SSA HDRify

SSA HDRify 是一款桌面工具，能将 SSA/ASS 字幕的颜色转换为适合 HDR 播放的值，同时提供时间轴偏移、字体嵌入和批量重命名等辅助功能。 它是 gky99/ssaHdrify（Python 原版）的 Tauri 桌面重写版。

SSA HDRify is a desktop tool that converts SSA/ASS subtitle colors into values suitable for HDR playback, with companion tools for timing shift, font embedding, and batch renaming. It is a Tauri desktop rewrite of gky99/ssaHdrify (the original Python version).

浅色主题（中文）/ Light Theme (Chinese)

HDR 转换 / HDR Convert	时间轴偏移 / Time Shift

字体嵌入 / Font Embed	批量重命名 / Batch Rename

深色主题（英文）/ Dark Theme (English)

HDR 转换 / HDR Convert	时间轴偏移 / Time Shift

字体嵌入 / Font Embed	批量重命名 / Batch Rename

下载 | Download

Windows 用户可从 Releases 页面下载免安装的便携版 exe。建议优先使用最新稳定版；预览版也会保留在同一页面，用于测试尚未进入稳定版的功能。

Note

预览版可能包含尚未进入稳定版的修复或改进。

ssahdrify*.exe — 图形界面（GUI），适合手动操作
ssahdrify-cli*.exe — 命令行（CLI），适合自动化流水线、批处理和脚本化场景

macOS / Linux 用户请参考下方「从源码构建」。

Windows users can download portable, no-install exe files from Releases. Use the latest stable build by default; preview builds remain listed on the same page for testing features that have not entered a stable release yet.

Note

Preview builds may include fixes or improvements that have not entered the stable line yet.

ssahdrify*.exe — graphical interface (GUI), for manual use
ssahdrify-cli*.exe — command line (CLI), for automation pipelines, batch jobs, and scripts

macOS / Linux users, see "Build from Source" below.

功能 | Features

标签页 / Tab	功能 / Description
HDR 色彩转换 / HDR Color Conversion	将字幕颜色转换为适配 BT.2100 PQ 或 HLG 的 HDR 色彩值 / Convert subtitle colors into HDR values suitable for BT.2100 PQ or HLG
时间轴偏移 / Timing Shift	批量调整字幕时间戳；可从指定时间点之后开始偏移，并实时预览效果 / Batch-adjust subtitle timestamps; optionally start after a chosen timestamp, with live preview
字体嵌入 / Font Embedding	自动检测字幕引用的字体，在系统字体库或本地字体源中匹配，并把子集化后的字体嵌入 ASS 文件 / Detect fonts referenced by the subtitle, match them from system or local font sources, and embed subset fonts into the ASS file
批量重命名 / Batch Rename	自动匹配视频和字幕，并按视频文件名重命名字幕；当同一视频匹配到多个候选字幕时，可手动选择并调整配对，也可以启用多字幕模式，为每个视频保留多种语言的外挂字幕文件。 / Automatically match videos and subtitles, then rename subtitles after the video filename; when one video has multiple subtitle candidates, manually choose and adjust pairings, or enable multi-subtitle mode to keep multiple language sidecar subtitles for each video
样式编辑 / Style Edit	批量预览并修改 ASS/SSA `Style:` 行的字体族和字号。两项操作可独立启用，也可以只替换指定的原字体；行内 `\fn` / `\fs` 标签保持不变。 / Preview and batch-edit font family and size in ASS/SSA `Style:` rows. Enable either operation independently, optionally filter one source family, and leave inline `\fn` / `\fs` tags untouched.

Tip

完整支持中文路径 — 包含中文、日文或其他非 ASCII 字符的文件路径都可以正常处理。Tauri 和 Rust 底层使用 Unicode API，不受传统 ANSI 编码限制。

Non-ASCII paths are fully supported — File paths containing Chinese, Japanese, or other non-ASCII characters are handled correctly. Tauri and Rust use native Unicode APIs under the hood.

支持格式 | Supported Formats

不同功能的格式支持范围并不完全相同；下表是当前行为。HDR 色彩转换 会先将 .srt、.sub、.vtt 转换为 ASS 再进行处理。

Format support is not identical across workflows. The table below describes current behavior. HDR Color Conversion first converts .srt, .sub, and .vtt to ASS before processing.

功能 / Workflow	`.ass` / `.ssa`	`.srt`	`.sub`	`.vtt`	`.sup`
HDR 色彩转换 / HDR Color Conversion	原生处理 / native	转换为 ASS / convert to ASS	转换为 ASS / convert to ASS	基本文本 cue 转换为 ASS / basic text cues convert to ASS	不支持 / no
时间轴偏移 / Timing Shift	保持格式 / preserve format	保持格式 / preserve format	保持格式 / preserve format	重建基础 cue；不保留全部 WebVTT 元数据 / rebuilds basic cues; does not preserve all WebVTT metadata	不支持 / no
字体嵌入 / Font Embedding	支持 / yes	不支持 / no	不支持 / no	不支持 / no	不支持 / no
样式编辑 / Style Edit	支持 / yes	不支持 / no	不支持 / no	不支持 / no	不支持 / no
`diagnose-fonts`	支持 / yes	不支持 / no	不支持 / no	不支持 / no	不支持 / no
批量重命名 / Batch Rename	配对、复制或重命名 / pair, copy, or rename	配对、复制或重命名 / pair, copy, or rename	配对、复制或重命名 / pair, copy, or rename	配对、复制或重命名 / pair, copy, or rename	作为外挂字幕文件配对、复制或重命名（不解析内容） / opaque sidecar only
`chain`	取决于步骤 / depends on steps	取决于步骤 / depends on steps	取决于步骤 / depends on steps	仅支持本身接受 `.vtt` 的步骤 / only where the chosen step accepts `.vtt`	不支持 / no

Note

这里的 .sub 指 MicroDVD 文本字幕。Blu-ray PGS .sup 和 VobSub .sub/.idx 属于图像字幕，不适用于 HDR 文本颜色转换、时间轴偏移、ASS 字体嵌入或 diagnose-fonts。.sup 只会在「批量重命名」中作为外挂字幕文件进行配对、复制或重命名，不会解析其中的内容。如需把图像字幕变成文本字幕，请先使用专门的字幕转换/OCR 工具。

Here .sub means MicroDVD text subtitles. Blu-ray PGS .sup and VobSub .sub/.idx are image subtitle formats, so HDR text color conversion, Timing Shift, ASS font embedding, and diagnose-fonts do not apply. .sup is only paired, copied, or renamed as an opaque sidecar in Batch Rename. To turn image subtitles into text subtitles, use a dedicated subtitle conversion/OCR tool first.

