Skip to content

v0.12.10 - Face-aware cover cropping, custom actress photos & pre-organize editing

Choose a tag to compare

View all tags
@github-actions github-actions released this 25 Jul 15:28
· 606 commits to main since this release
v0.12.10
3b82f66

English version below ↓

一行安裝（推薦）

macOS:

curl -fsSL https://raw.githubusercontent.com/slive777/OpenAver/main/install.sh | bash

Windows (PowerShell):

irm https://raw.githubusercontent.com/slive777/OpenAver/main/install.ps1 | iex

💡 不想開 PowerShell 的 Windows 用戶：也可以到下方 Assets 下載 OpenAver-Windows-Setup.bat，雙擊即可跑完同一套安裝。

自上次發布（v0.11.12）以來累積了整個 v0.12 系列的更新，主軸有五個：無碼封面自動對準人臉裁切、女優照片可以自己上傳、唯讀來源會直接讀用現成資料不再重刮、拖檔整理前可先手動修正演員與發售日，以及手機瀏覽體驗更完整。

🎯 無碼封面自動對準人臉

無碼片（FC2、素人、uncensored 廠牌）的封面會自動偵測人臉裁切。有碼片維持原樣。

  • 可手動微調：點焦點 icon → 自動對焦（跑一次偵測）→ 左右拖曳亮窗微調到你要的位置 → ✓ 存 / ✗ 取消。就算偵測抓不到臉（歪臉、藝術照），也能直接拖到想要的位置存起來。
  • 唯讀來源、換封面重刮也一併對焦；桌面寬螢幕看片時封面本來就完整，對焦鈕只在手機／窄視窗（封面被裁成直式）時才出現。
  • 媒體伺服器海報也對臉：給 Jellyfin／Emby／Kodi 掃描的實體海報檔（-poster.jpg）在產生時也會以臉為中心裁切。

📤 女優照片可以自己上傳，也能手動對焦

女優牆的照片以前只能靠系統自動抓（線上圖庫或本機影片封面），想換成自己喜歡的照片得手動塞檔案。現在可以直接上傳。

  • 燈箱右上角一顆上傳鈕，桌面選檔或手機開相簿都能用（JPEG／PNG／WebP／GIF 都吃），選好就直接變成主圖。
  • 女優頭像也能手動對焦（和影片同一個位置的對焦鈕）：照片明顯偏寬、真的有左右可調空間時才出現，只左右拖、與影片一致；抓不到臉也能手動拖到想要的位置存。
  • 改過名（有別名）的女優換圖優化：重抓時雲端查詢會自動輪流用她的每個名字查，撈得到掛在舊名下的線上照片。

📀 唯讀來源更聰明：直接讀用現成資料，不再無腦重刮

唯讀來源（通常是雲端分享盤）如果本身已經是整理好的媒體庫——影片旁邊就附了 .nfo 和封面圖——OpenAver 現在會直接讀用它現成的資料，不再無視它、傻傻重新上網刮一次。更快、更省流量，撞號抓錯的問題在已整理的庫直接消失。

  • 產生列表時，來源影片旁若有 .nfo 就直接讀它寫入資料庫（片名／番號／女優／片商／標籤／日期…），完全不連網；旁邊有封面圖就複製一份使用。
  • 即使檔名被改成片名、抽不出番號，只要 NFO 裡有番號一樣能就地建立。
  • 已整理成 Jellyfin／Emby 格式（-poster + -fanart）的圖會原樣複製，完整保留你 curate 的海報與背景圖。
  • 唯讀來源的片現在也能用放大鏡補料、齒輪重刮、補劇照（以前是鎖住的）——所有寫入一律導向獨立輸出夾，來源盤零寫入。

✏️ 拖檔整理前，可先手動修正演員與發售日

把影片檔拖進搜尋頁、整理進庫之前，現在可以先看一眼並手動修正。scraper 抓錯／漏抓演員、無碼片沒有發售日，都能在整理前直接改，修正會隨整理一起寫進 NFO ＋ 資料庫。

  • 演員可編：點鉛筆改名／加人／刪人／修「現名(舊名)」這種括號格式都在同一格；多人用頓號「、」、全形逗號「，」或半形逗號「,」分隔皆可。共演漏抓、無碼片沒帶演員都能在這裡補齊（逐字存，打什麼存什麼）。
  • 發售日補日期：已經抓到的直接顯示、不動它；只有沒抓到日期的無碼片才出現一個日曆，讓你點一下挑日期補上（只能日曆點選，不會有格式跑掉的問題）。
  • 純用關鍵字搜尋瀏覽時一律唯讀——那時沒有「整理」這個動作，不會冒出任何編輯鈕。
  • 整理時送出你實際挑選的那個來源版本：以前切到第二、第三個版本去看，按整理卻固定送出「第一個」——現在一律送出你當下正在看的那一版。

