ESPHome 2025.5.0 S3 Box 3 reboots on end of voice assistant announcement

[chrisdunnname]

Since updating to ESPHome 2025.0 the S3 box reboots on the end of any conversation when it stops the mixing speaker.

[01:13:50][D][voice_assistant:626]: Speech recognised as: "Hi."
[01:13:50][D][voice_assistant:456]: State changed from STOPPING_MICROPHONE to AWAITING_RESPONSE
[01:13:50][D][voice_assistant:598]: Event Type: 5
[01:13:50][D][voice_assistant:631]: Intent started
[01:13:50][D][i2s_audio.microphone:467]: Task is stopping, attempting to unload the I2S audio driver
[01:13:50][D][i2s_audio.microphone:472]: Task is finished, freeing resources
[01:13:50][D][voice_assistant:598]: Event Type: 6
[01:13:50][D][voice_assistant:598]: Event Type: 7
[01:13:50][D][voice_assistant:656]: Response: "Hello from Home Assistant."
[01:13:50][D][voice_assistant:598]: Event Type: 8
[01:13:50][D][voice_assistant:678]: Response URL: "http://192.168.210.124:8123/api/tts_proxy/N8NZ5AQTlsBHZM689FoQhQ.flac"
[01:13:50][D][voice_assistant:456]: State changed from AWAITING_RESPONSE to STREAMING_RESPONSE
[01:13:50][D][voice_assistant:463]: Desired state set to STREAMING_RESPONSE
[01:13:50][D][media_player:074]: 'Media Player' - Setting
[01:13:50][D][media_player:081]:   Media URL: http://192.168.210.124:8123/api/tts_proxy/N8NZ5AQTlsBHZM689FoQhQ.flac
[01:13:50][D][media_player:087]:  Announcement: yes
[01:13:50][D][speaker_media_player:408]: State changed to ANNOUNCING
[01:13:50][D][voice_assistant:598]: Event Type: 2
[01:13:50][D][voice_assistant:697]: Assist Pipeline ended
[01:13:50][D][sensor:093]: 'S3 Temperature': Sending state 21.09534 °C with 2 decimals of accuracy
[01:13:50][D][sensor:093]: 'S3 Humidity': Sending state 60.02083 % with 2 decimals of accuracy
[01:13:50][D][ring_buffer:034][ann_read]: Created ring buffer with size 1000000
[01:13:50][D][speaker_media_player.pipeline:114]: Reading FLAC file type
[01:13:50][D][speaker_media_player.pipeline:124]: Decoded audio has 1 channels, 48000 Hz sample rate, and 16 bits per sample
[01:13:50][D][ring_buffer:034]: Created ring buffer with size 9600
[01:13:50][D][speaker_mixer:312]: Starting speaker mixer
[01:13:50][D][speaker_mixer:320]: Started speaker mixer
[01:13:50][D][ring_buffer:034][speaker_task]: Created ring buffer with size 48000
[01:13:50][D][i2s_audio.speaker:117]: Starting Speaker
[01:13:50][D][i2s_audio.speaker:122]: Started Speaker
[01:13:52][D][speaker_media_player:408]: State changed to IDLE
[01:13:52][D][voice_assistant:329]: Announcement finished playing
[01:13:52][D][voice_assistant:456]: State changed from STREAMING_RESPONSE to RESPONSE_FINISHED
[01:13:52][D][voice_assistant:463]: Desired state set to RESPONSE_FINISHED
[01:13:52][D][micro_wake_word:360]: Starting wake word detection
[01:13:52][D][micro_wake_word:378]: State changed from STOPPED to STARTING
[01:13:52][D][voice_assistant:456]: State changed from RESPONSE_FINISHED to IDLE
[01:13:52][D][voice_assistant:463]: Desired state set to IDLE
[01:13:52][D][ring_buffer:034][mww]: Created ring buffer with size 3840
[01:13:52][D][micro_wake_word:261]: Inference task has started, attempting to allocate memory for buffers
[01:13:52][D][micro_wake_word:266]: Inference task is running
[01:13:52][D][micro_wake_word:378]: State changed from STARTING to DETECTING_WAKE_WORD
[01:13:52][D][i2s_audio.microphone:455]: Task has started, attempting to setup I2S audio driver
[01:13:52][D][speaker_mixer:325]: Stopping speaker mixer
WARNING esp32-s3-box-3-2e8c34 @ 192.168.210.49: Connection error occurred: [Errno 104] Connection reset by peer
INFO Processing unexpected disconnect from ESPHome API for esp32-s3-box-3-2e8c34 @ 192.168.210.49
WARNING Disconnected from API
INFO Successfully connected to esp32-s3-box-3-2e8c34 @ 192.168.210.49 in 0.009s
INFO Successful handshake with esp32-s3-box-3-2e8c34 @ 192.168.210.49 in 0.165s
[01:14:39][D][sensor:093]: 'WiFi db': Sending state -48.00000 dBm with 0 decimals of accuracy
[01:14:39][D][sensor:093]: 'WiFi Signal': Sending state 100.00000 % with 0 decimals of accuracy