使用方法 | Usage

HDR 色彩转换 / HDR Color Conversion

选择 EOTF 曲线（PQ 或 HLG）/ Select EOTF curve (PQ or HLG)
设置字幕目标亮度（默认 203 nits）/ Set target subtitle brightness (default: 203 nits)
选择字幕文件（支持多选）/ Select subtitle files (multi-select supported)
点击「转换」/ Click Convert
默认输出扩展名为 .hdr.ass（可修改）/ Default output extension is .hdr.ass (customizable)

参数说明 | Parameter Guide

参数 / Parameter 默认值 / Default 说明 / Description

EOTF curve PQ PQ (ST 2084) 用于 HDR10/杜比视界；HLG 用于广播 HDR / PQ for HDR10/Dolby Vision; HLG for broadcast HDR

Target brightness 203 nits SDR 字幕峰值亮度（BT.2408 标准值）。如果字幕太亮就调低，太暗就调高 / Peak SDR subtitle brightness (BT.2408 reference value). Lower it if too bright; raise it if too dim

时间轴偏移 / Timing Shift

选择字幕文件 / Select a subtitle file
输入偏移量（毫秒），并选择「提前」或「延后」/ Enter offset amount (ms), then choose Faster or Slower
可选：只偏移指定时间点之后的字幕行 / Optionally shift only lines after a specific timestamp
实时预览调整结果 / Preview the result in real time
导出 / Export

字体嵌入 / Font Embedding

点击「选择字幕文件 / Select Subtitle File」，选择一个或多个 ASS/SSA 字幕文件 / Click Select Subtitle File to pick one or more ASS/SSA files
工具会自动检测字幕引用的字体，并优先尝试从系统字体库匹配 / The tool detects fonts referenced by the subtitle and first tries to match them against the system font library
主面板会实时显示本地字体源覆盖情况（覆盖 N / M）和尚未匹配的字体；每个字体都会标注来源（本地 / 系统）和状态（已找到 / 缺失）/ The main panel shows live local-source coverage (Coverage: N / M) and lists any still-missing families; each detected font is tagged with its source (Local / System) and status (Found / Missing)
选择输出位置：默认保存到源字幕旁，也可以保存到指定文件夹；指定文件夹模式会将输出平铺到该文件夹，重复输出名会自动跳过 / Choose the output location: save beside each source subtitle by default, or save into a chosen folder; chosen-folder mode writes flat outputs into that folder and skips duplicate output names
点击「嵌入已选字体」，将子集化后的字体数据写入 .embedded.ass 输出文件 / Click Embed Selected Fonts to write the subset font data into .embedded.ass output files

字幕组排版常用字体通常没有安装在系统中。打开「字体来源 / Font Sources」面板：选择「添加文件夹 / Add Folder」只扫描所选文件夹这一层；选择「添加字体库 / Add Font Library」会扫描所选根目录及其所有子文件夹。这些字体无需系统安装，也可以参与匹配。

Fonts commonly used in fan-sub typesetting are often not installed system-wide. Open Font Sources: Add Folder scans only the selected folder, while Add Font Library scans the selected root and all of its subfolders. These fonts can be matched without installing them into the OS.

支持大型字体文件夹和多层字体库；发现文件与解析字体时都会显示进度，也可以随时取消。扫描包含约 5000 个字体文件或内容总量约 5 GiB 以上的来源前，程序会先弹出确认对话框。扫描和缓存写入都有安全上限；超大、扫描中变化、无法完整读取或被取消的来源可能提前停止，或仅用于本次会话而不写入持久化缓存，并会在界面/日志中提示。递归扫描不会跟随符号链接、junction（目录联接）或其他 reparse point（重解析点）。

Large font folders and nested libraries are supported; discovery and parsing show progress and can be cancelled. Before scanning a source with about 5000 font files or about 5 GiB of content, the app asks for confirmation. Scanning and cache writes are bounded by safety ceilings; unusually large, changing, unreadable, or cancelled sources may stop early or remain available only for the current session instead of being written to the persistent cache, with a visible UI/log message. Recursive scans do not follow symbolic links, junctions, or other reparse points.

字体名称匹配 / Font Name Matching

工具会读取字体文件的 OpenType name 表，并索引受支持的本地化 family、typographic family、full-face 和 PostScript 名称变体（英文、中文等），同时内置了异常字体防护上限。ASS 脚本无论引用哪个受支持名称，都能匹配到同一个字体文件；@家族名 这类竖排前缀也会按同一字体处理。

The tool reads each font's OpenType name table and indexes supported localized family, typographic-family, full-face, and PostScript name variants (English, Chinese, etc.), with built-in safety caps against abnormal fonts. An ASS script referencing any supported name resolves to the same font file; the ASS @FamilyName vertical-writing prefix is treated as the same font.

对 .ttc / .otc 字体集合，工具会按匹配到的 face index 抽出对应字形并嵌入为单 face 子集。ASS [Fonts] 里的 fontname: 是生成的嵌入项标签，例如 dream_han_serif_sc_w22.ttf，不代表源文件必须是 .ttf；实际匹配依赖子集字体内部保留的 name 表。

For .ttc / .otc collections, the matched face index is extracted and embedded as a single-face subset. The ASS [Fonts] fontname: line is a generated attachment label, such as dream_han_serif_sc_w22.ttf; it does not mean the source file had to be .ttf. Matching relies on the preserved internal name table in the subset font.

样式编辑 / Style Edit

选择一个或多个 ASS/SSA 文件；应用会读取每个文件的 [V4+ Styles] 或 [V4 Styles] 表，并在写入前列出全部 Style: 行 / Select one or more ASS/SSA files; the app reads each [V4+ Styles] or [V4 Styles] table and lists every Style: row before writing
独立启用「更改字体族」和/或「更改字号」。字体族操作还可以限定为只替换一个指定的原字体 / Independently enable Change font family and/or Change font size. The family operation can optionally replace only one specified source family
在预览表中逐行确认旧值和新值；默认勾选所有会发生变化的行，也可以取消任意行 / Review old and new values row by row; every effective change is checked by default, and any row can be cleared
点击「写入」。输出固定为源文件旁的 .styled.ass / .styled.ssa 新文件；如果目标已存在，本次不会覆盖它 / Click Write. Output is always a new sibling .styled.ass / .styled.ssa file; an existing target is never overwritten

编辑器只修改样式表中的 Fontname 和 Fontsize 字段，并按文件自己的 Format: 列顺序解析。它不会全局替换对话或注释文字，也不会更改行内 \fn / \fs 覆盖标签。输出保留源文件的受支持文本编码、BOM（字节顺序标记）和换行符；如果源文件在预览后发生变化，安全写入会拒绝继续执行这个已过期的写入计划。

The editor changes only Fontname and Fontsize fields in the style table and follows each file's own Format: column order. It does not globally replace dialogue or comment text, and it leaves inline \fn / \fs override tags untouched. Output preserves the source's supported text encoding, byte-order mark (BOM), and line endings; if a source changes after preview, the safe writer rejects the stale plan.

Note

样式编辑目前仅在 GUI 中提供，不属于 CLI 或 chain。每个源文件上限为 50 MiB；单次选择上限为 500 个文件、200 MiB 源文件字节、内存中保留的 200 MiB 解码文本和 2,000 个 Style 行。超过任一上限时会拒绝整个选择，以保证写入前的预览完整。

Style Edit is currently GUI-only and is not part of the CLI or chain. Each source file is capped at 50 MiB; one selection is capped at 500 files, 200 MiB of source bytes, 200 MiB of decoded text retained in memory, and 2,000 Style rows. Exceeding any limit rejects the whole selection so the pre-write preview remains complete.