📱 手機體驗更完整，開機自動看一眼有沒有新版

手機瀏覽收藏更好看，桌面版開機會在背景幫你看一下有沒有新版本。

  • 手機上封面吃滿寬度：卡片左右的大片留白收窄，封面明顯變大（同樣 3 欄、每張大一圈），跟頁面上方的搜尋列一樣貼近邊緣；桌面／平板寬螢幕完全不變。
  • 觸控時封面更純粹：手機／平板上封面不再常駐壓著一排操作 icon（播放、開資料夾、補資料），想操作就點封面進燈箱——播放／補資料都在燈箱裡。缺封面的破圖卡例外，補資料仍是一點到位。搜尋結果牆一併統一，桌面滑鼠 hover 浮出 icon 的行為完全不變。
  • 「開資料夾」鈕在純觸控裝置上收起：手機、平板、LAN 端瀏覽器上這顆鈕只能把一段別人桌面的路徑複製到剪貼簿、開不了遠端資料夾，按了沒用。接了滑鼠／觸控板的二合一裝置照常顯示。
  • 女優牆卡片大小不再「乎大乎小」：以前同一個視窗寬度下，女優卡與影片卡的大小關係不固定（窄視窗女優偏大、寬螢幕女優偏小，切換模式會「乎大乎小」）。本版改為兩段式對齊——手機／平板（窄視窗）女優牆與影片牆同欄同寬、切換模式完全一致；桌面寬螢幕因為影片封面在這個尺寸會切成橫式劇照、女優照永遠是直式，硬要同寬會讓女優卡高度接近影片卡兩倍，所以女優牆改用自己的一組欄數（1920px 全螢幕下女優 7 欄、影片 5 欄），對齊的是「一列的高度」而不是卡片寬度。結果：每個視窗寬度下女優卡都不會比影片卡大，大小關係固定可預期。
  • 啟動時自動檢查更新（桌面版）：開機後背景靜靜去 GitHub 看一次有沒有新版，真的有才在側邊欄鈴鐺推一則通知（已是最新就完全安靜、不打擾）；Help 頁多一個「啟動時檢查更新」開關（預設開）。離線／逾時一律安靜略過，不拖慢開機。

⚡ 其他改進與修復

  • 深色主題幾顆按鈕看清楚了：深色（dim）主題下 Help「檢查更新」、選資料夾、搜尋錯誤態的上一個／下一個等邊框式按鈕，原本邊框與文字幾乎與卡片同色、看不見，本版一次修好整類問題。
  • 滑鼠滾輪也能切上一片／下一片：搜尋頁詳情、燈箱（影片／女優）、劇照瀏覽——橫向滾輪撥＝左右方向鍵；燈箱和劇照這類全螢幕檢視裡垂直滾輪也能切。
  • 拖檔的番號辨識更完整、與貼上一致：以前拖檔比貼上少認幾種格式，同一個檔拖進去可能失敗；本版兩邊走同一套規則，認不出番號時也會明確提示。
  • 另外也做了多項資料正確性、穩定性與安全性修正。

English

This release bundles the whole v0.12 series since the last published release (v0.11.12). Five highlights: uncensored covers auto-center on the face, actress photos you can upload yourself, read-only sources that read existing metadata instead of re-scraping, editable actors / release date before organizing a dropped file, and a more complete mobile browsing experience.

🎯 Uncensored covers auto-center on the face

Uncensored covers (FC2, amateur, uncensored labels) are now auto-cropped around the detected face. Censored covers are left as-is.

  • Manual fine-tuning: click the focal icon → auto-focus (runs one detection) → drag the highlight window left/right to where you want it → ✓ save / ✗ cancel. Even when no face is detected (tilted faces, art shots), you can just drag to the spot you want and save.
  • Read-only sources and cover re-scrapes get focused too; on a wide desktop screen the cover is already complete, so the focal button only appears on phones / narrow windows (where the cover is cropped to a portrait poster).
  • Media-server posters get face-centered too: the physical poster file (-poster.jpg) that Jellyfin / Emby / Kodi scan is also cropped around the face when generated.

📤 Upload your own actress photos, with manual focus

Actress-wall photos used to come only from automatic sources (online galleries or local video covers) — swapping in one you actually like meant dropping a file in by hand. Now you can just upload.

  • An upload button in the top-right of the lightbox, usable from a desktop file picker or your phone's photo library (JPEG / PNG / WebP / GIF all work); pick one and it becomes the main image immediately.
  • Actress portraits can be manually focused too (the same focal button as videos): it only appears when the photo is clearly wide with room to adjust left/right, drags left/right only, consistent with videos; even with no face detected you can drag to the position you want and save.
  • Renamed (aliased) actresses' photo swaps improved: on re-fetch, the online lookup now cycles through each of her names automatically, so it can pull photos filed under her old name.