I have a suspicion this is related to the new features that were just released.
https://www.esphome.io/changelog/2025.5.0.html

I also noted that there is a new i2s_audio driver although I have tested the legacy mode and the issue still replicates:
esphome/esphome#8703

I can see there is pending updates to the esphome wakewordvoiceassistants to support the new features which I suspect would resolve this:
esphome/wake-word-voice-assistants#110

It looks like a refactor of the way the voice assistant is configured as you can't assume the order of an interaction is wake word->stt->tts->done

I have been trying to handle how it ends the announcement but I haven't been able to mitigate this yet so I think the media_player and voice_assistant code may need to be updated similar to the pending update from esphome. I will wait until they release their changes.

I wanted to raise this as I expect others will start reporting it soon as it appears there is a breaking change in this release.

[chrisdunnname]

I have created a pull request that addresses this and many other fixes and enhancements and aligns the code closer with the latest HA wake word voice assistant but keeps all the good stuff.
#160

[ashimokawa]

@chrisdunnname

For me also with your changes, the box reboots with 2025.5.0
I am using the Original S3 Box though with patches to the .yaml file to make it work.

Here is the stacktrace

[14:56:48][D][micro_wake_word:378]: State changed from STARTING to DETECTING_WAKE_WORD
[14:56:48][D][i2s_audio.microphone:455]: Task has started, attempting to setup I2S audio driver
[14:56:48][D][speaker_mixer:325]: Stopping speaker mixer
[14:56:49]Guru Meditation Error: Core  0 panic'ed (LoadProhibited). Exception was unhandled.
[14:56:49]
[14:56:49]Core  0 register dump:
[14:56:49]PC      : 0x403836d5  PS      : 0x00060030  A0      : 0x82085fa9  A1      : 0x3fce2d50  
WARNING Decoded 0x403836d5: xQueueSemaphoreTake at /home/xxx/.platformio/packages/framework-espidf/components/freertos/FreeRTOS-Kernel/queue.c:1657
[14:56:49]A2      : 0x00e800dc  A3      : 0xffffffff  A4      : 0x00060020  A5      : 0x00060023  
[14:56:49]A6      : 0xb33fffff  A7      : 0xb33fffff  A8      : 0x00000001  A9      : 0x00000000  
[14:56:49]A10     : 0x00060023  A11     : 0x00000000  A12     : 0xb33fffff  A13     : 0x00060023  
[14:56:49]A14     : 0xb33fffff  A15     : 0xb33fffff  SAR     : 0x00000000  EXCCAUSE: 0x0000001c  
[14:56:49]EXCVADDR: 0x00e8011c  LBEG    : 0x00000000  LEND    : 0x00000000  LCOUNT  : 0x00000000  
[14:56:49]
[14:56:49]
[14:56:49]Backtrace: 0x403836d2:0x3fce2d50 0x42085fa6:0x3fce2d90 0x42012784:0x3fce2dc0 0x42012ac3:0x3fce2de0
WARNING Found stack trace! Trying to decode it
WARNING Decoded 0x403836d2: xQueueSemaphoreTake at /home/xxxi/.platformio/packages/framework-espidf/components/freertos/FreeRTOS-Kernel/queue.c:1657 (discriminator 1)
WARNING Decoded 0x42085fa6: i2s_channel_disable at /home/xxxi/.platformio/packages/framework-espidf/components/driver/i2s/i2s_common.c:1033 (discriminator 2)
WARNING Decoded 0x42012784: esphome::i2s_audio::I2SAudioMicrophone::stop_driver_() at /home/xxx/esp/esphome.s3box/.esphome/build/esp32-s3box/src/esphome/components/i2s_audio/microphone/i2s_audio_microphone.cpp:332
WARNING Decoded 0x42012ac3: esphome::i2s_audio::I2SAudioMicrophone::mic_task(void*) at /home/xxx/esp/esphome.s3box/.esphome/build/esp32-s3box/src/esphome/components/i2s_audio/microphone/i2s_audio_microphone.cpp:386

[chrisdunnname]

Well that's frustrating. I have two S3box3 both updated with TTS and va both working without crashing or getting stuck playing.