批量重命名 / Batch Rename

拖入包含视频和字幕的文件夹（或点击「选择文件 / Select Files」手动选择）；程序会按文件格式自动归类 / Drop a folder containing both videos and subtitles (or click Select Files to pick manually); the app categorizes files by format automatically
应用会按字幕组常见命名方式提取剧集号，并预填配对表 / The app extracts episode numbers from common fan-sub naming patterns and pre-fills the pairing table
如果出现错配或漏配，可直接在对应行的下拉框中手动选择字幕；选中后该行会自动加入重命名队列 / If a row is paired incorrectly or missed, choose a subtitle from that row's dropdown; once selected, the row is automatically added to the rename queue
如果要为同一视频保留多个语言字幕，启用「保留每个视频的多个字幕」；带语言标签的字幕会写成 Video.sc.ass、Video.tc.ass 这类文件名 / To keep multiple language subtitles for the same video, enable Keep multiple subtitles per video; tagged subtitles are written as names such as Video.sc.ass and Video.tc.ass
选择输出策略：原文件直接改名 / 复制到视频所在目录 / 复制到自定义目录 / Pick the output strategy: rename in place, copy to the video's directory, or copy to a custom directory
点击「运行」；如果目标路径已存在按同一规则生成的同名文件，程序会先弹出覆盖确认对话框 / Click Run; if a file with the generated name already exists at the target path, an overwrite confirmation appears first

配对算法 | Pairing Algorithm

处理流程：清理括号内容 → 按优先级尝试剧集号正则（S\d+E\d+、][NN][、- NN、第N话、EP\d+）→ 分季并行扫描 → (season, episode) 配对键 → LCS 回退 → 最后由手动选择兜底。规则已在多组真实字幕组命名样本上验证过，包括中日双语标题、外挂多语字幕、季度后缀变体等。

Pipeline: bracket cleanup → priority-ordered episode regex (S\d+E\d+, ][NN][, - NN, 第N话, EP\d+) → parallel season-aware scan → (season, episode) pairing key → LCS fallback → manual selection as the final safety net. Pattern coverage has been validated against multiple real-world fan-sub naming samples, including bilingual CJK titles, external multi-language subtitles, and season-suffix variants.

CLI 使用 | CLI Usage

ssahdrify-cli 与 GUI 从同一份源代码构建，并提供 HDR 转换、时间轴偏移、字体嵌入和批量重命名。新增的样式编辑器目前仅在 GUI 中提供。CLI 另外提供 chain（一次调用串联多个步骤，只有最后一步写入文件）、refresh-fonts（构建或刷新 CLI 字体缓存）和 diagnose-fonts（只诊断字体解析，不写字幕）等子命令。

ssahdrify-cli is built from the same source as the GUI and provides HDR Convert, Time Shift, Font Embed, and Batch Rename. The new Style Edit workbench is currently GUI-only. The CLI also provides chain (run multiple steps with only the last step writing), refresh-fonts (build or refresh the CLI font cache), and diagnose-fonts (read-only font diagnostics).

快速示例 | Quick Examples

下面示例中的 <font-folder>、<series-folder> 等都是占位符；请替换成你自己电脑上的实际路径。

Placeholders such as <font-folder> and <series-folder> stand for real paths on your own computer; replace them before running the commands.

# HDR 色彩转换（PQ 曲线）/ HDR conversion (PQ curve)
ssahdrify-cli hdr --eotf pq input.ass

# 时间轴偏移 +500ms / Timing shift +500ms
ssahdrify-cli shift --offset +500ms input.ass

# 时间轴映射：按多个时间段应用不同偏移 / Timing map: apply different offsets by segment
ssahdrify-cli shift --map timing-map.json input.srt

# 字体嵌入：从指定文件夹查找字体 / Font embed: search a folder for fonts
ssahdrify-cli embed --font-dir "<font-folder>" input.ass

# 字体嵌入：递归搜索整个字体库树 / Font embed: recursively search a font-library tree
ssahdrify-cli embed --recursive-font-dir "<font-library-root>" input.ass

# 字体解析诊断：不写输出文件 / Font diagnostics: no output subtitle writes
ssahdrify-cli diagnose-fonts --font-dir "<font-folder>" input.ass

# 持久化字体缓存：先扫描一次，后续 embed 复用 / Persistent font cache: scan once, reuse later
ssahdrify-cli refresh-fonts --recursive-font-dir "<font-library-root>"
ssahdrify-cli embed input.ass            # 自动使用缓存 / uses cache automatically

# 链式调用：一次完成 HDR 转换和时间轴偏移，只有最后一步写文件 / Chain: HDR + shift in one command, only the final step writes
ssahdrify-cli chain hdr --eotf pq + shift --offset +500ms input.ass

# 批量重命名：默认复制到视频所在目录 / Batch rename (default: copy sub next to video)
ssahdrify-cli rename "<series-folder>"

# 批量重命名：复制到自定义目录 / Batch rename: copy to a chosen directory
ssahdrify-cli rename --mode copy-to-chosen --output-dir "<output-folder>" "<series-folder>"

# 多个外挂字幕文件：保留语言后缀，避免 sc/tc 同扩展字幕互相覆盖 / Multiple sidecar subtitles: keep language suffixes
ssahdrify-cli rename "<series-folder>" --langs all --dry-run

rename --mode 控制文件操作。默认的 copy-to-video 会把字幕复制到匹配视频所在的目录；rename 会在字幕原目录中直接重命名源文件；copy-to-chosen 会把字幕复制到 --output-dir 指定的目录。copy-to-chosen 必须同时传入 --output-dir，另外两种模式则不接受该参数。

rename --mode controls the file operation. The default copy-to-video copies each subtitle beside its matched video; rename renames the source subtitle in its existing directory; and copy-to-chosen copies it to the directory supplied through --output-dir. --output-dir is required with copy-to-chosen and is rejected with the other two modes.

rename --langs auto 保持和 GUI 一致的默认行为：每个视频只选一个字幕，输出文件名精确匹配视频 stem（如 Video.ass）。rename --langs all 或显式列表（如 --langs sc,jp）可以为同一个视频规划多个字幕，并写成带语言后缀的文件名（如 Video.sc.ass、Video.jp.srt）；没有语言标记的字幕仍使用精确视频名（如 Video.ass）。如果多行会写入同一个目标路径，CLI 会在写入前拦截这些存在冲突的条目。

rename --langs auto keeps the same default behavior as the GUI: one subtitle per video, with the output filename matching the video stem exactly (Video.ass). rename --langs all or an explicit list such as --langs sc,jp can plan multiple subtitles for the same video and write them with language suffixes such as Video.sc.ass and Video.jp.srt; untagged subtitles still use the exact video name (Video.ass). If multiple rows would write to the same target path, the CLI blocks those conflicting rows before writing.

