what is the inference time in GPU?.

[ajaysg-zz]

[alokprasad]

Depends on the platform ? Its not Tacotron2 is not realtime yet, but if we are able to compile
libtensorflow-core.a for Android using ndk16 or above ( rather than ndk14b) , i think with compiler optmization and memory mapped model , it could be realtime.

[ajaysg-zz]

I am trying to use the model in GTX1080 card. Will I be able to get realtime inference with less GPU memory usage?.

[alokprasad]

i dont have 1080 , so haven't done inference on it.

[ajaysg-zz]

how long LPCtron takes to synthesize 2 secs of audio in CPU?.

[mingkzhou]

how long LPCtron takes to synthesize 2 secs of audio in GPU?@ajaysg

[xxoospring]

how long LPCtron takes to synthesize 2 secs of audio in CPU?

have you done this experiments?