SuperPicky 4.5.0 RC5

最新测试版(候选发布)· 2026-07-13 Latest Beta (Release Candidate) · 2026-07-13

这是 4.5.0 的候选发布版本,供愿意尝鲜的用户测试。日常使用推荐 4.3.0 LTS 稳定版。 遇到问题欢迎到 GitHub Issues 反馈。 This is a release candidate for 4.5.0, offered for early testing. For day-to-day use we recommend the 4.3.0 LTS stable build. Please report issues on GitHub Issues.
安装包 Package 下载渠道 Downloads
macOS (Apple Silicon) ≈ 800 MB GitHub 大陆镜像CN Mirror Google Drive Baidu (renc)
Windows Lite ≈ 200 MB,首启自动下载模型 ≈ 200 MB, models fetched on first launch GitHub 大陆镜像CN Mirror Google Drive Baidu (zjaf)
Windows Full (CPU) ≈ 750 MB,模型全内置 ≈ 750 MB, models bundled GitHub 大陆镜像CN Mirror Google Drive Baidu (mijt)
Windows Full CUDA ≈ 2.1 GB,NVIDIA GPU 加速 ≈ 2.1 GB, NVIDIA GPU acceleration Google Drive Baidu (knw4) 体积超 GitHub 附件限制,仅网盘分发 Exceeds GitHub asset size limit — cloud drives only

自 4.3.0 以来的变化 What's New since 4.3.0

评星算法全新升级:批内相对 + 配额 Batch-Relative Star Rating

星级不再依赖固定阈值——3 星 = 本批照片里最好的前 N%(默认 20%,5–50% 可调,与摄影水平档位联动)。开启识鸟时配额按鸟种分配:每种鸟都保住自己最好的照片, 罕见鸟至少保底 1 张,常见鸟的大连拍不再霸榜。美学评分改为只看鸟体裁剪区,背景不再干扰。 想用旧的绝对阈值算法?设置中心一键切回。 Star ratings are now batch-relative: the best N% of each batch get 3 stars (default 20%, adjustable 5–50%, mapped to skill levels). With Bird ID on, the quota is applied per species — rare birds always keep their best shot, and a long burst of one common bird no longer crowds out the others. Aesthetics now scores the bird crop instead of the whole frame. The legacy threshold algorithm remains available in Settings.

平铺模式:不移动任何文件(Lightroom 用户必看) Flat Layout — Zero File Moves (Lightroom users, read this)

新增「平铺」目录布局:照常检测、评星、写关键字和色标,但所有文件留在原地—— Lightroom 已导入目录的引用不会断。连拍也可以只检测不分子文件夹。 New "Flat" folder layout — rate in place, no file moves. Photos are detected, rated and tagged as usual, but every file stays exactly where it was — your Lightroom folder references never break. Burst detection can now also skip creating subfolders.

鸟种写入 Lightroom 关键字 Species Keywords

识鸟结果同步写入标准关键字(XMP Subject),LR 关键字面板直接按鸟种筛选;只增不删, 你自己的关键字绝不会被动,重跑也不会产生重复。 Species names are written to standard keywords (XMP Subject) — filter by species right in Lightroom. Merge-only: your own keywords are never touched, and re-runs never duplicate.

色标新含义(老用户请注意) New Color-Label Defaults (please note)

含义Meaning Old New
飞鸟Bird in flight 绿Green Blue
精准合焦Critical focus Red 绿Green
脱焦Soft / out of focus none Red
如果你在 LR 建过「绿=飞鸟」的智能收藏夹,请改为蓝色。 If you built Lightroom smart collections on "green = flying", update them to blue.

全新设置中心 & 浏览器增强 Settings Center & Results Browser

所有设置统一进带左侧导航的设置中心,首页保留快捷面板双向同步。结果浏览器支持 键盘打星(数字键 0–3 直接设星,↑/↓ 加减星),缩略图同时显示鸟名 + 文件名,「精选」皇冠角标,可直接修改鸟种。 All settings unified into one Settings Center with sidebar navigation. The Results Browser adds keyboard rating (0–3 to set stars, arrows to step), species + filename shown together, crown badges for picks, and in-browser species editing.

更快、更稳 Faster & More Reliable

ExifTool 读写分离双常驻进程,消除每约 30 张卡顿 3 秒的问题(152 张实测总耗时 75.5s → 62.1s);识鸟驻留服务秒级就绪,Lightroom 插件不再等待十几秒;大量稳定性修复。 另:4.5.0 起应用内不再检测在线更新,新版本请关注本站或 GitHub Releases ExifTool now runs as dedicated read/write resident processes — no more periodic stalls (152-photo test: 75.5s → 62.1s). The Bird ID service is ready within seconds; many stability fixes. Note: in-app update checking is retired as of 4.5.0 — watch this site or GitHub Releases.

完整变更明细(ChangeLog)Full ChangeLog

特别感谢 Paul Gregory 的细致测试与反馈——本版的平铺模式、鸟种关键字、 键盘打星与色标新默认值均源自他的建议。 Special thanks to Paul Gregory for his thorough testing and feedback — the flat layout, species keywords, keyboard rating and new color-label defaults all grew out of his suggestions.