hide

直播

89 days ago

audio loopback 基本的訪談逐字稿

Unfiled. Edited by che wei liu 89 days ago

做法搜集

IBM Watson 批次語音轉文字網站：https:/Python SpeechRecognition 套件/speech-to-text-demo.mybluemix.net/

Watson 目前需要廣播節目等級的收音（硬體、空間）條件，才能達到「人工校正比重新打字快」。

Linux 系統 audio loopback：

https://www.freedesktop.org/wiki/Software/PulseAudio/Documentation/User/Modules/#index63h3

Python 的 SpeechRecognition 套件

https://youtu.be/31DZfkYRvI4

Macpaul Lin> it works! 透過一條音源線，把Speaker out put導到 MIC去，然後就可以一邊放youtube，然後另外一邊用Google語音辨識，讓電腦幫你作逐字稿！（Firefox x Chrome) 但是Google語音辨識速度明顯比人講話慢。猜測應該跟網路頻寬，以及計算能力有關。所以理論上，對於這種背景雜音乾淨的語音檔，可以接Google web 語音辨識的API，就能輕鬆做出基本的訪談逐字稿 https://www.facebook.com/photo.php?fbid=10152824899882910&set=a.172718147909.122769.532627909&type=1&theater&notif_t=comment_mention

https://cloudplatform.googleblog.com/2017/04/Cloud-Speech-API-is-now-generally-available.html?m=1

Scott Tsai> 自製做法 https://www.facebook.com/700484037/posts/10153115500289038

Show full change Show only preview

Stop sharing the collection with ?

This pad is open to "", so will still be able to access it.

Logout

Contact Support

Log in / Sign up