Video2Sub

Desktop subtitle extraction

Extract subtitles from video — on your machine.

Video2Sub reads hardcoded on-screen text with OCR or transcribes dialogue with ASR, then helps you translate and review — all in a desktop app for Windows and macOS.

Local-first processing · PaddleOCR + faster-whisper · DeepL translation

Same tabs as the app

Four steps from video file to reviewed subtitles.

01

Process video

Open mp4, mkv, or other supported files. Draw crop boxes for OCR, or switch to ASR for spoken dialogue.

02

Generate subtitles

Run PaddleOCR or faster-whisper locally — or send cropped frames / audio to your own GPU server.

03

Translate

Turn source SRT into Korean, English, Japanese, Chinese, and more with DeepL or local Qwen.

04

Review

Play the video with subtitle overlay, edit cues, run frame OCR, and save SRT next to the file.

Local-first by design

Your full video stays on your computer.

No cloud upload

OCR and ASR run on your PC. Output SRT files are written beside the source video.

Remote GPU, minimal data

Optional GPU server mode sends cropped frames or Opus audio — never the entire video file.

Keys on device

DeepL and OpenAI keys are stored locally. Translation calls go through the desktop app, not the browser.

Built for real workflows

Production tools, not a generic uploader.

Video2Sub reuses proven engines — PaddleOCR, faster-whisper, and optional remote GPU — inside a focused subtitle pipeline.

Hardsub OCR

Read burned-in subtitles with crop zones, dual-zone mode for two lines, and SSIM-based frame sampling.

Speech ASR

Transcribe dialogue with faster-whisper. Anime, balanced, and clear-speech presets tune VAD and gain.

Translation

DeepL Free API, local Qwen, or remote GPU translation. Auto-translate right after extraction if you want.

Review workspace

Edit cues on a timeline, find & replace across languages, overlay subtitles on video, and spot-fix with frame OCR.

Batch queue

Drop a folder, queue multiple episodes, and let OCR or ASR run while you move on to the next task.

Remote GPU server

Use a LAN or hosted GPU machine for heavy OCR/ASR while keeping video open and cropped on your laptop.

Formats & languages

What the app actually supports.

Input video

mp4 · mkv · avi · mov · wmv · webm · m4v · ts · flv

Output files

SRT beside the video · MyVideo.ko.srt · MyVideo.en.srt · …

OCR languages

Korean · Japanese · Chinese · English · Spanish · French · German · +11 more

Translation

16 primary targets · 24 additional via DeepL · local Qwen optional

Who uses Video2Sub

Made for subtitle work that already happens on desktop.

Fansub & anime teams

Pull hardsubs or dialogue from MKV releases, refine timing, and ship Korean or English SRT.

Post-production

Turn screen text or rough dialogue into timed cues before hand-off to QC or mastering.

Localization vendors

Keep source and translated tracks aligned in review, with per-language SRT files side by side.

Solo creators

Work from one machine — or offload OCR/ASR to a home GPU box without sending full uploads.

Pricing

Start free. Scale when you need to.

Pay for translated characters and collaborators as your volume grows.

Basic

$0 / month
  • 1,000 translation characters / month
  • 1 collaborator
  • Free forever
Start free

Max

$99 / month
  • 3,000,000 translation characters / month
  • 15 collaborators
  • $79.20 / month billed annually
Choose Max

FAQ

Questions, answered.

What is Video2Sub?

A desktop app for Windows and macOS that extracts subtitles from video using OCR (on-screen text) or ASR (speech), then helps you translate and review the result as SRT files.

Should I use OCR or ASR?

Use OCR when subtitles are burned into the picture. Use ASR when you only have spoken audio. The app runs one extraction mode at a time.

Is my video uploaded anywhere?

No. Processing happens on your computer. If you enable remote GPU mode, only cropped frames or compressed audio are sent to your server — not the full video file.

What do I need to run it?

A valid license.key next to the app, the bundled OCR/ASR engines (installed on first setup), and optionally a DeepL API key when you want translation.

How are translation characters counted?

Each source character counts once for every target language you translate into.

Do I need a website account?

No. Download and use the desktop app without signing in. Website accounts are optional and reserved for future online features.

Ready to extract your first SRT?

Install Video2Sub on Windows or macOS, place license.key beside the app, and open your first video in minutes.