Google Imagen 3 (NB2) can take flat black-on-white text and render it as hand-painted letters—physical paint, visible texture, stroke weight variation, conforming to surface geometry—but give it alien glyphs instead of recognizable script and the perspective warp… well it don’t work so great. This breakdown documents the annotated reference image technique that solved the placement problem, the Photoshop workaround that solved the warp problem, and the follow-up experiments that turned a workflow quirk into a genuinely strange finding about what NB2 does and doesn’t understand about geometry. Hope you have your sacrificial exotic chicken handy. You’re gonna need it.
Editing Images with Gemini 3.1 Flash Image
Google’s own documentation doesn’t mention it. Nobody else has reported it. And yet: Google’s new GenAI flagship “Nano Banana 2” reads annotations in reference images—red dots, text labels, arrows pointing at specific parts—and acts on them with precision. I’ve reproduced it across multiple generations. It’s not luck. It can also transfer materials between images, swap elements without touching the rest of the frame, and apply patterns to draped fabric with enough fidelity that the result looks like the cloth actually exists. All of that is impressive. None of it is the interesting part. Here’s what it can do, how far it goes, and where it still falls apart.
AI is Erasing Entire Ethnic Groups by Default—And So Are Artists
AI image generators can’t see my character. Sarai—a Central Asian woman with copper-bronze skin and freckles—doesn’t exist in their training data. After dozens of failed generations across thirteen different models, I documented exactly what goes wrong and built a workflow to fix it: using AI as compositional scaffolding while correcting ethnicity, skin tone, and features manually. This piece breaks down the specific failure modes (phenotype collapse, extreme skin tone overcorrection, Instagram-mom glamorization), shows the eight-step process I used to get accurate results (well, the best I could manage anyway), and explains why “just commission a human artist” produced the same erasure. For writers with characters from underrepresented populations, here’s what you’re up against—and how to fight it.
Picking the Best AI Video Model for Book Promo Videos and Trailers
I’ve been drowning in AI video model options while building promo videos and trailers for “Doors to the Stars.” Google Veo 3? Kling 2.5? Runway Gen-4? Sora 2? The marketing claims all sound identical—until you actually test them. Mixing the wrong model to the wrong shot wastes hours (and dollars) generating unusable footage. I’ve researched the major players included in Freepik’s umbrella subscription to figure out which models excel at what, and the differences matter more than you’d think. The lessons I learned about matching AI video tools to actual storytelling needs will save you time and money.
Why I Don’t Stress About Audiobook Deadlines (And You Don’t Have to Either)
My publisher expected weeks of work on Podium Entertainment’s audiobook production forms. Monday morning they’ll find complete submission materials in their inbox—detailed character breakdowns, comprehensive content warnings, and a 200+ entry pronunciation guide covering four distinct linguistic systems. Total time invested: one Friday evening. Here’s the exact process I used with Claude Sonnet 4 to compress what’s typically 2-3 weeks of tedious work into 5 focused hours. For a 136,000-word space opera with body-swapping operatives, multiple aliases, and invented languages, systematic beats casual every time. Use AI for data processing, not creative judgment. Keep your weekends free.