Microsoft · 2022–24 · NS 1–3
Microsoft’s ‘human parity’ line: NaturalSpeech, then NS 2 (latent diffusion, zero-shot), then NS 3. The LJSpeech clip is from the StyleTTS 2 comparison page, same sentence as StyleTTS 2 and VITS. A second clip is from the original NaturalSpeech demo. A third is NaturalSpeech 2’s official zero-shot LibriSpeech example.
Claiming human-level MOS, then scaling to zero-shot.
Vendor / paper demo
LJSpeech (StyleTTS 2 comparison) · NaturalSpeech sample hosted on the StyleTTS 2 demo page, LJSpeech ID 00005.
After the construction and action of the machine had been explained, the doctor asked the governor what kind of men he had commanded at Goree,
Vendor / paper demo
LJSpeech (official NS demo) · Official NaturalSpeech demo, sample 0.
The lax discipline maintained in Newgate was still further deteriorated by the presence of two other classes of prisoners who ought never to have been inmates of such a jail.
Vendor / paper demo
NaturalSpeech 2 (zero-shot LibriSpeech) · Official NaturalSpeech 2 project-page zero-shot sample (WAV converted to MP3).
Indeed, there were only one or two strangers who could be admitted among the sisters without producing the same result.