img.dsxzai.com：Image 2.5 案例 + 提示词画廊（大师的AI小灶） img.dsxzai.com: Image 2.5 case + prompt gallery (by @dashiAIxz)

来源Sources

为什么推荐Why we recommend it Image 2.5 提示词看板的第一张卡：一个可逛的第三方提示词画廊，不是模型卡，也不是 Grok Bot 用法。主源是 img.dsxzai.com（2026-09-14 本库打开返回 200，Cloudflare 前置），社区源是作者 @dashiAIxz 的 X 帖（可达，正文与附图已核）。「复制就能生成同款」「能力很强没多少人关注」均为作者自述；站内每条提示词与原帖作品的对应关系以该条的整理说明为准，本库未逐条打开。站点带「提交案例」与「管理」入口，是作者自建站，本库不为其背书，也未测试提交流程。不写点赞、浏览等互动数字；条数只记接口当日自报值并标明口径。预览图取自作者 X 帖附图（站点首屏截图，裁掉浏览器边框后取中 3:2），非本库现场截图 — 无头浏览器打开时首屏图片仍在渐入动画中，故未用现场截图。 First card on the Image 2.5 prompts board: a browsable third-party prompt gallery — not a model card, not a Grok Bot playbook. Primary source is img.dsxzai.com (returned 200 behind Cloudflare when this library opened it on 2026-09-14); the community source is the author @dashiAIxz’s X post (reachable; text and attached image checked). “Copy it and you get the same result” and “strong but under-noticed” are the author’s claims; the mapping between each on-site prompt and its original post is whatever that case’s curation note says — not opened one by one here. The site carries “submit a case” and “admin” entries and is the author’s own build; no endorsement, and the submission flow was not tested. No likes, views, or other engagement figures are recorded; the only count noted is the API’s self-reported total on the day, with its scope stated.

要点（作者称 / 本库核到）Highlights (author’s claims / checked here)

• 作者帖原文：「GPT6 太耀眼了，导致 Image 2.5 虽然能力很强没多少人关注到。最近花时间整理了一下网上 Image2.5 相关的案例和提示词，做了个网站来呈现。每个案例都有提示词和来源，复制就能生成同款。」帖子发于 2026-09-14，附一张站点截图。 The author’s post, verbatim (translated): “GPT-6 is so dazzling that Image 2.5, strong as it is, got little attention. I spent some time collecting Image 2.5 cases and prompts from around the web and built a site to present them. Every case has its prompt and source — copy it and you get the same result.” Posted 2026-09-14 with one site screenshot.
• 站点可核到的：页面标题「生成提示词案例库」，顶部按模型分栏（截图里可见 全部 / Image2.5 / Nano banana / Midjourney），标签行含品牌视觉、信息图、摄影、海报、字体设计、清爽、未来感、插画、极简、复古、空间建筑、角色IP、时尚、食物、人像、产品、界面；带搜索框、「提交案例」入口与 中/En 切换。 Visible on the site: page title 「生成提示词案例库」; model tabs across the top (the screenshot shows 全部 / Image2.5 / Nano banana / Midjourney); a tag row with brand visual, infographic, photography, poster, typography, clean, futuristic, illustration, minimal, retro, architecture, character IP, fashion, food, portrait, product, UI; plus a search box, a “submit a case” entry, and a 中/En toggle.
• 每条案例的结构：标题、一句整理说明（如「中文整理提示词，原帖公开一句具体提示词；无需参考图」）、完整提示词、模型名、来源 URL 与来源作者、来源发布时间、标签与图片；站内按钮为「复制提示词」「查看来源」。提示词是否与原帖作品逐字一致，以每条的整理说明为准，本库未逐条核对。 Per-case structure: title, a one-line curation note (e.g. “prompt curated in Chinese; the original post published one concrete prompt; no reference image needed”), the full prompt, model name, source URL and source author, source publish time, tags and images; on-site buttons are “copy prompt” and “view source.” Whether each prompt matches the original post verbatim is whatever that case’s note says — not checked one by one here.
• 站点是 Vite/React 单页应用，公开接口 /api/cases 分页返回案例；「全部」分栏的接口 total 字段在本库 2026-09-14 打开时为 485，但这是全部模型合计、由站点自报，Image 2.5 单独条数本库未复数。 The site is a Vite/React single-page app; a public /api/cases endpoint pages through the cases. Its total field for the “all” tab read 485 when this library opened it on 2026-09-14 — but that is the site’s own figure across all model tabs; the Image 2.5-only count was not re-tallied here.