[FR] Support For fastest TTS model

[mablue]

Hi. we tested all models also vocos+metcha but still the speed is low for use as screen reader and still espeak is the best choise for a blind couse of speed. So we go to try mixer tts but its need a big labeled data to train and its not good for use in screen reader. because much of languages haven't a very big dataset to train. also we cant know and dont need to lable the texts passing from screen readers. so we find this thing: https://github.com/roatienza/efficientspeech

its faster than mixertts and can be a good alternative to espeak but I think its not supported in sherpa onnx. please support it its the best and faster model in train just with 100 voice as dataset.and also its very fast in generating voices. thanks img

[csukuangfj]

its faster than mixertts

Please show speed numbers

[csukuangfj]

It is way slower than matcha + vocos