4 ms·
The dataset as described in the paper linked below- "In the initial evaluation of our system, we use three voices; one English male and female (M1, F1), and on
by Gargoyle 9y ago
The dataset as described in the paper linked below-
"In the initial evaluation of our system, we use three voices; one English male and female (M1, F1),
and one Spanish female (F2). The recordings consist of short sentences sung at a single pitch and an
approximately constant cadence. The sentences were selected to favor high diphone coverage. For
the Spanish dataset there are 123 sentences, for the English datasets 524 sentences (approx. 16 and
35 minutes respectively, including silences). Note that these datasets are very small compared to the
datasets typically used to train TTS systems, but this is a realistic constraint given the difficulty and
cost of recording a professional singer"
- bubo_bubo 9y ago"but this is a realistic constraint given the difficulty and cost of recording a professional singer" Just because a singer is professional doesn't mean they're any good. My wife copes with adversity by singing and she can sing "fuck fuck fuck shit shit shit" in soprano, on key, from the kitchen. The only thing keeping her from singing in public is her stage fright. There are a /lot/ of people like her, that would answer an ad in the newspaper (or craigslist) that would like to /volunteer/ and contribute to a geeky project as long as they got credit in the paper. At that point, the largest non-tech cost winds up being the studio rental fee, if you have one.
- mwcampbell 9y agoYou could also go to karaoke bars and observe who the good singers are.