Scooby-Doo Character Studio
A character system that had to look like 1970s Scooby-Doo, even after the fan mixed every trait in the library.
Fans open a web app, talk to Scooby and Shaggy, build a Mystery Inc.–style avatar, and can buy a shirt, a poster, or a custom book with that character in What a Night for a Knight.
Duncan Studios supplied a lot of the source art. The website had to drive the combinations. In the middle of that, the Blender side had to assemble, rig, pose, preview, and render characters that still read as Scooby-Doo no matter what the fan picked.
That was my job.
What I did
The problem was not making one Scooby-Doo character. The problem was making thousands of them look like they belonged in the same book.
I defined the technical requirements for the traits: profile heads and hair, expressions, hand gestures, shoe profiles, eye direction, alignment, consistent scale. Then I sat with Duncan Studios and walked through why the vector line and fill work had to be built a certain way, and why scale mistakes would break every pose downstream.
I built an Illustrator automation pass to clean and organize traits, then the Blender Character Builder used to assemble, rig, pose, preview, and render configurable bodies. Around that I kept expanding GrooveTech_Character_Builder_Toolkit — pose and rest controls, pose rendering, trait indexes, line-thickness per trait and per pose, pose libraries, hair + cap + pose corrections, profile/front filters. The toolkit went well past v90 because production kept finding new ways to break.
I designed the JSON contract for trait names, indexes, and colors so Blender and the website were talking about the same character. When the site sent null colors, bad glasses colors, skirt issues, or trait order problems, I diagnosed our side and theirs.
I also ran the posing/QA operation. I created the channel, the training, the Miro boards, the pose maps, the tracking sheets, and the buddy/QA process. I assigned work, reviewed poses, and personally updated the ones that had to match franchise taste: expressions, line thickness, hair and cap collisions, the expanded pose set.
On the output side I built the rendering path for books, shirts, posters, social graphics, and the back-cover text alignment. Local and server-facing QA for books and consumer products sat on top of that, because a bad combination is worse if it ships printed.
Constraints that mattered
The art had to hold the 1970s line look after it was assembled in 3D. Thin lines, white gaps, and overlapping traces all showed up in merch.
Hair, caps, and expressions could not be treated as independent stickers. Mixing them produced clipping and broken silhouettes unless the builder knew about the combination.
The website and Blender had to share exact indexes. Close enough was a broken character.
What I do not claim
Duncan Studios made a lot of the source artwork. Several artists rigged and posed assigned bodies inside the system I built. Website and backend implementation was shared with development. I owned the Blender character system, the toolkit, the data contract, the QA, and the team that posed the library.
Awards
The production received an Honorable Mention at the 2025 Hermes Creative Awards (Electronic Media / Web Creative / Use of Multi-Media).
Project frames
Production stills
