I2S Audio for ESP32~

build_flags                 = ${env:tasmota32_base.build_flags}
                              -DUSE_I2S_ALL

The main difference to the older ESP8266 sound driver is the configuration of the various settings at runtime with the command i2sconfig, which uses a hidden driver file.

I2S (Inter-IC Sound) is a serial, synchronous communication protocol that is usually used for transmitting audio data between two digital audio devices.
The I2S framework of the ESP-IDF supports 3 communication modes which are standard, PDM and TDM. TDM is the most advanced mode and very uncommon in the IOT world - there is no support for it in Tasmota yet.

Support for different I2S modes varies across the ESP32 family:

Note the limited support for PDM microphones.

Audio settings~

Without any additional argument i2sconfig will print the current audio configuration to the console in JSON format, e.g.:

MQT: stat/tasmota_4359CC/RESULT = 
{"I2SConfig":
{"Sys":{"Version":2,"Duplex":0,"Tx":0,"Rx":1,"Exclusive":0,
"MclkInv0":0,"MclkInv1":0,"BclkInv0":0,"BclkInv1":0,"WsInv0":0,
"WsInv1":0,"Mp3Preallocate":1},
"Tx":{"SampleRate":16000,"Gain":10,"Mode":0,"SlotMask":3,
"SlotConfig":0,"Channels":2,"APLL":1},
"Rx":{"SampleRate":32000,"Gain":30,"Mode":1,"SlotMask":1,
"SlotWidth":32,"Channels":1,"DCFilterAlpha":32511,
"LowpassAlpha":17719,"APLL":1,"WsWidth":32,"WsPol":0,"BitShift":1,
"LeftAlign":1,"BigEndian":0,"LsbOrder":0,
"DMAFrame":768,"DMADesc":5}}}

Some properties are easy to understand like number of channels or sample rate.
The "weird ones" a very likely data structures from the I2S audio framework of the ESP-IDF. Numeric values for these properties are integer representations of enums in many cases.
This allows for a very flexible audio setup to support a wide range of hardware combinations with one firmware image.

Changing values is done by passing a JSON with the new key-value-pair.

Examples:

i2sconfig {"Rx":{"Mode":0}} switch input of the microphone to standard modes, where all other default values are set to use an INMP441 with left-channel-configuration.
i2sconfig {"Rx":{"SlotMask":2}} would change the microphone in standard mode to the right channel.

i2sconfig {"Rx":{"SampleRate":48000}} changes sample rate of the microphone.
i2sconfig {"Rx":{"DMAFrame":768}} use DMA buffer for higher sample rate.

How to apply complex settings? The ugly truth ...~

In order to support basically any type of microphone in a combination with a certain ESP32 type, we need to find documentation about the needed settings (this works rarely or never) or find/generate demo code, which can be converted to Tasmotas I2S driver.
In example the INMP441 does not need the same setting on different SOC's of the ESP32 family.

For the ESP32-S2 this microphone works with the following configuration macro from Espressif (found with trial and error):

#define I2S_STD_PHILIPS_SLOT_DEFAULT_CONFIG(bits_per_sample, mono_or_stereo) { \
    .data_bit_width = bits_per_sample, \
    .slot_bit_width = I2S_SLOT_BIT_WIDTH_AUTO, \
    .slot_mode = mono_or_stereo, \
    .slot_mask = (mono_or_stereo == I2S_SLOT_MODE_MONO) ? \
                I2S_STD_SLOT_LEFT : I2S_STD_SLOT_BOTH, \
    .ws_width = bits_per_sample, \
    .ws_pol = false, \
    .bit_shift = true, \
    .msb_right = true, \
}

Now we have to translate this.

I2S_SLOT_BIT_WIDTH_AUTO is an enum with an int value of 0, so we need:

i2sconfig {"Rx":{"SlotWidth":0}}

I2S_STD_SLOT_LEFT and I2S_STD_SLOT_BOTH are enums too with values of 1 and 3. For mono:

i2sconfig {"Rx":{"SlotMask":1}}

The next values are pretty obvious:

i2sconfig {"Rx":{"WsWidth":16}}
i2sconfig {"Rx":{"WsPol":0}}
i2sconfig {"Rx":{"BitShift":1}}

The C struct member .msb_right does only exist on ESP32 and ESP32-S2 and is not part of the driver settings with this name.
Instead the settings key LsbOrder is reused for these 2 SOC's:

i2sconfig {"Rx":{"LsbOrder":1}}

The rest of the data is inferred from the currently fixed sample rate of the Tasmota driver of 16 bits and channel number with the driver key channels.

Audio Output~

For audio output an I2S digital audio decoder (DAC) board is required. It is recommended to use an external DAC

Internal DAC~

ESP32 has two 8-bit DAC (digital to analog converter) channels, connected to GPIO25 (Channel 1) and GPIO26 (Channel 2).

Those channels can be driven via the I2S driver when using the “built-in DAC mode” enabled with USE_I2S_NO_DAC

Commands~

Audio Input~

For microphone input an I2S microphone must be connected. It is pretty common, that the captured audio signal has a very low volume.
Hence a relatively high gain factor is needed, that can be configured with:
i2sconfig {"Rx":{"Gain":30}}

I2S Microphone - standard mode~

Use command i2sconfig {"Rx":{"Mode":0}} to switch audio input to standard mode.

If you're using only the microphone without a DAC you still need to set pin I2S_DOUT to an unused GPIO.

I2S Microphone - PDM mode~

Pulse density modulation (PDM) microphones are handled as I2S microphones in the ESP-IDF and need only two data wires in comparison to standard I2S microphones. They're used in ESP32-S3-BOX, Seeed Xiao Sense and others.

Use command i2sconfig {"Rx":{"Mode":1}} to switch audio input to PDM mode.

When using PDM microphones the microphone CLK pin is configured as I2S_WS in Tasmota.

Commands~

Optional: i2sconfig {"Sys":{"Mp3Preallocate":1}} to turn on buffer preallocation at boot time. Can solve issues, if there is not enough contiguous RAM available later.

I2S Audio Bridge~

Starts an UDP audio service to connect 2 ESP32 devices as an audio intercom (an example).

Needs audio output and microphone on 2 devices (no PSRAM needed)

build_flags                 = ${env:tasmota32_base.build_flags}
                              -DUSE_I2S_ALL
                              -DUSE_I2S_BRIDGE

If a push to talk button is defined the bridge goes to write mode if the button is pushed and to read mode if the button is released