CSTR / Nagoya Institute of Technology · 2002–16 · HMM / DNN parametric

Merlin / HTS

Statistical parametric TTS: predict vocoder features (often with HMMs, later DNNs), then filter. The buzzy, muffled colour is the vocoder, not a bad speaker. This clip is the Merlin baseline from the FastSpeech demo page.

Known for

The pre-neural baseline every 2017–2019 paper had to beat.

Vendor / paper demo

LJSpeech (Merlin) · Merlin baseline from the FastSpeech demo page, sample 1.

I will quote an extract from the reverend gentleman's own journal.

Year / era
2002
License class
research
Cloning
No
Streaming
Yes
Emotion controls
No
Languages (tags)
en
Official
Source
Kokoro 82M · af_heart