shift --map <FILE> 使用一个只读时间轴映射文件，不运行 Sushi、alass、FFmpeg 或任何音频自动同步工具。JSON 格式可以写成 {"rules":[{"start":"00:00:00.000","end":"00:05:00.000","offset":"+1s","label":"opening"},{"startMs":5000,"offsetMs":-500}]}。也可以使用简单 CSV 行：start,end,offset,label,enabled，例如 00:00:00.000,00:05:00.000,+1s,opening,true。start 为闭区间，end 为开区间；规则出现重叠时，按文件顺序以最先匹配的为准。映射文件会在批处理开始前先解析和校验，失败时不会写任何字幕。

shift --map <FILE> uses a read-only timing-map file; it does not run Sushi, alass, FFmpeg, or any audio auto-sync helper. JSON can look like {"rules":[{"start":"00:00:00.000","end":"00:05:00.000","offset":"+1s","label":"opening"},{"startMs":5000,"offsetMs":-500}]}. A simple CSV shape is also accepted: start,end,offset,label,enabled, for example 00:00:00.000,00:05:00.000,+1s,opening,true. start is inclusive, end is exclusive, and overlapping rules use the first matching row in file order. The map is parsed and validated before batch processing starts; if validation fails, no subtitles are written.

全部子命令 | All Subcommands

每个子命令都可通过 --help 查看完整参数。

Each subcommand can show the full parameter reference with --help.

ssahdrify-cli --help
ssahdrify-cli hdr            --help
ssahdrify-cli shift          --help
ssahdrify-cli embed          --help
ssahdrify-cli rename         --help
ssahdrify-cli diagnose-fonts --help
ssahdrify-cli refresh-fonts  --help
ssahdrify-cli chain          --help

全局选项 | Global Options

选项 / Option	说明 / Description
`--lang <en\|zh>`	输出语言；不指定时按系统区域设置自动检测（zh* → zh，否则 en）/ Output language; auto-detected from OS locale when omitted (zh* → zh, otherwise en)
`--json`	为支持的子命令输出机器可读 JSON 报告；详见下方 JSON 模式 / Emit machine-readable JSON for supported subcommands; see JSON Mode below
`--verbose`	显示更详细的进度 / Show more detailed progress
`--quiet`	隐藏常规进度输出 / Suppress normal progress output
`--dry-run`	预览计划执行的操作，不写入文件 / Preview planned work without writing files
`--overwrite`	允许覆盖已存在的输出文件 / Replace existing output files instead of skipping
`--output-dir <DIR>`	将输出重定向到指定目录 / Redirect output to a specific directory
`--no-cache`	跳过本次运行的字体缓存；缓存文件本身保持不变 / Skip the font cache for this run; leave the cache file untouched
`--cache-file <PATH>`	使用指定缓存文件路径，覆盖默认路径 / Use a specific cache file path instead of the OS default (see Cache Location below)
`--fail-fast`	任一文件失败即停止处理后续输入；已成功写出的文件会保留，失败输入的目标位置可能残留未完整写入的文件 / Abort the batch on the first failed file; previously-succeeded outputs are kept, but the failed input may leave a partial-write artifact at its destination

JSON 模式 | JSON Mode

--json 目前适用于 hdr / shift / embed / rename 和 diagnose-fonts。常规子命令会输出固定 schema 的报告，按文件列出 status (written / planned / skipped / failed)、output path、encoding、warnings 等字段；stderr 仍可输出供人阅读的诊断信息。diagnose-fonts --json 直接输出诊断报告。chain v1 会明确提示不支持 JSON，并改用纯文本报告；refresh-fonts 使用 stderr 输出状态。

--json currently applies to hdr / shift / embed / rename and diagnose-fonts. Normal subcommands emit a fixed-schema report listing per-file status (written / planned / skipped / failed), output path, encoding, warnings, and related fields; stderr can still print human-readable diagnostics. diagnose-fonts --json emits the diagnostic report directly. chain v1 explicitly reports that JSON output is not supported and uses a plain-text report instead; refresh-fonts reports status on stderr.

启用 --diagnose 时，JSON 会额外包含完整 diagnostics 对象，即使人类输出模式是默认的 summary。未启用 --diagnose 时，常规 JSON schema 保持不变。

When --diagnose is enabled, JSON additionally includes the full diagnostics object even when human output uses the default summary mode. Without --diagnose, the normal JSON schema is unchanged.

终端字符串插值安全注意事项 | Terminal string-interpolation safety note

--json 按 RFC 8259 输出；BiDi 控制符（U+200E/U+202E 等）、零宽字符，以及 U+2028/U+2029 行分隔符在 JSON 字符串中都是合法字符，因此不会额外转义。如果用 jq -r 将 .input / .output 等字段还原后再插入到终端（例如 echo、提示符或其他 CLI 参数），恶意构造的文件名可能影响终端显示。下游脚本应在终端输出边界自行过滤（如 jq 的 gsub 或 shell 包装工具）。CLI 自身供人阅读的输出（未启用 --json）已在所有打印点调用 sanitize_for_display，不受此问题影响。

--json output follows RFC 8259, but BiDi format characters (U+200E/U+202E etc.), zero-width characters, and the U+2028/U+2029 line separators are valid in JSON strings and are not additionally escaped. If you use jq -r to restore fields such as .input / .output and then insert or interpolate those values into a terminal (echo, prompts, or another CLI's arguments), crafted filenames may affect terminal display. Downstream scripts should sanitize at the terminal output boundary (for example with jq's gsub or a wrapping shell tool). The CLI's own human-readable output (without --json) already passes every print site through sanitize_for_display and is not affected.

诊断输出 | Diagnostics

hdr / shift / embed / rename 支持 --diagnose[=summary|full]。--diagnose 与 --diagnose=summary 等价，会在命令完成后附加紧凑诊断；--diagnose=full 会列出逐文件细节，embed 还会列出字体解析层级（本次传入的字体源、持久化缓存、系统字体）和缓存状态。chain 与 refresh-fonts 不支持 --diagnose，传入会报错而不是静默忽略。

hdr / shift / embed / rename support --diagnose[=summary|full]. --diagnose and --diagnose=summary are equivalent and attach compact diagnostics after the command finishes; --diagnose=full lists per-file details, and embed also lists font-resolution tiers (font sources passed for this run, persistent cache, system fonts) plus cache status. chain and refresh-fonts do not support --diagnose; passing it returns an error instead of being silently ignored.

diagnose-fonts 是独立的详细诊断命令，默认输出 verbose 报告，而且只读：不写输出字幕、不刷新或修改字体缓存。它接受字幕输入和字体解析选项：--font-dir、--recursive-font-dir、--font-file、--no-system-fonts、--no-cache、--cache-file、--lang、--json。需要确认字体文件是否真的能被子集化时，可显式加入 --subset-check；该检查只在内存中运行，不写出字幕。

diagnose-fonts is the standalone detailed diagnostic command. It is verbose by default and read-only: it does not write output subtitles and does not refresh or modify the font cache. It accepts subtitle inputs plus font-resolution options: --font-dir, --recursive-font-dir, --font-file, --no-system-fonts, --no-cache, --cache-file, --lang, and --json. Add --subset-check explicitly when you need to confirm whether resolved font files can actually be subset; the check runs in memory and does not write subtitles.

diagnose-fonts 和带 --diagnose 的 embed 还会输出 package-level Font QA 状态：complete、incomplete 或 blocked。complete 表示已检查文件里的字体引用全部解析成功；incomplete 表示有缺失字体、警告或被跳过的可选子集化检查；blocked 表示文件诊断失败、字体解析错误，或 --subset-check 明确失败。

diagnose-fonts and embed --diagnose also report a package-level Font QA status: complete, incomplete, or blocked. complete means all font references in the checked files resolved successfully; incomplete means missing fonts, warnings, or skipped optional subset checks remain; blocked means a file diagnostic failed, a font resolution errored, or --subset-check explicitly failed.

