Sayboard
Sayboard copied to clipboard
ElishaAz
sherpa-ncnn as backend?
hi. i found that there is another speech recognition project which may provide better result than vosk. do you have interest of integrate it with sayboard? https://github.com/k2-fsa/sherpa-ncnn
it might be a good idea to add this language model too https://github.com/coqui-ai/STT-models
Sounds interesting. I'll take a look
Any updates? I would like to provide any help if you have any issues.
By the way, http://github.com/k2-fsa/sherpa-onnx is also a good candidate. It supports both streaming and non-streaming speech-to-text as well as text-to-speech.
Any updates? I would like to provide any help if you have any issues.
Sorry, I'm busy right now.
The main problem with these projects is that they don't provide binaries on Maven, which might cause trouble with F-Droid (specifically with Reproducible Build)
Sorry, I'm busy right now.
The main problem with these projects is that they don't provide binaries on Maven, which might cause trouble with F-Droid (specifically with Reproducible Build)
can you tell me which binary are you struggling on? i tried to build sherpa-onnx (rather than ncnn, because it have nnapi support) recently. the only blob i can find is https://github.com/csukuangfj/onnxruntime-libs . which have a build script in gh action. so that it is possible to build it from scratch
The main problem with these projects is that they don't provide binaries on Maven, which might cause trouble with F-Droid (specifically with Reproducible Build)
I did tried to package the sherpa demo for F-Droid. Not sure about reproducible build but I thought it's not a big problem.
