AISubs: Automated Video Transcription & Subtitle Translation Pipeline with WebUI

Showcase
1

Built this over the past week using Cursor. Here’s the result :backhand_index_pointing_down:

A fully automated system that watches your video library, queues and processes files, transcribes audio, and translates subtitles into any desired language :clapper_board::memo::globe_showing_europe_africa:

Models Used:

  • Claude 4-sonnet MAX
  • Claude 4-opus MAX

Cursor Usage:

  • Rules: RIPER-5
  • Token Cost: ~$80

:wrench: Tech Stack Overview

Core System:

  • :spouting_whale: Dockerized with full GPU support
  • :high_voltage: NVIDIA PyTorch container (CUDA 12.6)
  • :headphone: WhisperX + RTX acceleration
  • :globe_with_meridians: GPT-4o translation engine with parallel processing
  • :card_file_box: SQLite database for queue + state tracking
  • :eyes: Auto file monitoring and video detection
  • :hammer_and_wrench: Complete error handling + logging
  • :globe_with_meridians: Flask WebUI
  • :card_index_dividers: Batch mode + file scheduling
  • :laptop: API support
  • :bar_chart: Job status dashboard
  • :open_file_folder: Manual reprocessing queue
  • :chequered_flag: Per-file language selection
  • :page_facing_up: External .srt file detection + translation
  • :muted_speaker: Silent, Regular, and Focus modes

:rocket: Enhancement Pipeline

  • :open_mailbox_with_raised_flag: Telegram notification alerts
  • :brain: Metadata fetcher for context (TMDB, etc.)
  • :artist_palette: Subtitle styling (font, color, positioning, format)

I know tools like Subgen exist, but they don’t offer multilingual subtitle generation. English isn’t my first language—and after years of subtitle hunting, this is the solution my wife and I needed.

GitHub release with setup instructions coming soon :wrench::package:

WebUI
WebUI1890×896 115 KB