为避免诊断命令在异常字幕/字体组合上长时间运行，--subset-check 有命令级预算：最多 128 次子集化调用，累计子集输出约 100 MiB；超出后剩余检查会标记为 skipped，并输出一次 budget exhausted 警告。

To prevent diagnostics from running too long on unusual subtitle/font combinations, --subset-check uses one command-level budget: up to 128 subset calls and about 100 MiB of cumulative subset output. After that, remaining checks are marked skipped and one budget-exhausted warning is emitted.

# 附加紧凑诊断 / Attach compact diagnostics
ssahdrify-cli embed --diagnose input.ass

# 附加完整诊断 / Attach full diagnostics
ssahdrify-cli embed --diagnose=full --font-dir "<font-folder>" input.ass

# 只诊断字体解析，不写字幕 / Diagnose font resolution only, no subtitle writes
ssahdrify-cli diagnose-fonts --font-dir "<font-folder>" input.ass

# 额外检查已解析字体能否子集化 / Also check whether resolved fonts can be subset
ssahdrify-cli diagnose-fonts --subset-check --font-dir "<font-folder>" input.ass

# 下游打包必须完整嵌入字体时推荐 / Recommended when downstream packaging requires every font
ssahdrify-cli embed --font-dir "<font-folder>" --on-missing fail --fail-fast --diagnose input.ass

embed 默认仍使用 --on-missing warn：能嵌入的字体会继续嵌入，缺失或子集化失败的字体会变成 warning。此时输出文件可能已经写出，但 summary 会明确显示 written with warnings / incomplete，避免把部分成功误读成「全部字体都成功」。

embed still defaults to --on-missing warn: fonts that can be embedded continue to be embedded, while missing fonts or subset failures become warnings. In that case the output file may already be written, but the summary explicitly says written with warnings / incomplete so partial success is not mistaken for “all fonts succeeded.”

字体缓存 | Font Cache

embed 每次启动通常都要扫描每个 --font-dir（浅层）或 --recursive-font-dir（递归）下的字体文件，构建查找表（一般几秒到几十秒；5000+ 字体可能需要几分钟）。持久化字体缓存能将这一过程变为一次性操作：先运行 refresh-fonts，将字体元数据写入磁盘上的 SQLite 文件；之后 embed 会在缓存仍有效时直接复用它，跳过扫描。对于字幕组按集批量处理尤其有用。

The embed subcommand normally scans the font files under each shallow --font-dir or recursive --recursive-font-dir every time it starts to build its lookup table (usually seconds to tens of seconds; minutes for 5000+ font collections). The persistent font cache turns this into a one-time step: run refresh-fonts to write font metadata into a SQLite file on disk, then later embed calls reuse it while the cache is still valid. This is especially useful for fan-sub teams processing episodes in batches.

工作流 | Workflow

# 一次性扫描字体目录，构建缓存 / Scan once to build the cache
ssahdrify-cli refresh-fonts --font-dir "<anime-font-folder>" --font-dir "<latin-font-folder>"

# 也可以把一棵字体目录树作为一个来源 / Or track one whole directory tree as one source
ssahdrify-cli refresh-fonts --recursive-font-dir "<font-library-root>"

# 后续 embed 自动复用缓存（不再扫描） / Subsequent embed uses cache (no scan)
ssahdrify-cli embed input.ass

# 也可以继续加 --font-dir，临时合并额外字体源（缓存 + 额外目录） /
# You can still pass --font-dir to merge extra dirs with the cache
ssahdrify-cli embed --font-dir "<project-font-folder>" input.ass

# 本次强制不用缓存 / Force no-cache for one run
ssahdrify-cli --no-cache embed --font-dir "<font-folder>" input.ass

# 字体目录变更后刷新缓存 / Refresh cache after fonts change
ssahdrify-cli refresh-fonts --font-dir "<anime-font-folder>" --font-dir "<latin-font-folder>"

缓存位置 | Cache Location

默认位置按操作系统决定（与 GUI 缓存独立，避免锁竞争）：

Windows: %APPDATA%/ssahdrify/cli_font_cache.sqlite3
macOS: $HOME/Library/Application Support/ssahdrify/cli_font_cache.sqlite3
Linux: ${XDG_DATA_HOME:-$HOME/.local/share}/ssahdrify/cli_font_cache.sqlite3

--cache-file <PATH> 可改用指定路径。

Default locations are OS-specific and separate from the GUI cache to avoid lock contention:

Windows: %APPDATA%/ssahdrify/cli_font_cache.sqlite3
macOS: $HOME/Library/Application Support/ssahdrify/cli_font_cache.sqlite3
Linux: ${XDG_DATA_HOME:-$HOME/.local/share}/ssahdrify/cli_font_cache.sqlite3

Use --cache-file <PATH> to choose a different path.

漂移检测 | Drift Detection

embed 启动时会校验每个缓存来源的目录清单和候选字体文件元数据，包括递归来源的子文件夹。如果发现漂移（说明你添加 / 删除 / 替换 / 重命名了字体文件或文件夹），CLI 会在 stderr 列出发生变化的来源及其扫描范围，本次运行自动退回无缓存模式（使用显式字体来源或系统字体），并提示你运行 refresh-fonts 更新。缓存不会被静默重建——缓存写入必须由 refresh-fonts 显式触发。

At startup, embed validates each cached source's directory inventory and candidate-font metadata, including nested folders for recursive sources. If drift is detected (meaning you added, deleted, replaced, or renamed font files or folders), the CLI lists the changed source and its scan scope on stderr, automatically falls back to no-cache mode for this run (using explicit font sources or system fonts), and tells you to run refresh-fonts. The cache is never silently rebuilt — cache writes are always triggered explicitly by refresh-fonts.

