TL;DR
Get everyday essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Thorsten Meyer says he built a hand-drawn video engine in one evening, then revised it after the first full video felt slow and visually sparse. Measurements across the first three scenes showed frame fill rising from 15% to 66%, still time falling from 29 seconds to zero, and camera moves increasing from three to 27; those measures now inform production gates.
Thorsten Meyer says he built a new hand-drawn video engine in one evening, reviewed its first full-length result the next morning, and revised it after identifying pacing and visual issues. His account describes a production process involving AI coding and review, human approval, and performance checks before publication.
Meyer writes that the idea followed seeing a video whose camera moved across a hand-drawn whiteboard scene. He says he recorded only broad style traits for his own project and used no drawings, assets or layouts from the original. Claude built the engine, according to the post, while OpenAI’s Codex reviewed the code. Meyer says five findings from the first review were fixed that evening, with a test for each that failed against the earlier code.
A roughly one-minute test video received Meyer’s approval and was published that evening in an article about local votes against data centers, he says. The first full video in the style appeared the next morning. Meyer reports identifying white space, icons he considered unrelated to the narration and scenes that remained still after a drawing was complete. He says he measured the first three scenes before making changes.
In those scenes, the proportion of the frame occupied by content rose from 15% to 66%. Time spent with the picture standing still fell from 29 seconds to zero, while camera moves increased from three to 27. Meyer says the resulting checks now act as gates: a build stops if a scene is too empty or an icon illustrates nothing being said. The account does not provide the full video’s runtime, a broader sample of scenes or independent evaluation of the measurements.
From AI idea to production One evening, then a rethink
Thorsten Meyer built a hand-drawn video engine, then revised it after the first full video felt slow and visually sparse. His account describes checks that help decide whether a scene is ready to proceed.
From idea to first video
The account describes a compressed build, review and publication process.
Claude built it using Meyer’s own design system and broad style traits.
Codex reviewed the code; Meyer says five findings were fixed with tests.
Meyer approved a roughly one-minute test video before it was published.
What changed after review
Meyer measured the first three scenes and used the findings to guide revisions.
- Too much empty space
- Icons unrelated to the narration
- Scenes stayed still after drawing finished
- More of the frame filled with content
- Still time reduced to zero
- Camera moves increased
Checks that can stop a build
The measurements inform two production gates before publication.
- Stop if a scene is too empty.
- Stop if an icon does not illustrate the narration.
- Keep human review and approval in the process.
How Visual Checks Shape Publication
The account describes how Meyer revised an AI-assisted video prototype after reviewing its first full-length version. He reports identifying pacing and visual issues and making changes before publication. The episode distinguishes technical completion from editorial and visual review in his production process.
The reported measurements turn some of Meyer’s observations into checks that can be applied to later scenes. He says the checks are intended to flag sparse frames or illustrations that do not correspond to the narration before publication. The post does not establish whether those checks improve communication or viewer response; Meyer describes them as part of a process that also includes a person watching and approving the result.
Meyer also describes separating code generation from code review. He says one AI system builds and another reviews, with findings tied to tests, while he retains the publication decision. The account is a first-person description and does not independently establish how reliably the approach works across projects.
Products Worth Considering
Samsung Galaxy Tab A11 8.7-Inch Wi-Fi Tablet, 64GB, 4GB RAM, Silver
Apple MacBook Air 13-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, 12MP Center Stage Camera, Touch ID, Wi-Fi 7, Midnight
LG UltraGear 34GX90SA-W 34-inch WQHD Curved OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
From Style Reference to Original Engine
The source is a personal account published by Thorsten Meyer AI on November 6, 2026, and marked as created with AI assistance. Meyer frames the project as a response to a video style he liked: a hand-drawn explainer presented as a continuous world, rather than a sequence of slides. He says he translated that impression into a list of style traits and built a separate engine, using his own design system and an openly licensed handwriting font.
The sequence described is compressed into a single evening and the following morning: an initial engine and review, a short approved test video, a full video, then a revision based on measurements and publication. The post says the one-minute test appeared in an article about local votes against data centers. Its focus is Meyer’s own build notes and experiments, rather than a formal study or an independently audited production report.
“It finishes is not it’s right.”
— Thorsten Meyer
What the Measurements Do Not Show
The figures cover only the first three scenes, and the post does not specify the complete video’s length or provide measurements for the full production. It also does not report viewer testing, engagement data or an independent assessment of whether the revised video communicates better. The reported changes are therefore evidence of alterations in the selected visual measures, not proof of audience impact.
The account does not detail the thresholds used for the two gates, how they are calculated, or whether every later video has passed them. Meyer’s descriptions of the video as boring and the revision as much better are his own judgments. The post does not include the full review exchange or enough information to assess the five code findings independently.
Checks for Future Videos
Meyer says the two visual checks are now part of the build process and can stop a video from proceeding when a frame is too empty or an icon does not illustrate the narration. The post does not announce a further release date or provide a schedule for new videos. The next development described is continued use of those gates in later production.
For readers following the project, a fuller account of subsequent videos, their scene-level measurements and any audience feedback would show whether the workflow performs consistently beyond this first revision. Meyer’s post does not say when such results will be available.
Key Questions
What did Thorsten Meyer build?
He says he built an engine for hand-drawn explainer videos with a moving camera and a continuous illustrated world. The account says Claude built the engine and that Meyer used his own design system.
What changed after the first full video?
Meyer says he reduced empty space, added movement and revised how icons supported the narration. Across the first three scenes, frame fill rose from 15% to 66%, still time fell from 29 seconds to zero, and camera moves increased from three to 27.
Who reviewed and approved the work?
Meyer says OpenAI’s Codex reviewed the code after Claude built it. Meyer watched a roughly one-minute test video and approved it for publication; the post presents that human approval as a separate step.
Do the reported figures prove the revised video worked better?
No. They describe visual changes in the first three scenes, based on Meyer’s build notes. The post does not provide independent review, audience testing or engagement results.
Source: Thorsten Meyer AI
