
What happened
kensaurus said on November 13 that he made six new art styles of his mascot Kenji for seven apps, using fal's google/nano-banana-2.1/edit and fal-ai/birefnet/v2, with 63 generations costing $4.68.
Why it matters
The post shows a repeatable pipeline — an approved style sheet, a fixed identity paragraph in every prompt, and common foot alignment — that keeps one character recognizable across styles, and four of the seven apps already ship it.
Summaries like this, every morning.
The post is a build log rather than a product launch: kensaurus, who runs the one-person app studio kensaurus, set out on November 13 to give seven of his apps the same mascot without forcing one visual style on all of them. In a Thai learning game the character appears as a pixel-art sage; in a philosophy book app, as a woodblock-print moon reader; in a budgeting app, as a stamp-pressing bookkeeper. The record covers the six newly drawn styles (the glot sage predates this effort and is excluded), and of the seven apps four already contain the character.
The method hinges on a single approved style sheet per app. Each pose is generated against that approved sheet at 1:1 and 2K, then cut out with fal-ai/birefnet/v2 and re-centered so the feet sit on the same baseline. That foot alignment is what lets poses swap on screen without the figure jumping vertically. The character description is kept as one fixed block pasted verbatim into every prompt, because rewriting the wording slightly shifts the face from generation to generation.
Costs are logged line by line in spend.jsonl: eight style sheets at $0.59, 49 poses at $3.64, and six icon and stamp candidates at $0.45, totaling 63 generations and $4.68. The figure uses $0.0743 per 2K edit from fal's published prices and is not a measured invoice, and the 48 cutouts are left out because birefnet's API page shows no price. Beyond the art, kensaurus assigned each app a role and tone for Kenji — a stage-side coach for lets-talk, a moon reader for the wanting mind, a bookkeeper for yen-yen — with lists of things the character must not say, and rules such as one Kenji per screen and no mascot in permission prompts or shared images.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Prada's Holiday 2026 campaign, "Drawing Prada: Animated by Hand," notched 3.3 million YouTube views, while a f…

The item is published under the heading "AI NSFW 2026", framing a collection of reference tools, reusable work…

The developer traced that Codex read his Notion profile each time it made an image, because an AGENTS.md instr…

SHINING 3D announced the EINSTAR EZ GO, a wireless AI-equipped 3D scanner that connects to smartphones, tablet…

A Zenn guide says the "AI-ish" look on landing pages comes from vague prompts, not bad taste, and lays out a 9…

Using PixAI's Tsubaki.3, the author made six expressions of one character
