{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreia2ggykn37kwzpalurdslmxnujhcmosopkkvwpndyvz6khtie2ogq",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mmpwvh4dkjr2"
  },
  "path": "/t/gpt-image-2-image-creator-and-editor-app-python-desktop-ui-message-13/1370104#post_17",
  "publishedAt": "2026-05-26T00:45:30.000Z",
  "site": "https://community.openai.com",
  "textContent": "## Update 2025-05-25\n\n_See updated reply #13 above to grab it._\n\n### Feature changelog\n\n  * Prompt inserts: a quick list of text additions to the end of a prompt, good for tune-ups or reused style. Hard-coded, for your manual additions. (and don’t forget a right-click to ask an AI about your selection).\n\n\n  * Aspect ratio size selector: quickly find only landscape or portrait resolutions, and the portrait aspects are copied from landscape model presets in the code. The resulting drop-down of resolutions is shorter - and no other choice needed in models not `gpt-image-2`.\n\n\n  * Resolution ratio annotation: automatically generated UI help, with model presets matched against common nearby image format ratios they indicate, otherwise decimal.\n\n\n  * Improved transparency features and algorithms\n\n    * Expansion: remove more of AI’s edge blending into an image around detected transparency pixels, heuristically\n    * Background: apply an underlying background color to the transparency to match where it will be shown, for tinting the smoothing, or a fallback when transparency fails\n    * normalize an image immediately, maximize contrast either to improve detection of bright or dark transparency, or to increase the remaining contrast within transparent image contents. With “normalize all colors”, it can also shift color balance from a sepia-tone `gpt-image-1` symptom.\n\n\n  * A file load to canvas no longer targets a set DALL-E size by default, as we can send arbitrary size to the remaining models’ vision, within reason. Encourages 1:1 loading for applying transparency tool to image files.\n\n\n\n\n\nAfter this code bloat, refactored and unified API requests and history items, and a mini-edit canvas at startup of an unseen “size” that can’t be accidentally sent.",
  "title": "Gpt-image-2 image creator and editor app - Python, desktop UI (message #13)"
}