Ask HN: What is your favorite solution for transcription?

2 points·DanBlake··4 comments

Anyone have any experience with the following?

CMU Sphinx vs IBM Viavoice vs Dragon Server edition vs Julius

Also worth noting this is for English spoken language.

View on HN

The open source IBM Viavoice disappeared.

Dragon Server is expensive and proprietary but likely the best.

HTK/Julius is good.

Cmu Sphinx is my preference. Sphinx4 is java and works good if real time is not an issue.

Cmu's pocketsphinx is written in C and fast. Also works well on embedded devices/mobile phones.

To start I would download the sphinx's, they come with decent acoustic/language models on sf and try the demos out to see if they suit your needs. Once you want good accuracy thats when the fun begins...

Thanks, Il definitely start with Sphinx. When you say dragon is expensive, what #'s we talking about here? I dont mind that its proprietary if the quality is worth the cost.

Caligula, Do you mind if I contact you about a project I'd like to start with Sphinx4? wheresjoe @ hotmail.com

Learn to type faster! I went from 35 wpm to over 100 in just over a year of 4-hrs-a-day practice. I have a USB foot pedal so I can start and stop, rewind and fast-forward mp3 recordings. (It would be better if I could do this with speech command - my foot gets tired) Programs to alter the speed of playback typically suffer from pitch alteration.

If you want to do this for a living, look at http://www.careerstep.com

Reply on news.ycombinator.com