If you have the original box you could try changing the i2s_audio driver to prior version.
I think you just add use_legacy:true. It was mentioned in the pull request in the release notes.

The adaptations i made came from the latest code from the esp home wake word voice assistants earlier today. You could test their new release and see if it crashes.

[ashimokawa]

@chrisdunnname
I tried, changing to use_legacy: true. it still reboots

[aalaei]

I face the same problem (with or without legacy deriver) here is the stack trace:

0x4037bf2e: panic_abort at /config/.esphome/platformio/packages/framework-espidf/components/esp_system/panic.c:466
0x40382ea9: esp_system_abort at /config/.esphome/platformio/packages/framework-espidf/components/esp_system/port/esp_system_chip.c:84
0x40389e31: assert_func at /config/.esphome/platformio/packages/framework-espidf/components/newlib/assert.c:81
0x403832f3: xQueueGenericSend at /config/.esphome/platformio/packages/framework-espidf/components/freertos/FreeRTOS-Kernel/queue.c:873
0x42031cf6: esphome::Mutex::unlock() at /config/.esphome/build/esp32-s3box-3/src/esphome/core/helpers.cpp:656
0x42013cdb: esphome::i2s_audio::I2SAudioComponent::unlock() at /config/.esphome/build/esp32-s3box-3/src/esphome/components/i2s_audio/microphone/../i2s_audio.h:92
0x42013cdb: esphome::i2s_audio::I2SAudioMicrophone::stop_driver() at /config/.esphome/build/esp32-s3box-3/src/esphome/components/i2s_audio/microphone/i2s_audio_microphone.cpp:346
0x42013cdb: esphome::i2s_audio::I2SAudioMicrophone::stop_driver() at /config/.esphome/build/esp32-s3box-3/src/esphome/components/i2s_audio/microphone/i2s_audio_microphone.cpp:305
0x42013fcf: esphome::i2s_audio::I2SAudioMicrophone::mic_task(void*) at /config/.esphome/build/esp32-s3box-3/src/esphome/components/i2s_audio/microphone/i2s_audio_microphone.cpp:386
0x4037fffe: esp_cpu_wait_for_intr at /config/.esphome/platformio/packages/framework-espidf/components/esp_hw_support/cpu.c:121

[ashimokawa]

@aalaei
The original box? or the -3 ?
Did you use the config from the PR #160 ?

[aalaei]

@aalaei The original box? or the -3 ? Did you use the config from the PR #160 ?

esp32-s3-box-3 with sensor deck. Yes I used config from #160

[chrisdunnname]

I have had two s3-box-3 running with no problem on #160 (with HA 2025.5.3 and esphome 2025.5.0) since I shared it but If you are using the s3-box-3 and getting reboots you may find the updated config in #163 addresses this. It is a very significant refactor but I believe the problem with reboots may have been caused by the continued conversation behaviour and I think #163 addresses this better due to how it uses the voice phases configuration from the core repository. I don't have an original s3 box to test the other changes suggested in #161

[ashimokawa]

@chrisdunnname
Thanks, I tried your updated config with my changes on my original s3-box, and it still reboots

@aalaei
How about you, does it fix things for you?

[ashimokawa]

This seems to also reboot if just using piper as a media source to stream to the box!
No need to actually use speech to text, or wake words!

Media Sources -> Text-to-speech -> piper

[ashimokawa]

@chrisdunnname
@aalaei
For me this is fixed by decreasing buffer_duration from 500 to 100ms, like in the original firmware!

ecb6163

I also updated #161 accordingly

[aalaei]

Yes I can confirm, Decreasing buffer_duration from 500 to 100ms fixes the problem

[chrisdunnname]

Doh!! My apologies. I have my devices running at 100ms but when I merged the code back i missed that line.

I have updated #160 and #163.