31 lines
880 B
Markdown
31 lines
880 B
Markdown
Title: Building an Android speech recognition prototype
|
|
Date: 2026-08-05
|
|
Category: News
|
|
Slug: ai-speech
|
|
Lang: en
|
|
|
|
<video controls width="700">
|
|
<source src="../../vid/2026-08_ai-speech.mp4" type="video/mp4"/>
|
|
</video>
|
|
|
|
# July
|
|
|
|
AI came firmly into my life because I started creating a prototype of
|
|
a speech recognition application for Android. So far, I have implemented
|
|
both Speech-to-text and Text-to-speech.
|
|
|
|
Speech-to-text is powered by [Vosk][vosk] and Text-to-speech by
|
|
[Sherpa-Onnx][sherpa]. Both libraries work on-device without sending data
|
|
to a server.
|
|
|
|
# August
|
|
|
|
I'll integrate an LLM between STT and TTS to enable the application to provide
|
|
voice consultations based on the provided dataset.
|
|
|
|
P.S. I publish weekly reports in the [TG channel][tg].
|
|
|
|
[sherpa]: https://k2-fsa.github.io/sherpa/onnx
|
|
[tg]: https://t.me/kotlindialect
|
|
[vosk]: https://alphacephei.com/vosk
|