diff --git a/esphome/components/audio/__init__.py b/esphome/components/audio/__init__.py index 7ecc45db5a..80fd328e48 100644 --- a/esphome/components/audio/__init__.py +++ b/esphome/components/audio/__init__.py @@ -429,3 +429,4 @@ async def to_code(config): ) if data.wav_support: cg.add_define("USE_AUDIO_WAV_SUPPORT") + add_idf_component(name="esphome/micro-wav", ref="0.2.0") diff --git a/esphome/components/audio/audio_decoder.cpp b/esphome/components/audio/audio_decoder.cpp index 156704fb86..7abd03a36e 100644 --- a/esphome/components/audio/audio_decoder.cpp +++ b/esphome/components/audio/audio_decoder.cpp @@ -79,7 +79,6 @@ esp_err_t AudioDecoder::start(AudioFileType audio_file_type) { this->flac_decoder_ = make_unique(); this->free_buffer_required_ = this->output_transfer_buffer_->capacity(); // Adjusted and reallocated after reading the header - this->decoder_buffers_internally_ = true; break; #endif #ifdef USE_AUDIO_MP3_SUPPORT @@ -87,7 +86,6 @@ esp_err_t AudioDecoder::start(AudioFileType audio_file_type) { this->mp3_decoder_ = make_unique(); this->free_buffer_required_ = this->output_transfer_buffer_->capacity(); // Adjusted and reallocated after reading the header - this->decoder_buffers_internally_ = true; break; #endif #ifdef USE_AUDIO_OPUS_SUPPORT @@ -95,16 +93,12 @@ esp_err_t AudioDecoder::start(AudioFileType audio_file_type) { this->opus_decoder_ = make_unique(); this->free_buffer_required_ = this->output_transfer_buffer_->capacity(); // Adjusted and reallocated after reading the header - this->decoder_buffers_internally_ = true; break; #endif #ifdef USE_AUDIO_WAV_SUPPORT case AudioFileType::WAV: - this->wav_decoder_ = make_unique(); - this->wav_decoder_->reset(); - - // Processing WAVs doesn't actually require a specific amount of buffer size, as it is already in PCM format. - // Thus, we don't reallocate to a minimum size. + this->wav_decoder_ = make_unique(); + // 1 KiB suffices to always make progress while avoiding excessive CPU spinning for decoding this->free_buffer_required_ = 1024; if (this->output_transfer_buffer_->capacity() < this->free_buffer_required_) { this->output_transfer_buffer_->reallocate(this->free_buffer_required_); @@ -181,10 +175,8 @@ AudioDecoderState AudioDecoder::decode(bool stop_gracefully) { // Decode more audio - // Only shift data on the first loop iteration to avoid unnecessary, slow moves - // If the decoder buffers internally, then never shift - size_t bytes_read = this->input_buffer_->fill(pdMS_TO_TICKS(READ_WRITE_TIMEOUT_MS), - first_loop_iteration && !this->decoder_buffers_internally_); + // Never shift the input buffer; every decoder buffers internally and consumes only what it processed. + size_t bytes_read = this->input_buffer_->fill(pdMS_TO_TICKS(READ_WRITE_TIMEOUT_MS), false); if (!first_loop_iteration && (this->input_buffer_->available() < bytes_processed)) { // Less data is available than what was processed in last iteration, so don't attempt to decode. @@ -401,50 +393,38 @@ FileDecoderState AudioDecoder::decode_opus_() { #ifdef USE_AUDIO_WAV_SUPPORT FileDecoderState AudioDecoder::decode_wav_() { - if (!this->audio_stream_info_.has_value()) { - // Header hasn't been processed + // microWAV's samples_decoded counts individual channel samples; e.g., for + // 16-bit stereo, 4 input bytes results in 2 samples_decoded. + size_t bytes_consumed = 0; + size_t samples_decoded = 0; - esp_audio_libs::wav_decoder::WAVDecoderResult result = - this->wav_decoder_->decode_header(this->input_buffer_->data(), this->input_buffer_->available()); + micro_wav::WAVDecoderResult result = this->wav_decoder_->decode( + this->input_buffer_->data(), this->input_buffer_->available(), this->output_transfer_buffer_->get_buffer_end(), + this->output_transfer_buffer_->free(), bytes_consumed, samples_decoded); - if (result == esp_audio_libs::wav_decoder::WAV_DECODER_SUCCESS_IN_DATA) { - this->input_buffer_->consume(this->wav_decoder_->bytes_processed()); + this->input_buffer_->consume(bytes_consumed); - this->audio_stream_info_ = audio::AudioStreamInfo( - this->wav_decoder_->bits_per_sample(), this->wav_decoder_->num_channels(), this->wav_decoder_->sample_rate()); - - this->wav_bytes_left_ = this->wav_decoder_->chunk_bytes_left(); - this->wav_has_known_end_ = (this->wav_bytes_left_ > 0); - return FileDecoderState::MORE_TO_PROCESS; - } else if (result == esp_audio_libs::wav_decoder::WAV_DECODER_WARNING_INCOMPLETE_DATA) { - // Available data didn't have the full header - return FileDecoderState::POTENTIALLY_FAILED; - } else { - return FileDecoderState::FAILED; + if (result == micro_wav::WAV_DECODER_SUCCESS) { + if (samples_decoded > 0 && this->audio_stream_info_.has_value()) { + this->output_transfer_buffer_->increase_buffer_length( + this->audio_stream_info_.value().samples_to_bytes(samples_decoded)); } + } else if (result == micro_wav::WAV_DECODER_HEADER_READY) { + // After HEADER_READY, get_bits_per_sample() returns the output bit depth + // (16 for A-law/mu-law, 32 for IEEE float, original value for PCM). + this->audio_stream_info_ = + audio::AudioStreamInfo(this->wav_decoder_->get_bits_per_sample(), this->wav_decoder_->get_channels(), + this->wav_decoder_->get_sample_rate()); + } else if (result == micro_wav::WAV_DECODER_NEED_MORE_DATA) { + return FileDecoderState::MORE_TO_PROCESS; + } else if (result == micro_wav::WAV_DECODER_END_OF_STREAM) { + return FileDecoderState::END_OF_FILE; } else { - if (!this->wav_has_known_end_ || (this->wav_bytes_left_ > 0)) { - size_t bytes_to_copy = this->input_buffer_->available(); - - if (this->wav_has_known_end_) { - bytes_to_copy = std::min(bytes_to_copy, this->wav_bytes_left_); - } - - bytes_to_copy = std::min(bytes_to_copy, this->output_transfer_buffer_->free()); - - if (bytes_to_copy > 0) { - std::memcpy(this->output_transfer_buffer_->get_buffer_end(), this->input_buffer_->data(), bytes_to_copy); - this->input_buffer_->consume(bytes_to_copy); - this->output_transfer_buffer_->increase_buffer_length(bytes_to_copy); - if (this->wav_has_known_end_) { - this->wav_bytes_left_ -= bytes_to_copy; - } - } - return FileDecoderState::IDLE; - } + ESP_LOGE(TAG, "WAV decoder failed: %d", static_cast(result)); + return FileDecoderState::FAILED; } - return FileDecoderState::END_OF_FILE; + return FileDecoderState::MORE_TO_PROCESS; } #endif diff --git a/esphome/components/audio/audio_decoder.h b/esphome/components/audio/audio_decoder.h index 8769e0b38b..58e982317c 100644 --- a/esphome/components/audio/audio_decoder.h +++ b/esphome/components/audio/audio_decoder.h @@ -30,9 +30,9 @@ #include #endif -// esp-audio-libs +// micro-wav #ifdef USE_AUDIO_WAV_SUPPORT -#include +#include #endif namespace esphome { @@ -135,7 +135,7 @@ class AudioDecoder { #endif #ifdef USE_AUDIO_WAV_SUPPORT FileDecoderState decode_wav_(); - std::unique_ptr wav_decoder_; + std::unique_ptr wav_decoder_; #endif std::unique_ptr input_buffer_; @@ -146,20 +146,12 @@ class AudioDecoder { size_t input_buffer_size_{0}; size_t free_buffer_required_{0}; -#ifdef USE_AUDIO_WAV_SUPPORT - size_t wav_bytes_left_{0}; -#endif uint32_t potentially_failed_count_{0}; uint32_t accumulated_frames_written_{0}; uint32_t playback_ms_{0}; bool end_of_file_{false}; -#ifdef USE_AUDIO_WAV_SUPPORT - bool wav_has_known_end_{false}; -#endif - - bool decoder_buffers_internally_{false}; bool pause_output_{false}; }; diff --git a/esphome/idf_component.yml b/esphome/idf_component.yml index e8cad29439..14fb11ace5 100644 --- a/esphome/idf_component.yml +++ b/esphome/idf_component.yml @@ -13,6 +13,8 @@ dependencies: version: 0.2.0 esphome/micro-opus: version: 0.4.1 + esphome/micro-wav: + version: 0.2.0 espressif/esp-dsp: version: "1.7.1" espressif/esp-tflite-micro: