If your spoken sentences fall into a predictable pattern, you may be able to compute them in advance using just a Pi. With batch produced speech you can output whatever you want: reading news articles, private notes, whatever takes your fancy. Whilst that speed rules out interactive speech scenarios, there are still plenty of scenarios where it can add value. Speech is currently generated around six times slower than real time, but there are still plenty of practical uses Admittedly the current implementation runs fairly slowly, but the fact that a reasonably compute intensive task like this can be done on a cheap $55 computer at all is remarkable in itself. What this article shows is that you can generate high quality speech using a pre-trained model even on a lowly Raspberry Pi. As the project has progressed the output from the best models has become more and more human-like in quality. In practical terms it means anyone with a computer can train a synthetic voice model with data and then have the system output whatever speech you wish. Mozilla TTS is a marvellous project that has been going since early 2018 as part of their Common Voice initiative, aimed at opening up voice software and resources. So after a hearty breakfast, I set off and what’s below condenses down the solutions I found whilst struggling to get it successfully installed and working. That thought crossed my mind recently, but it was one of those “ because it’s there” kind of challenges and I found myself in possession of a shiny new Raspberry Pi 4, facing a long weekend where I would spend much of the time indoors (and with recent global circumstances, I’m not alone in that last respect I’m sure!).
Given the compute needed for high quality speech synthesis and the comparative low power of a Raspberry Pi you may wonder why on Earth anyone would decide to try this at all. I’ll briefly outline some context and explain the motivations for trying this out yourself.
The title of this article should give a fairly unambiguous idea of what it is about but before we dive into the instructions. Raspberry Pi 4 ready to have TTS installed!
0 kommentar(er)
