releases or describe a planned independent assessment.
For now, the account documents a workflow and its early results rather than a broadly tested production system. Further videos, published evaluation methods and information about how often the gates catch issues would help show whether the process holds up beyond the examples Meyer describes.
Key Questions
What did Thorsten Meyer publish?
He published a first-person account of building an AI-assisted engine for hand-drawn explainer videos, identifying problems in its first full video and adding measurements and build checks.
How did the first full video fall short?
Meyer said it had too much white space, icons that did not support the narration and drawings that became still when finished. He described it as slow and boring.
What changed in the rebuilt video?
For the first three scenes, Meyer reported that frame fill rose from 15% to 66%, still time dropped from 29 seconds to zero, and camera moves rose from three to 27. He says those measures were turned into checks used before rendering.
Was the video workflow independently tested?
The provided account reports Meyer’s own project measurements and does not describe an independent evaluation or results across a larger set of videos.
What did the legal-model experiment show?
Meyer reported that a model completed 12 of 15 legal-benchmark answers after a weight edit, compared with four before, while its legal score remained zero. The supplied account does not provide enough benchmark detail to assess what the change means for legal accuracy.
Source: ThorstenMeyerAI.com
