tkwong94 / video-ocr

Extract text from local video file using google-video-intelligence text recognition

0. install node environment and do npm install

1. dowload youtube video using node-ytdl

ytdl "http://www.youtube.com/watch?v=_HSylqgVYQI" > myvideo.mp4

2. split video using ffmpeg (higher chance of failure when the video is long, split to <10 min is recommended)

ffmpeg -i source-file.mp4 -ss 0 -t 600 -c copy first-10-min.mp4

3.set google cloud api creditenal

export GOOGLE_APPLICATION_CREDENTIALS="KEY_PATH" (ref: https://cloud.google.com/docs/authentication/getting-started)

4.npm run dev

node index.js "video"

Future enhancement

Crop video before ocr to reduce noise (ref: https://video.stackexchange.com/a/4571)