限制 | Limitations

--font-dir 和 GUI 的「添加文件夹 / Add Folder」始终只扫描一层；--recursive-font-dir 和「添加字体库 / Add Font Library」才会递归扫描。两种范围可以同时用于同一个根目录，并作为独立来源管理。
字体缓存最多记录 256 个来源。单个递归扫描还受 4096 个真实目录、64 层深度、50,000 个候选字体文件、128 GiB 候选文件总量和 200,000 个目录项等安全上限约束。
单个缓存来源最多安全写入 32,768 个 font faces。超过上限、用户取消、遇到读取不完整或扫描期间来源变化时，结果不会作为完整缓存发布；GUI 已成功读取的部分字体可以仅在当前会话继续使用，并会明确提示。
单个字体文件的扫描/子集化读取上限为 64 MiB；超过会被拒绝并报告错误。
递归扫描不会跟随符号链接、Windows junction 或其他 reparse point（重解析点），从而避免越过所选字体库根目录或形成目录循环。
GUI 和 CLI 各自使用独立缓存文件，避免 SQLite 锁竞争；同一个可执行文件同一时间只会打开一个缓存文件（默认路径或 --cache-file 覆盖路径）。chain v1 暂不读取缓存（其中的 embed 步始终使用显式 --font-dir、--recursive-font-dir 或系统字体）。
跨版本不会自动迁移缓存结构。递归来源支持目前使用 cache schema v6；旧缓存会进入明确的一次性重建流程。CLI 会提示删除旧缓存文件并重新运行 refresh-fonts，GUI 会在确认后重建，不会静默改写。
--font-dir and GUI Add Folder always scan one level; only --recursive-font-dir and Add Font Library recurse. The same root can be tracked independently in both scopes.
The font cache tracks at most 256 sources. A recursive scan is also bounded to 4096 real directories, 64 levels, 50,000 candidate font files, 128 GiB of candidate-file data, and 200,000 directory entries.
A single cached source can safely persist 32,768 font faces. A ceiling hit, user cancellation, incomplete read, or source mutation prevents publication as a complete cache; successfully parsed GUI results may remain available for the current session with an explicit notice.
A single font file is capped at 64 MiB for scanning/subsetting; larger files are refused with an error.
Recursive scans do not follow symbolic links, Windows junctions, or other reparse points, which prevents traversal outside the chosen library root and directory cycles.
GUI and CLI use separate cache files to avoid SQLite lock contention; a single binary opens exactly one cache at a time (default path or --cache-file override). chain v1 does not consult the cache; its embed step always uses explicit --font-dir, --recursive-font-dir, or system fonts.
Cache schemas are not migrated silently. Recursive source support uses cache schema v6, so an older cache goes through a visible one-time rebuild flow: the CLI asks you to remove it and rerun refresh-fonts, while the GUI rebuilds only after confirmation.

使用场景 | Background

SSA/ASS 字幕自身不带色彩空间元数据，渲染器通常会按 SDR 处理，结果是字幕在 HDR 画面里显得过饱和、过亮。播放 HDR 视频时，显示设备会进入 HDR 模式，但字幕仍按 SDR 混合，色差就来自这里。

SSA/ASS subtitles do not carry color-space metadata, so renderers usually treat them as SDR content. The result is that subtitles can look oversaturated and overly bright in HDR video. When an HDR video plays, the display enters HDR mode, but subtitles are still blended as SDR; that is where the color mismatch comes from.

如果你的播放器已经能正确处理字幕亮度（例如 mpv 的 blend-subtitles=video，或 madVR 配合 xy-SubFilter 的字幕色彩管理），则不需要本工具。

If your player already handles subtitle brightness correctly (e.g. mpv with blend-subtitles=video, or madVR with xy-SubFilter color management), you don't need this tool.

相关讨论 / Related discussion: libass/libass#297

相关工具 / Related tool: arition/SubRenamer 也是视频与字幕重命名工作流的一个选择（按字母序 + 下标配对）。本项目的批量重命名功能（Tab 4）则使用面向字幕组命名习惯的正则配对流程，代码独立实现。

For video-and-subtitle rename workflows, arition/SubRenamer is another option (alphabetical + index pairing). This project's Batch Rename feature (Tab 4) uses an independently implemented regex pairing flow built around common fan-sub naming patterns.

HDR 转换原理 | How HDR Conversion Works

SSA/ASS 字幕颜色 (sRGB)
├─ 1. sRGB → rec2100-linear（Color.js 色彩空间转换）
├─ 2. 亮度缩放：Y × (targetBrightness / 203)
├─ 3. rec2100-linear → rec2100pq 或 rec2100hlg
└─ 4. 输出 RGB

SSA/ASS subtitle colors (sRGB)
├─ 1. sRGB → rec2100-linear (Color.js color space conversion)
├─ 2. Luminance scaling: Y × (targetBrightness / 203)
├─ 3. rec2100-linear → rec2100pq or rec2100hlg
└─ 4. Output RGB

精度说明 | Accuracy Note

PQ 模式已验证与 Python 原版（colour-science）逐像素一致。HLG 模式使用手动实现的 BT.2100 逆 OOTF + OETF（绕过 Color.js 的 rec2100hlg 空间），同样与 Python 原版完全一致。

PQ mode is verified pixel-exact against the Python version (colour-science). HLG mode uses a manually implemented BT.2100 inverse OOTF + OETF (bypassing Color.js's rec2100hlg space) and also matches the Python version exactly.

由于字幕混合链路和 HDR 显示环境很复杂（HDMI 元数据协商、显示器色调映射等），实际效果主要保证「红还是红、蓝还是蓝」的基础观感，不适合严格校色场景。

Due to the complexity of subtitle blending pipelines and HDR display environments (HDMI metadata negotiation, display tone mapping, etc.), the result is mainly intended to preserve the basic impression that “red stays red and blue stays blue.” It is not meant for strict color-accuracy or color-grading workflows.

从源码构建 | Build from Source

前置条件 | Prerequisites

Node.js (v22.13 minimum, v24 LTS recommended, or v26 Current)
npm 11.19.0（由 packageManager 声明并由 CI 强制 / declared by packageManager and enforced in CI）
Rust 工具链 / Rust toolchain（最低 1.91；rustup 会自动安装仓库锁定且经过测试的稳定工具链 / 1.91 minimum; rustup automatically installs the repository-pinned tested stable toolchain）
Windows: WebView2 (Windows 10/11 已预装 / pre-installed on Windows 10/11)
macOS / Linux: 参考 / see Tauri prerequisites

开发 | Development

cd ssaHdrify-tauri
npm ci
npm run tauri dev

构建 | Production Build

# GUI portable executable
npm run tauri build

# CLI portable executable
npm run build:cli

# Build both GUI and CLI
npm run build:all

npm run build:all 会依次生成 GUI 和 CLI 两个原生发布版可执行文件；GUI 阶段由 Tauri 的 beforeBuildCommand 自动运行 TypeScript 7 类型检查和 Vite 前端构建。

npm run build:all produces both native release executables in sequence. During the GUI stage, Tauri's beforeBuildCommand automatically runs the TypeScript 7 type-check and Vite frontend build.

在 Windows 上，便携式 exe 会生成到 src-tauri/target/release/，可直接运行，无需安装。tauri.conf.json 目前设置了 bundle.active: false，因此默认生成便携式二进制文件，而不是安装包。

On Windows, portable executables are produced under src-tauri/target/release/ and can be run directly with no installation required. tauri.conf.json currently sets bundle.active: false, so the default output is portable binaries rather than installers.