📀 Read-only sources got smarter: read existing data instead of blindly re-scraping

If a read-only source (usually a mounted cloud drive) is already an organized library — with .nfo and cover art sitting right next to each video — OpenAver now reads that existing data directly instead of ignoring it and dumbly re-scraping online. Faster, less bandwidth, and ID-collision mismatches simply disappear on an already-organized library.

  • When generating the list, if a source video has a .nfo next to it, it's read straight into the database (title / ID / actress / studio / tags / date…), fully offline; a cover image next to it is copied for use.
  • Even if the filename was changed to the title and no ID can be extracted, as long as the NFO has an ID it can still be created in place.
  • Art already in Jellyfin / Emby format (-poster + -fanart) is copied as-is, fully preserving the poster and backdrop you curated.
  • Read-only videos can now use enrich (magnifier), re-scrape (gear), and fetch-stills (previously locked) — all writes go to a separate output folder, zero writes to the source drive.

✏️ Fix actors and release date before organizing a dropped file

Before you drop a video into Search and organize it into your library, you can now take one last look and fix things by hand. Scraper got the actors wrong or missed some, an uncensored title has no release date — all editable before organizing, and the fixes are written into the NFO + database along with the organize.

  • Editable actors: click the pencil to rename / add / remove people, or fix the "current name (old name)" bracket format — all in one field; separate multiple people with a Chinese comma "、", full-width comma "，", or half-width comma ",". Missed co-stars or an uncensored title with no listed actors can be filled in here (saved verbatim — what you type is what's stored).
  • Fill in the release date: dates already scraped are shown and left untouched; only an uncensored title with no scraped date gets a calendar to click a date in (calendar-only, so there's no date format to get wrong).
  • Keyword-only search stays read-only — there's no "organize" action there, so no edit buttons appear.
  • Organize submits the exact source version you picked: previously, if you flipped to the 2nd or 3rd version to look, organize still submitted the "first" one — now it always submits the version you're currently viewing.

📱 A more complete mobile experience, with an on-launch update check

Collection browsing looks better on the phone, and the desktop app checks for a new version in the background on launch.

  • Covers fill the width on phones: the big side margins are narrowed and covers get noticeably bigger (same 3 columns, each one size larger), hugging the edges just like the search bar above; desktop / tablet wide screens are unchanged.
  • Cleaner covers on touch: on phones / tablets, covers no longer permanently carry a row of action icons (play, open folder, enrich) — to act, tap the cover to open the lightbox, where play / enrich live. Broken-image cards with no cover are the exception, keeping enrich one tap away. The search result wall is unified the same way, and desktop mouse-hover icons are completely unchanged.
  • "Open folder" is hidden on touch-only devices: on phones, tablets, and LAN browsers that button could only copy someone else's desktop path to the clipboard — it can't open a remote folder, so it did nothing useful. 2-in-1 devices with a mouse / trackpad still show it.
  • Actress-wall cards no longer jump bigger/smaller: previously, at the same window width, the size relationship between actress and video cards drifted (actress cards larger in narrow windows, smaller on wide screens, jumping when switching modes). This release aligns them in two tiers — on phones / tablets (narrow windows) the actress wall matches the video wall column-for-column and width-for-width, so switching modes looks identical; on wide desktop screens, video covers are cropped to landscape stills while actress photos are always portrait, so matching the width would make actress cards nearly twice as tall — the actress wall therefore uses its own column ladder (at 1920px full screen: 7 actress columns vs 5 video columns), aligning row height rather than card width. The result: at every window width the actress card is never bigger than a video card, and the relationship is fixed and predictable.
  • On-launch update check (desktop): after launch it quietly checks GitHub once for a new version, pushing a single bell notification only if there really is one (completely silent if you're already up to date); the Help page gets a "Check for updates on launch" toggle (on by default). Offline / timeout is silently skipped and never slows down launch.

⚡ Other improvements & fixes

  • A few dark-theme buttons are legible now: in the dark (dim) theme, outlined buttons like Help's "Check for updates", the folder picker, and prev/next in the search error state had borders and text almost the same color as the card and were invisible — this fixes the whole class at once.
  • The mouse wheel can switch to the previous / next title too: on Search detail, the lightbox (video / actress), and stills browsing — a horizontal wheel nudge = left/right arrow keys; in full-screen views like the lightbox and stills, the vertical wheel switches too.
  • Dragged-in files recognize IDs as fully as pasting: dragging used to recognize fewer formats than pasting, so the same file could fail on drag; both now go through the same rules, and it clearly tells you when an ID can't be recognized.
  • Plus a number of data-correctness, stability, and security fixes.
Assets 5
Loading