@zhouyousong For anyone else who is curious about how to get around this problem, the playRaw() method of M5.Speaker is non-blocking, but requires you to provide the PCM data directly as a bytearray or similar, as well as the sample rate.

If you look up the encoding of a .WAV file, it is fairly trivial to extract the sample rate (it's bytes 24:28 in the 44-byte header included at the start of any .Wav file). Getting the PCM data is then simply a case of reading the byte values from byte 44 onwards, until you reach the end of the file.

Here's a basic sketch:

fd = open('wavfile.wav', 'rb') # Open .wav file in read (bytes) mode header = fd.read(44) # Read .wav file header sample_rate = int.from_bytes(header[24:28], 'little') # Extract sample-rate ... # Can also extract other header information, such as bits per sample, number of channels, etc. buffer_time = 1 # 1s worth of audio data buf = bytearray(sample_rate*(bits_per_sample//8)*buffer_time) # Buffer large enough to contain 1s of audio # Loop fd.readinto(buf, sample_rate) # Read from file into buffer Speaker.playRaw(buf, sample_rate) # Non-blocking playback of audio data in buffer ... # and repeat for each additional second of audio in .wav file

This is a useful link to understand the contents of .WAV file header.

The only potentially tricky part in this is timing the file reads so they don't block the rest of your program for a long time.
I found that reading 1 second of 8000 samples per second, 16-bit, mono audio from the SD card takes about 67ms for the M5 FIRE (v1.0). Using a smaller buffer (0.5s/0.25s of audio etc.) will mean correspondingly shorter read times, with the caveat that you will have to read more often.

You can try the following code in the Block Designer: (note - it uses Timer 0 to time the periodic file reads, so if you are using Timers for other things, you will need to adjust accordingly).

""" file NonBlockingSpeaker time 2026-02-24 author Tom Fahey email tomp.fahey@gmail.com license MIT """ from machine import Timer from M5 import Speaker import micropython """ file NonBlockingSpeaker time 2026-02-24 author Tom Fahey email tomp.fahey@gmail.com license MIT License """ class NonBlockingSpeaker: """ note: en: '' details: color: '#0fb1d2' link: https://github.com/m5stack image: '' category: Custom example: '' """ def __init__(self): """ label: en: '%1 init' """ self.tim = Timer(0) def __setup(self, wav_file): """ label: en: ' %1 setup, wav_file: %2' params: wav_file: name: wav_file """ self.fd = open(wav_file, 'rb') self.header = self.fd.read(44) self.file_size = int.from_bytes(self.header[0:4], 'little') self.fmt_type = int.from_bytes(self.header[20:22], 'little') self.channels = int.from_bytes(self.header[22:24], 'little') self.sample_rate = int.from_bytes(self.header[24:28], 'little') self.bits_per_sample = int.from_bytes(self.header[34:36], 'little') self.data_size = int.from_bytes(self.header[40:44], 'little') self.buffer = bytearray(self.sample_rate*(self.bits_per_sample//8)) def playWav(self, wav_file): """ label: en: ' %1 playWav, wav_file: %2' params: wav_file: name: wav_file """ self.__setup(wav_file) self.counter = self.data_size//len(self.buffer) self.fd.seek(44) self.fd.readinto(self.buffer, len(self.buffer)) self.tim.init(mode=Timer.ONE_SHOT, period=1000, callback=self.__timer_cb) Speaker.playRaw(self.buffer, self.sample_rate) def __continue_playback(self, x): """ label: en: ' %1 continue_playback, _: %2' params: x: name: x """ self.fd.readinto(self.buffer, len(self.buffer)) Speaker.playRaw(self.buffer, self.sample_rate) def __timer_cb(self, _): """ label: en: 'method %1 param1 param2 ' """ if self.counter > 0: self.counter -= 1 self.tim.init(mode=Timer.ONE_SHOT, period=1000, callback=self.__timer_cb) micropython.schedule(self.__continue_playback, 1)