As you add new recordings to the web recorder, the system occasionally builds examples of your synthetic voice for you to listen to. When a new one is built, a new number will briefly flash next to the Voices button at the top of the screen. The number indicates how many of these example voices have been built for you to listen to and compare. To listen to one of these voices, select it from the dropdown menu, type a sentence into the box, and click “listen”.
New voices are built whenever we think you have added enough new material to make a perceptible difference in the voice quality. This amounts to about every 25 sentences at first, then less frequently as you go along. A new voice is built every 400 sentences after reaching 800 sentences because after that, it takes a lot of additional recording to make a difference you can easily hear. These example voices are build with default parameters that are not necessarily well-tuned to your speech and so the voice quality is probably not quite as good as the final voice we will build for you, but they will give you a reasonable sense of progress as you do the recording. Moreover, you will not hear any example voices based on our DNN technology because they require too many computer resources to build.
NOTE: If you are recording the new “Gen3” inventory, the samples generated by the Voices button may be particularly poor and will not reflect the final voice we will build for you. This is because the Gen3 inventory was designed to be used with our latest DNN technology, which as mentioned above we cannot generate through the web recorder at this time. Moreover, the Gen3 inventory is particularly short and contains many sentence which encourage prosodic (pitch) variation, which the older online synthesis technology has particular trouble with. However, the shortness of the Gen3 inventory will allow you to quickly finish your recordings, so that you can listen to your final synthetic voice sooner, via the Audition process.