VoxForge
VoxForge
Yesterday, I discovered that I can help the project by recording things directly in that Java application.
Then, after recording, I saw that there has not been any activity since September (nobody added the recordings to the downloadable list).
As the 'listen' button in the Java application doesn't work for me, I cannot find out if the quality and loudness level of my recording is high enough. (Bug?) And if the recording isn't added to the downloads, I can never check... (female, norddeutsch, anonymous, 2013)
I would go on 'donating my voice', but if there is no activity and maybe no interest and if I cannot check if the quality is good enough, there is no reason for me to do so.
Could someone tell me if any activity on improving / making a German language model is still desired?
>has not been any activity since September
These submissions will be processed when I get to them. Each submission needs to be vaildated before it can be placed in the download section - validation is just making sure that the audio can be used by the speech recognition engine trainer... we get a surprising number of spam/junk submissions... sometimes a single prompt is bad, but the remainder of the submission is good, and this needs to be cleaned up...
>'listen' button in the Java application doesn't work for me
Which O/S, browser and Java versions are you using. Is there anything that appears in your logs?
Ken
I'm using Firefox 17.0.1 on Debian Squeeze. The Java Version is an older one which is deemed insecure, and it seems that something happened after I had to use it on your site.
Now (almost) everytime a web page is loaded in firefox, a sound is played (always one I recognize, though I don't know where it could come from, as I have emptied my browser cache after I noticed this).
Seems like I should somehow find out if the Java plugin is the source of this strange behaviour... I would really like to contribute, but using audacity and installing that prompting program first seems too much work to me (as I do not even use speech recognition) - doing this 'on the fly' using a web-based application is a really good idea, in principle. Isn't there something that does not use Java? I don't know enough about this...
>but using audacity and installing that prompting program
when using audacity to record prompts, you just read prompts from the VoxForge website, tar it up as a submission, and upload it - no other program is required to be downloaded...
>Isn't there something that does not use Java?
waiting for browser support of HTML media capture
Thank you! When I find the time, I'm going to try this, that way I can check for quality issues before submitting. Thanks for the links, that way I'll be faster :-) >waiting for browser support of HTML media capture Come on Mozilla, Google + Opera, hurry up :-) !
There's a problem with the prompts files: they only exist in English.
While I certainly could add my German-accented voice to your English speech model, I would prefer to do this for the German speech model...
Are there any prompts for German anywhere? Or can I just read anything (e.g. a Gutenberg book) that is available if I provide the text file with the reading?
>Or can I just read anything (e.g. a Gutenberg book) that is
>available if I provide the text file with the reading?
yes.
Just make sure whatever text you are reading is *not* protected by Copyright.
Public domain texts (like Gutenberg) are perfect from a licensing perspective, and from a acoustic model training perpsective. Because you are reading lots of different words in different contexts (which is better than the current VoxForge applet which only has a fixed list of words... for now).