Lots of TTS examples, what about Speech to Text?

[MaxMaeder]

I see a lot of examples of text-to-speech implementations, however I was wondering if anyone had experience implementing speech-to-text using this library and a API like the Azure Speech Service.

As for my specific hardware, I'm trying to use a I2S microphone and an esp32 with WiFi

[pschatzmann]

I have never tried, but I don't see any reason why this should not work...

[MaxMaeder]

For these APIs I need to POST the audio data as a binary wav, is there a way for me to Stream this data to a HTTP post endpoint in Arduino?
docs

[pschatzmann]

Sure: have a look at the HttpRequest class

[MaxMaeder]

Oh perfect!

[kaloprojects]

Great thread, thank you both!. @MaxMaeder : that's exactly what i am missing to my use case too!.

Background (standalone ESP32 OpenAI chatbot): I'm also using an ESP32 wifi connected (Arduino IDE) .. entering a user input via Serial .. sending as request to OpenAI API (POST message) and receiving and printing out show response. Meanwhile also implemented a Open AI Text-To-Speech (works great, using schreibfaul1 audio.openai_speech()), so i can listen to the AI answer via I2S amplifier (MAX98357) ....
... BUT what i totally miss is a connection to any Speech-To-Text API via my connected I2S microphone INMP441 (having any STT API call would allow me to replace my keyboard with my voice and having a standalone solution).

I might not have the skills for implementing the STT by myself :| (even with @pschatzmann great link) .. so i cross my finger you can have progress to share ? :)

What i wished is something like (any code snippets welcome!), using @pschatzmann's library:

• recording an audio (e.g. 3-10 sec, as long i press a btn) with my I2S microhone (in case a I2S SD card storage needed i could add)
• sending this wav file/buffer (or even as stream ?, would be best) to an STT Service (via POST message) and receiving text transcription from API.

I would love most this STT API from Open AI), or alternativelly Deepgram, both seem very powerful, not expensive, multi languages detection, and most user (me included) have an account already, so no complex Azure registration or costs needed. But i accept any vendor (e.g. Azure), as i currently have no STT idea at all.

@MaxMaeder : do you think you can share some first code examples how to call such an SST API (with pschatzmann help above) ?

[arduibag]

@MaxMaeder : do you think you can share some first code examples how to call such an SST API (with pschatzmann help above) ?
It would be great!

[MaxMaeder]

Made a demo, see below

[MaxMaeder]

I made a little Assistant demo: uses Azure SST/TTS + OpenAI API.

https://github.com/MaxMaeder/arduino-audio-tools-assistant