Expected Windows release build outputs:

src-tauri/target/release/ssahdrify.exe
src-tauri/target/release/ssahdrify-cli.exe

测试 | Testing

npm run test:run                                  # 前端单元测试 / Frontend unit tests
npm run typecheck:all                             # TypeScript 7 release gate + TypeScript 6 API compatibility
cargo test --manifest-path src-tauri/Cargo.toml   # Rust 后端测试 / Rust backend tests

npm test 默认进入 watch 模式（开发用）；npm run test:run 是单次运行。

npm test defaults to watch mode (development); use npm run test:run for a single-pass run.

发布构建使用原生 TypeScript 7 类型检查；ESLint 和 typescript-eslint 通过官方 @typescript/typescript6 兼容包继续使用 TypeScript 6 的程序化 API。CI 会检查两个工具链。请使用上面的 npm run typecheck:* 脚本；不要依赖裸 tsc / npx tsc 的解析顺序。

Release builds use native TypeScript 7 for type-checking. ESLint and typescript-eslint continue to use TypeScript 6's programmatic API through the official @typescript/typescript6 compatibility package, and CI checks both toolchains. Use the npm run typecheck:* scripts above instead of relying on bare tsc / npx tsc resolution order.

Dependency Watch 是只读的每周检查：它监视直接 npm 和 Cargo 依赖、TypeScript 别名工具链，以及锁定到完整提交哈希的 GitHub Actions。发现普通更新时，它只会将工作流运行标红并写入汇总；不会创建分支或拉取请求。依赖安全漏洞由 GitHub Dependabot Alerts 单独监视。

Dependency Watch is a read-only weekly check for direct npm and Cargo dependencies, the aliased TypeScript toolchain, and GitHub Actions pinned to full commit hashes. When an ordinary update exists, it only marks the workflow run red and writes a summary; it cannot create a branch or pull request. GitHub Dependabot Alerts separately monitor dependency security vulnerabilities.

架构 | Architecture

┌──────────────────────────────────────────────────────────────────────────────┐
│  Shared TypeScript modules                                                   │
│  - GUI: HDR, Shift, Embed, Rename, and Style Edit                            │
│  - CLI: HDR, Shift, Embed, and Rename (Style Edit is GUI-only)               │
│  - Color.js (PQ/HLG color math), ass-compiler (font collection)              │
│  - Custom subtitle parser, fan-sub regex pairing engine                      │
└──────────────┬───────────────────────────────────────┬───────────────────────┘
               │                                       │
   imported as React modules              bundled via esbuild (IIFE)
               │                                       │
┌──────────────┴───────────────┐ ┌─────────────────────┴───────────────────────┐
│  GUI binary                  │ │  CLI binary                                 │
│  ssahdrify.exe               │ │  ssahdrify-cli.exe                          │
│                              │ │                                             │
│  Tauri 2 + React +           │ │  clap (argv parsing)                        │
│  Tailwind frontend           │ │  deno_core / V8 (embedded JS bundle)        │
│  - 5 tabs                    │ │  - feature and utility subcommands          │
│  - i18n (zh/en),             │ │  - JSON reports + font diagnostics          │
│    dark/light/auto theme     │ │  - env_logger (stderr warnings)             │
│  - FontSourceModal UI        │ │  - sys-locale (--lang auto)                 │
└──────────────┬───────────────┘ └─────────────────────┬───────────────────────┘
               │                                       │
           Tauri IPC                       execute_script + JSON
               │                                       │
               └────────────────────┬──────────────────┘
                                    │
┌───────────────────────────────────┴──────────────────────────────────────────┐
│  Shared Rust crates                                                          │
│  - font-kit (system font discovery + matching)                               │
│  - fontcull stack (subsetting + name-table reader)                           │
│  - chardetng + encoding_rs (encoding detection + conversion)                 │
│  - serde / serde_json (serialization)                                        │
│  - rusqlite (font cache + user font index)                                   │
└──────────────────────────────────────────────────────────────────────────────┘

致谢 | Credits

原项目 / Original project: ying (2021), gky99/ssaHdrify (2024-2025)
应用图标 / App icon: 为本项目制作的 H-transition 图形，源文件见 src-tauri/icons/source.svg / H-transition artwork created for this project; source: src-tauri/icons/source.svg
部分界面图标 / Some interface glyphs: Feather Icons（MIT）

许可证 | License

本项目采用 GNU 通用公共许可证 v3.0 或更高版本授权。

This project is licensed under the GNU General Public License v3.0 or later.

来源与衍生作品 | Origin and Derivative Work

本项目是 ssaHdrify 的 Tauri 桌面重写版，原项目由 ying (2021) 创建，后由 gky99 (2024-2025) 维护。原项目同样采用 GPL-3.0 授权。

This project is a Tauri desktop rewrite of ssaHdrify. The original project was created by ying (2021) and later maintained by gky99 (2024-2025). It is also licensed under GPL-3.0.

HDR 色彩转换算法由 TypeScript（基于 Color.js）重新实现，方案参考了 Python 原版（使用 colour-science）。没有逐字复制代码；实现本身是新的，但出于许可证考量，本项目仍按衍生作品处理。

The HDR color conversion algorithm was reimplemented in TypeScript (using Color.js) based on the approach in the Python version (which used colour-science). No code was copied verbatim; the implementation is new, but the project is treated as a derivative work for license purposes.

算法归属 | Algorithm Attribution

src/features/font-embed/font-collector.ts 中的字体收集算法受 Aegisub 的 FontCollector 设计（BSD-3-Clause）启发。未复制 Aegisub 代码；TypeScript 实现由本项目独立编写。

The font collection algorithm in src/features/font-embed/font-collector.ts is inspired by Aegisub's FontCollector design (BSD-3-Clause). No Aegisub code was copied; the implementation is original TypeScript written for this project.

第三方依赖 | Third-Party Dependencies

下表列出主要直接依赖和随应用分发的资产；完整的传递依赖链请分别参见 package-lock.json 和 src-tauri/Cargo.lock。

The tables below list the main direct dependencies and bundled assets. For the full transitive dependency chain, see package-lock.json and src-tauri/Cargo.lock, respectively.

运行时依赖（随应用分发）| Runtime (shipped with the application)

组件 / Component	许可证 / License	用途 / Usage
Tauri	Apache-2.0 OR MIT	桌面应用框架 / Desktop app framework
Tauri plugins	Apache-2.0 OR MIT	对话框、文件访问、日志和单实例启动保护 / Dialog, filesystem, logging, and single-instance startup protection
React / React DOM	MIT	UI 框架 / UI framework
React Window	MIT	大列表虚拟滚动 / Virtualized large lists
Color.js	MIT	HDR 色彩空间转换 (PQ/HLG) / HDR color space conversion
ass-compiler	MIT	ASS 字幕解析（字体收集）/ ASS subtitle parsing for font collection
font-kit	MIT OR Apache-2.0	跨平台系统字体发现 (Rust) / Cross-platform system font discovery
fontcull	MIT / MIT OR Apache-2.0	字体子集化（含 fontcull-klippa、fontcull-skrifa）/ Font subsetting (includes fontcull-klippa, fontcull-skrifa)
chardetng	MIT OR Apache-2.0	编码检测 (Firefox 引擎) / Encoding detection (Firefox's engine)
encoding_rs	(Apache-2.0 OR MIT) AND BSD-3-Clause	编码转换 / Encoding conversion
rusqlite	MIT	字体缓存和本地字体索引 / Font cache and local font index
serde / serde_json	MIT OR Apache-2.0	Rust 序列化 / Rust serialization
deno_core	MIT	嵌入式 V8 JS 运行时（CLI）/ Embedded V8 JS runtime (CLI)
V8	BSD-3-Clause	JavaScript 引擎（经 deno_core 嵌入，CLI）/ JavaScript engine via deno_core (CLI)
clap	MIT OR Apache-2.0	CLI 参数解析（CLI）/ CLI argument parsing (CLI)
env_logger	MIT OR Apache-2.0	CLI 日志后端 stderr（CLI）/ CLI logging backend on stderr (CLI)
sys-locale	MIT OR Apache-2.0	OS 区域设置检测（驱动 `--lang` 自动检测，CLI）/ OS locale detection driving `--lang` auto (CLI)
base64	MIT OR Apache-2.0	Rust 侧字体载荷 base64 编码 / Base64 encoding for Rust-side font payloads
unicode-normalization	MIT OR Apache-2.0	Unicode 路径 / 输出键规范化 / Unicode path and output-key normalization
sha2	MIT OR Apache-2.0	样式编辑源文件 SHA-256 修订指纹 / SHA-256 source-revision fingerprints for Style Edit
rfd	MIT	启动失败时的原生错误对话框 / Native error dialog for startup failures
Feather Icons	MIT	主题切换等内联界面图标 / Inline interface glyphs, including the theme toggle

捆绑字体（随应用分发）| Bundled Fonts (shipped with the application)

字体 / Font	许可证 / License	用途 / Usage
Inter · © The Inter Project Authors	SIL Open Font License 1.1 · OFL-1.1	英文界面正文与标题 / English UI body + display face
Smiley Sans 得意黑 · © 2022–2024 atelierAnchor	SIL Open Font License 1.1 · OFL-1.1	中文界面标题展示字体（仅用于标题）/ Chinese-mode application title display face (headline only)

OFL-1.1 允许这些字体与任何软件一起捆绑、嵌入和再分发，包括 GPL-3.0 项目；字体及其衍生作品必须继续以 OFL 授权，不得单独销售，且修改版本不得使用相应许可证声明的保留字体名称。随附的 Smiley Sans 许可证声明了 Smiley 和 得意黑；随附的 Inter 许可证未声明保留字体名称。

OFL-1.1 allows these fonts to be bundled, embedded, and redistributed alongside any software, including GPL-3.0 projects. The fonts and their derivatives must remain licensed under OFL, must not be sold on their own, and modified versions must not use Reserved Font Names declared by their respective licenses. The bundled Smiley Sans license declares Smiley and 得意黑; the bundled Inter license declares none.

在桌面版中，点击页脚的「许可证」即可离线阅读项目 GPL 正文、两款捆绑字体的完整 OFL 文本，以及 Feather Icons 的 MIT 声明。

In the desktop app, choose Licenses in the footer to read offline copies of the project GPL, both bundled fonts' complete OFL texts, and the Feather Icons MIT notice.

构建时依赖（不随应用分发）| Build-time only (not shipped)

组件 / Component	许可证 / License	用途 / Usage
Tailwind CSS	MIT	CSS 工具框架 / CSS utility framework
TypeScript	Apache-2.0	TypeScript 7 发布检查 + TypeScript 6 API 兼容层 / TS7 release gate + TS6 API compatibility layer
Vite	MIT	构建工具 / Build tool
Tauri CLI	MIT OR Apache-2.0	Tauri 构建入口 / Tauri build entry point
ESLint	MIT	代码检查 / Linting
Stylelint	MIT	CSS 代码检查 / CSS linting
Prettier	MIT	代码格式化 / Code formatter
Vitest	MIT	单元测试 / Unit testing
js-base64	BSD-3-Clause	测试侧 base64 wire-format 编码 / Test-side base64 wire-format encoding
esbuild	MIT	为 CLI 嵌入打包 engine.js / Bundles engine.js for CLI embedding
GitHub Actions	MIT	CI 检出、Node 配置、缓存和依赖审查 / CI checkout, Node setup, caching, and dependency review

Name		Name	Last commit message	Last commit date
.github		.github
docs/screenshots		docs/screenshots
public		public
scripts		scripts
src-tauri		src-tauri
src		src
.gitattributes		.gitattributes
.gitignore		.gitignore
.npmrc		.npmrc
.prettierignore		.prettierignore
.prettierrc.json		.prettierrc.json
.stylelintrc.json		.stylelintrc.json
LICENSE		LICENSE
README.md		README.md
eslint.config.js		eslint.config.js
index.html		index.html
package-lock.json		package-lock.json
package.json		package.json
rust-toolchain.toml		rust-toolchain.toml
tsconfig.app.json		tsconfig.app.json
tsconfig.json		tsconfig.json
tsconfig.node.json		tsconfig.node.json
tsconfig.ts7.app.json		tsconfig.ts7.app.json
tsconfig.ts7.json		tsconfig.ts7.json
tsconfig.ts7.node.json		tsconfig.ts7.node.json
vite.config.ts		vite.config.ts

参数 / Parameter	默认值 / Default	说明 / Description
EOTF curve	PQ	PQ (ST 2084) 用于 HDR10/杜比视界；HLG 用于广播 HDR / PQ for HDR10/Dolby Vision; HLG for broadcast HDR
Target brightness	203 nits	SDR 字幕峰值亮度（BT.2408 标准值）。如果字幕太亮就调低，太暗就调高 / Peak SDR subtitle brightness (BT.2408 reference value). Lower it if too bright; raise it if too dim

Latest commit

History

Folders and files

Repository files navigation

SSA HDRify

浅色主题（中文）/ Light Theme (Chinese)

深色主题（英文）/ Dark Theme (English)

目录 | Contents

下载 | Download

功能 | Features

支持格式 | Supported Formats

使用方法 | Usage

HDR 色彩转换 / HDR Color Conversion

时间轴偏移 / Timing Shift

字体嵌入 / Font Embedding

样式编辑 / Style Edit

批量重命名 / Batch Rename

CLI 使用 | CLI Usage

快速示例 | Quick Examples

全部子命令 | All Subcommands

全局选项 | Global Options

诊断输出 | Diagnostics

字体缓存 | Font Cache

工作流 | Workflow

缓存位置 | Cache Location

漂移检测 | Drift Detection

限制 | Limitations

使用场景 | Background

HDR 转换原理 | How HDR Conversion Works

精度说明 | Accuracy Note

从源码构建 | Build from Source

前置条件 | Prerequisites

开发 | Development

构建 | Production Build

测试 | Testing

架构 | Architecture

致谢 | Credits

许可证 | License

来源与衍生作品 | Origin and Derivative Work

算法归属 | Algorithm Attribution

第三方依赖 | Third-Party Dependencies

运行时依赖（随应用分发）| Runtime (shipped with the application)

捆绑字体（随应用分发）| Bundled Fonts (shipped with the application)

构建时依赖（不随应用分发）| Build-time only (not shipped)

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Used by

Contributors

Languages