[WIP] Various Improvements

[kahrendt]

Media playback is improved:

• A basic resampler adjusts sample rates (quality isn't great and this is slow)
• Handles both mono and stereo media
• Supports FLAC files
• Adds volume/mute control via the DAC (the wheel works for increasing/decreasing volume)
• Music Assistant streams work (both mp3 and flac), but since it requires resampling, the audio quality isn't great

Fixes a few issues:

• Reduces long delay in the voice assistant component marking the end after a TTS announcement is finished
• Eliminated the media player component from warning about delaying too long
• Eliminated the speaker hiss with no audio playback

This isn't very stable at the moment, hence the WIP PR.

• Internal memory use is high.. could potentially hit issues if decoding two streams
• Changing media streams causes issues (memory not properly freed, can't start a second stream without two attempts)
• Buffers are not optimized, especially on the announcement side (they are way too big for announcements, causing a delay in playing the response)

[github-actions]

Firmware built successfully! 🎉

Download and extract the firmware to install with https://web.esphome.io

Make sure to choose esphome-voice-kit-esp32s3/esphome-voice-kit-esp32s3.factory.bin.

[github-actions]

Firmware built successfully! 🎉

Download and extract the firmware to install with https://web.esphome.io

Make sure to choose esphome-voice-kit-esp32s3/esphome-voice-kit-esp32s3.factory.bin.

[kahrendt]

These new set of commits should improve stability, though internal memory usage is still too high for decoding and resampling two simultaneous streams. Main highlights:

• Basic wav header decoding (thanks @synesthesiam )
• Flac headers are decoded consistently
• Able to stop announcements on the wake word (I will improve how this functions when I work on the input/mWW/voice assistant side of things)
• Starting and stopping streams is a lot more stable
• Initial support for playing back local files (thanks @jesserockz for your file external component). I quickly added a button to show how to call it, but it should be easy to adapt it into a timer beeper
• Buffer sizes were increased/aligned, so it is much less likely to stutter after music playback has started (though it still happens at the beginning of a file on occasion)
• I apply two Biquad filters when resampling. This greatly improves the sound quality when playing back typical music files (at 44.1 kHz or 48 kHz). It does seem to reduce the sound quality when downsampling files with low sample rates (say 22.05 kHz).
• Modified mWW and voice assistant components to not use high frequency loopers, so another CPU core is available more often. This eliminates most stutters when decoding and mixing two audio files while wake word detection is going on (resampling is still the bottleneck here though)
• Switched to using the left microphone stream. This seems a lot cleaner with less false activations/better end of speech detection. I am having trouble finding documentation from XMOS regarding why the left and right channels seem different.