9 ms·
It's pretty awesome. I'm going through the italian verbs, it's very useful. Where did you get the audio files? EDIT: I noticed some audio files sometimes take
by rorrr 15y ago
It's pretty awesome. I'm going through the italian verbs, it's very useful. Where did you get the audio files?
EDIT:
I noticed some audio files sometimes take up to a minute to load (most play instantly though). Chrome/Win7 here.
EDIT2:
It would help if there were an option to play both Italian and English audio, just to use the auditory memory.
- philfreo 15y agoCool! The audio is text-to-speech, we licensed software from a couple different companies and combined/customized it. We tried to make it fast / preload a lot / are using a CDN (details about that in the post), hopefully the slowness isn't typical.
- kd1221 15y agoDid you check out Festival/Festvox/Cepstral? http://festvox.org/ http://festvox.org/ Festival/Festvox are free to use. I built an Arabic speech synthesizer with it in the spring of 2002 for a class project. It's quite flexible. Cepstral is a commercial offshoot of Festvox. I ran through a few of the foreign language quizzes (German, Spanish, Arabic) and they're pretty good. They do need a bit of curation, because some of the answers required superfluous punctuation (Food = "das Essen -"). Some data scrubbing is in order. The quizzes with non-Latin alphabets are a bit difficult to do, but understandably so. The beauty of these languages is in their writing anyway. Are you planning any iPad/tablet apps?
- asuth 15y agoWe decided the most realistic audio came from commercial vendors. In terms of mobile stuff, we've thought about doing our own but for now there are great apps being built on our API. We just launched V2 of our API with OAuth integration: http://quizlet.com/blog/quizlet-announces-api-2/ http://quizlet.com/blog/quizlet-announces-api-2/ If anyone's interested in building mobile stuff, feel free to play around with it. Or contact us (email in profile) and we may have a job for you :)
- koopajah 15y agoCan you tell which TTS engine you specifically use or would you rather keep it a secret? Do you plan adding voice recognition later to test our pronunciation too?