HairDeck: animation-ready stylized hair cards from a single image, exported as GLB or VRM
Tsinghua/CMU/CUHK researchers generate stylized hair cards with spring-bone chains from one image — per-card UVs, two-tone color, GLB/VRM export. Apache-2.0.
Published entries across all sections carrying the “Image generation” tag, newest first by publication date on this site.
32 entries
Tsinghua/CMU/CUHK researchers generate stylized hair cards with spring-bone chains from one image — per-card UVs, two-tone color, GLB/VRM export. Apache-2.0.
Tencent WeChat Vision's open VLM unifies detection, OCR, counting, segmentation and alpha matting in one model. 2B/4B sizes, NOASSERTION license, 1k stars.
yang0's handraw-style numbers 365 drawing styles, 170 layouts and 36 theme colors; pick numbers, get bilingual image prompts. MIT with attribution.
An open-source agent skill turning Chinese stories into 3:4 hand-drawn flipbook MP4s — 327 style assets, hand-written text in the art, Remotion rendering, MIT.
carolinaaafy's Codex skill turns travel photos into sticker cards via any image-capable agent, ready for Goodnotes journals. Non-commercial license.
Open-source clothing reference gallery: 1,407 model-free AI garment images, bilingual search, and an optional try-on backend wiring GPT Image or Gemini. MIT.
SuperCMO Skills is Apache-2.0: 23 marketing skills plus a local MCP server that turn a product photo and a brief into UGC or ad videos, with your own keys.
Canva released Craft on October 22: its first in-house image model plans a layout before rendering, tops out at 2K, and ships in Dream Lab and the API at once.
Three aurora alpha models flashed in OpenAI's API list on October 18 and were pulled; blind testers say they generate images. Unconfirmed.
A Codex skill from LiamGvchi that compresses a theme, sentence or reference set into a sparse paper poster with one colour accent, or a reusable image prompt.
A Codex skill from wnby that keeps the real photo on top and paints a quiet, paper-textured relic below, for series covers and art posters. MIT licensed.
Wardrobe is an MIT local-first tool: gpt-image extracts garments from photos into clean product shots, with the data kept on your machine.
Denmark’s Supreme Court upheld 18 months in prison for a 31-year-old with over 36,000 AI-generated CSAM images: fictional subjects are no ground for leniency.
Ant Group's Ling team open-sourced Ming-Image-0.1-Design on 22 September: two 6B models for text-heavy design generation and image layer splitting, MIT.
ComfyUI exposes every step of a diffusion pipeline as nodes: models, samplers and parameters all adjustable, with the largest open models runnable on 4GB VRAM.
The image model improves Chinese text rendering, adds 4K, and halves per-image cost: $0.050 at 2K and $0.076 at 4K.
A US test starts this month with labeled placements beside image generations, plus a wider ad-measurement partner ecosystem.
The four-line prompt from X that asks an AI to photograph its own ordinary life — realism comes from asking for imperfection, and every run differs.
A one-line 3D app icon prompt: a rounded tile in glass and chrome with disco-ball mosaic facets. Bilingual by design — swap the name and render the set.
An exploded-view product poster prompt written as JSON: layout, layers and callouts locked field by field — swapping products is a data edit. For GPT Image 2.
A watch-encyclopedia prompt written as a Python class: parents set the look, mutations set the info layers, weights set the mix — one render, a 2x2 grid.
A preschool vocabulary card prompt: whole fruit left, cut segment right, joined by a dotted arrow and a stick-figure guide — one image, both words.
A black-stage poster prompt where six laser planes need origin, perspective and thickness — with explicit bans against sliding into club-flyer neon.
A mid-century travel poster prompt held to three colors — cream paper, black linework, one accent. Swap the city and accent to print a whole series.
A system prompt for Apple-keynote-style nature posters: one hero animal at poster scale, facts in four icon columns, and a ban list that strips the cheap look.
A photorealistic metro portrait prompt: skin as color codes and optical events, the environment as an active interferer, five closing bans holding the style.
A typography poster prompt that treats the title as the artwork: weight, distortion and whitespace all serve the meaning — a new title is the only edit.
A vertical infographic prompt that maps a whole city from sky to bedrock, holding density with color-coded flows and numbered panels. For GPT Image 2.
A felt-diorama prompt that builds a country from three depth bands — street life, connecting flow, one landmark anchor — with element counts bracketed.
Black Forest Labs shipped FLUX 3: native 4K, coordinate-grid element layout, 10 positioned reference images and region editing; open weights within weeks.
An MIT-licensed open skill that puts hard limits on how an agent draws IP mascots: 4–7 large shapes per silhouette, three colors at most, legible at 32×32. It pitches three directions first, then generates six candidates in one batch. For developers tired of AI mascots that turn to mush at icon size.
A commercial still-life prompt for a perfume bottle. Subject, material, set, lighting, lens and composition each get their own line, so swapping in another product means editing only the first line. Written for image models such as GPT Image and Midjourney.