Description

Turn 1–8 characters into a single star-studded group portrait: anthropomorphized AI models, virtual idols, game characters, famous IP characters, and brand mascots are all supported. Each character gets their own colors, outfit, props, and pose, arranged at different depths and interacting with one another instead of standing in a row. Each person also gets a text card with their name, tags, and a line of dialogue. 🆕 Updated October 2026 · Upgraded to GPT Image 2.5: Flare is used by default for fast image generation; say “premium / poster / key visual / high resolution” to automatically switch to Sunburst for more accurate Chinese cards and a finish closer to an official promotional image · Generate directly when there’s enough information: the “show the complete plan first and wait for confirmation” step has been removed. When information is missing, all necessary questions are asked at once instead of asking about each character individually · Better at getting the references right: AI models and famous IP characters use widely recognized public personas and designs by default, such as DeepSeek’s blue, chubby fish · Less crowding: for 5–8 people, cards are automatically simplified to “name + one line of dialogue” · Uses fewer credits: the outdated multi-model fallback and retry chain has been removed, reducing wasted generations and back-and-forth conversation so the same image costs fewer credits Examples: “Turn Claude, GPT, Gemini, and DeepSeek into a group portrait using their public personas, landscape format” or “Draw portraits of my 4 original characters, vertical phone wallpaper”