Skip to content

v2.5.41 — Per-video workers and SupJav thumbnail fixes

Choose a tag to compare

View all tags
@Alos21750 Alos21750 released this 05 Aug 15:34
v2.5.41
9827da1

JableTV Downloader v2.5.41

English

Configurable workers per video

  • Added a persistent Per-video workers setting with a range of 1–16.
  • Available in Modern, SmallTool, the classic source GUI, CLI, and Docker.
  • Added the CLI option --max-workers-per-video N.
  • Added the Docker environment variable MAX_WORKERS_PER_VIDEO.
  • The existing CPU-derived default remains unchanged and is capped at 16.
  • HLS downloads use the selected value as their per-video segment-worker upper limit. SupJav direct downloads use up to the lower of four workers or the selected value.
  • Concurrent Downloads remains separate, so users can reduce per-video pressure on a proxy without reducing the number of queued videos.
  • Modern settings helper text now wraps cleanly at the minimum supported window size.

SupJav thumbnail reliability

  • SupJav thumbnails can now reuse the saved Cloudflare Bypass browser User-Agent, cf_clearance cookie, and correct Referer when required.
  • Saved clearance context is restricted to explicitly approved SupJav HTTPS hosts and is never forwarded to external, look-alike, or insecure HTTP thumbnail URLs.
  • If a saved bypass context has become stale, the thumbnail request is retried once without that saved context.
  • Failed thumbnail requests now display the localized No thumbnail state instead of remaining stuck on Loading....
  • If SupJav or all mirrors block an IP before the listing loads, an accessible network, proxy, or VPN may still be required.

AI subtitles and LLM API support

The existing local-first AI subtitle pipeline remains available. Japanese recognition always runs locally, and translation is local by default. Optional LLM translation supports OpenAI, Anthropic, Gemini, and custom OpenAI-compatible endpoints, with compatible presets for DeepSeek, OpenRouter, Groq, Ollama, and LiteLLM. xAI/Grok can be configured through the custom OpenAI-compatible fields.

繁體中文

可調整每部影片的下載執行緒

  • 新增可保存的「單片執行緒」設定，範圍為 1–16。
  • Modern、SmallTool、傳統原始碼 GUI、CLI 與 Docker 均可設定。
  • CLI 新增 --max-workers-per-video N。
  • Docker 新增 MAX_WORKERS_PER_VIDEO 環境變數。
  • 未手動調整時仍沿用原本依 CPU 自動計算的預設值，上限為 16。
  • HLS 下載會把此設定作為單片分段下載執行緒上限；SupJav 直接下載則最多使用「4 與設定值兩者中較小者」的執行緒數。
  • 「同時下載數」仍是獨立設定，因此可降低單片對 Proxy 的壓力，而不必減少排隊中的影片數量。
  • Modern 設定頁的說明文字在最小支援視窗尺寸下也會完整換行，不再被左右裁切。

改善 SupJav 縮圖載入

  • SupJav 縮圖在需要時會沿用已保存的 Cloudflare Bypass 瀏覽器 User-Agent、cf_clearance Cookie 與正確 Referer。
  • 已保存的 clearance 僅限明確允許的 SupJav HTTPS 主機使用，不會傳給外部圖片網域、偽裝網域或不安全的 HTTP 網址。
  • 若已保存的繞過資訊失效，縮圖會自動以不含該資訊的一般請求重試一次。
  • 圖片確實無法取得或解碼時，卡片會顯示本地化的「無縮圖」，不再永久停留於「載入中」。
  • 若 SupJav 本站或所有鏡像在清單載入前便封鎖目前 IP，仍可能需要可存取的網路、Proxy 或 VPN。

AI 字幕與 LLM API 支援

既有的本地優先 AI 字幕管線保持可用：日文辨識一律在本機執行，翻譯預設也在本機完成。使用者亦可選用 OpenAI、Anthropic、Gemini 或自訂 OpenAI-compatible 端點；相容模式提供 DeepSeek、OpenRouter、Groq、Ollama 與 LiteLLM 預設，xAI/Grok 可透過自訂 OpenAI-compatible 欄位設定。

Full changelog: v2.5.40...v2.5.41

Assets 7
Loading