Release v1.5.0

🚀 Release v1.5.0

🔍 Commits

✨ Features

• feat(ai): register whisper transcription tasks into global download and task progress queue (01fd175)
• feat(ai): add float button right-click quick action for whisper transcription and txt download with language selector (ab203f1)
• feat(ai): add transcription latency, realtime factor, and throughput metrics (220850f)
• feat(ai): maintain background whisper tasks across modal open-close and prevent session collision (9250141)
• feat(ai): persist transcription results per video episode across tab switching (66425de)
• feat(ai): support dual chunk callbacks and live progress percentage estimation (7cdab90)
• feat(ui): add WhisperTranscriber component and AI speech-to-text settings integration (2dbc2ae)
• feat(ai): add pure frontend Whisper speech-to-text module with WebGPU and CDN acceleration (f6c5cd1)

🐛 Fixes

• fix(ai): serialize webgpu model inference to prevent session already started collisions across episodes (6e8dc4e)
• fix(ai): avoid mutating frozen module namespace and record whisper errors to diagnostic log (0d14566)
• fix(ai): eliminate DOM script import.meta syntax error and enforce GM_xmlhttpRequest fetch proxy in all scopes (a28ab21)
• fix(ai): strict WebGPU device detection and add GM_xmlhttpRequest CORS proxy for model downloads (7660739)
• fix(build): remove SystemJS/CSP-eval dependency and bundle as pure IIFE userscript (bc424fe)

📚 Documentation

• docs: balance height proportions for demo and settings preview (f772391)
• docs: layout demo and settings preview side-by-side with picture theme switching (e828463)
• docs: use picture tag with prefers-color-scheme for responsive theme switching (f556372)
• docs: consolidate hero preview table with dark/light mode picture tags (ee1ca0b)
• docs: use fastly.jsdelivr CDN for README screenshots (0635349)
• docs: add dark/light mode picture tags and comparison tables to README (4b3cade)

💄 Style & Refactor

• style(ui): remove redundant WebGPU badge on AI transcription tab (9f4dae3)
• perf(ai): add event loop yielding, live chunk streaming, and memory cache disposal (0eb3db5)

🔧 Chore & Test

• chore(release): v1.5.0 (0a6beb3)
• test(ai): add comprehensive unit and integration tests for WebGPU Whisper transcription (8d2be52)
• chore(assets): fix demo (0327711)

Full Changelog: v1.4.0...v1.5.0

Release v1.4.0

🚀 Release v1.4.0

🔍 Commits

✨ Features

• feat(batch): stream batch tasks to external downloader immediately per episode (74912cc)
• feat: make CDN priority ordering configurable and unified across internal & external downloaders (554f0f5)
• feat: integrate external downloader selection and quick action routing (OCP) (14b106a)
• feat(downloader): integrate AB Download Manager with liveness probe and OCP architecture (ff62da1)
• feat(settings): make low-speed failover duration threshold configurable (7682455)
• feat(cdn): add intelligent CDN priority ranking and passive low-speed auto failover (c56393a)
• feat(download): prioritize actively progressing tasks at the top of the queue with auto-scroll (142a07d)
• feat(download): provide dual audio-muxed and video-only buttons with prioritized audio muxing (4f46bb5)

🐛 Fixes

• fix(batch): lightweight stream resolution and instant per-episode external dispatch (ba22017)
• fix(downloader): optimize ABDM headless queue integration, robust CDN fallback, and cookie headers (713f16e)
• fix(ui): isolate input keyboard events from Bilibili player and enhance input affordances (38dc404)
• fix(downloader): consolidate batch external downloads, pass auth headers, align toast, and add m4s notice (65da624)
• fix(ui): eliminate bottom blank space in settings modal (07ea325)
• fix(ui): replace all emojis with Lucide icons across UI components (46a15e7)
• fix(metadata): add 127.0.0.1 and localhost to @connect whitelist for external downloader probing (2b293fc)
• fix(logs): resolve copy button and action controls being squeezed out by trace filter (1301feb)
• fix(muxer, http): fix mp4box infinite loop on malformed boxes and validate chunk ranges (edbd7ac)
• fix(muxer): set audio track handler to soun and compute movie duration for seekable playback (5a48f7b)
• fix(ui): eliminate duplicate scrollbar in task queue and restore global font in log console (28c9137)
• fix(ui): optimize diagnostic log layout with adaptive docking and maximize mode (f81f7e9)
• fix(api): support nullish fields in Bilibili schemas and add defensive fallback for DASH parsing (aeff1f2)

💄 Style & Refactor

• refactor(downloader): abstract external downloader architecture and eliminate boilerplate (2785b8c)
• refactor(settings): remove File System Access API directory picker and related storage (ac8460d)
• refactor(types): eradicate all any types and introduce zod schemas (c083ce7)

🔧 Chore & Test

• chore(release): bump version to 1.4.0 (203170e)
• ci: enhance release notes generation with conventional commit categories (2ec2da0)
• chore(.github): add issue forms and pull request templates (ba26109)
• ci: configure test gate checks in lefthook and github actions ci workflow (f29befc)
• test: add comprehensive unit, integration and playwright e2e tests (b174fcc)
• ci(lint): elevate oxlint rules and typescript strict flags (59ab051)
• build(deps): install zod, vitest, testing-library and playwright dependencies (f9b2397)

Full Changelog: v1.3.4...v1.4.0

Release v1.3.4

Full Changelog: v1.3.3...v1.3.4

Release v1.3.3

Full Changelog: v1.3.2...v1.3.3

Release v1.3.2

Full Changelog: v1.3.1...v1.3.2

Release v1.3.1

Full Changelog: v1.3.0...v1.3.1

Release v1.3.0

Full Changelog: v1.2.0...v1.3.0

v1.2.0 - 精准分P字幕提取与独立文件命名

✨ v1.2.0 更新内容

1. 彻底解决字幕提取内容不匹配问题：
   • 接入 B 站现代播放器 WBI 鉴权签名（纯 TS RFC-1321 MD5 + 动态密钥混淆），改用 /x/player/wbi/v2\ 专用接口，彻底规避旧版接口 CDN 缓存串扰导致分 P 拿到错误字幕的问题。
2. 多分 P 剧集独立命名隔离：
   • 多分 P 视频自动注入 _P{序号}_{小节名}\，确保下载的 .srt\ 字幕、.ass\ 弹幕、.mp4\ 视频与 .m4a\ 音频文件名独立精准，不再相互覆盖或混淆。
3. 毫秒级时间轴对齐：
   • SRT 字幕与 ASS 弹幕转换重构为纯整型毫秒/百分秒计算，消除 JS 浮点取模偏差。
4. Bangumi / 多类型页面 BVID 解析容灾：
   • 增加对 \window.INITIAL_STATE\ 剧集状态的兜底提取。

Release v1.1.0

Full Changelog: v1.0.2...v1.1.0

Release v1.0.2

Full Changelog: v1.0.1...v1.0.2