Repository navigation
Releases: hongye007/open-pose-costume
Release list
v2.0.2 · 参考选择优化与文档更新
本次更新优化一键生图的参考选择体验,并更新中英文文档。
- 姿势、服饰选择弹窗统一留白、标题与卡片样式。
- 顶部搜索与分类常驻,素材列表独立滚动并自动加载。
- 姿势选择新增深度图开关,与生成参考同步。
- 更新中英文 README 的参考库、生成工作台截图与流程说明,补充服饰提取、本地历史及预览对齐。
验证:30 项测试通过,生产构建通过;已检查桌面与手机布局、深度图切换、空结果及滚动加载。
Improves reference selection in the generation studio and refreshes bilingual documentation.
- Refined pose and outfit picker layouts and image cards.
- Persistent search and category controls above an independently scrolling, automatically loading list.
- A depth-map toggle in the pose picker, synchronized with generation references.
- Updated Chinese and English library/studio screenshots and workflow, extraction, local-history and preview-alignment documentation.
Validation: all 30 tests and the production build passed; desktop/mobile layouts, depth switching, empty states and incremental loading verified.
v2.0.1 · 叠加对齐与滑杆优化
改进叠加预览中构图不一致时的对齐体验,并缩小对比滑杆。
更新
- 新增“调整对齐”模式,可分别缩放参考图和生成图,范围为 25%–400%。
- 支持拖动图片调整位置、方向键微调,按住 Shift 可增大移动步长。
- 对齐时半透明叠加两张图,点击“完成对齐”恢复滑杆对比;支持一键重置。
- 滑杆改为 24px 浅色圆钮,缩短分隔线,减少对画面的遮挡。
- 提供完整中英文文案,已检查手机布局。
对齐仅影响预览,不修改原图或生成文件。
验证
30 项测试、TypeScript/Vite 构建及格式检查通过;已验证缩放、鼠标拖动、键盘移动、重置和滑杆交互。
English
This patch adds independent zoom (25%–400%) and positioning for reference and generated images in overlay previews. Align with a transparent overlay, drag or nudge using arrow keys (Shift for larger steps), then return to the comparison slider. A reset control restores both images and the split position.
The slider now uses a compact 24px light handle and a shorter divider. Chinese/English copy and mobile layouts are included. Preview adjustments do not modify the original or generated files.
All 30 tests, the TypeScript/Vite build and formatting checks passed.
Full changelog: v2.0.0...v2.0.1
v2.0.0 · 一键生图工作台与服饰提取
将一键生图从弹窗升级为独立工作台,完成参考选择、服饰提取、生成预览与历史复用流程,对应提交 20d16d5。
更新
- 保留侧边栏导航,生成界面嵌入右侧内容区;切换图库时保留当前搭配和生成状态。
- 姿势、服饰、人物三类参考选择,支持随机搭配;素材选择器滚动到底自动加载。
- 上传服饰直接打开提取弹窗:左侧选图、中间提取、右侧查看结果,确认后才应用到搭配台;支持失败重试和取消。
- 可上传人物参考,编辑画风、尺寸、补充描述及高级提示词。
- 支持仅结果、双列和叠加预览,图片放大与下载,以及将生成结果用作人物参考继续创作。
- 右侧历史记录抽屉,在当前浏览器保留最近 20 次结果,可恢复搭配、删除记录,并将该次使用的参考原图打包下载为 ZIP。
- 完整中英文适配,缩小叠加对比控件并优化工作台布局。
使用说明
- 需自行配置支持
/v1/images/edits和浏览器 CORS 的生图服务,默认模型为gpt-image-2。 - 服饰提取会将上传图片发送至所配置的服务,并额外调用一次生图接口,费用以服务商为准。
- 历史记录仅保存在当前浏览器;成功生成时,结果及上传参考会保存在本地历史中。
验证
30 项测试通过;GitHub Pages 构建和部署成功。
English
Image generation now runs in a dedicated workspace within the existing sidebar layout. This release adds random reference combinations, automatic material loading, an outfit extraction dialog with explicit confirmation, uploaded outfit and character references, configurable prompts and preview modes, and a local history drawer with reference ZIP downloads.
Outfit extraction uses the configured image service and incurs an additional image request. History stores the latest 20 results and uploaded references in this browser. All 30 tests and the GitHub Pages build and deployment passed.
Full changelog: v1.0.0...v2.0.0
v1.0.0 · 参考图库与搭配台
改版前的参考图库与弹窗搭配台版本,对应提交 0cc039e。
功能
- 姿势、服饰、人物参考库,中英文搜索、筛选、收藏及原图查看和下载。
- 弹窗搭配台组合姿势、服饰和可选人物参考,通过兼容 OpenAI 的图片编辑接口生成角色。
- 双列预览、叠加滑杆对比,以及姿势视角与深度参考切换。
此 Release 保留旧版代码快照;在线网站继续运行最新版本。
English
The reference gallery and modal-based pairing workspace before the generation studio redesign, at commit 0cc039e. Includes bilingual browsing, favorites, optional character references, compatible image generation, split/overlay comparison and pose depth references.