fix(bookplayer): remove fades from page playback

This commit is contained in:
Adolfo Reyna
2026-08-20 19:45:17 -04:00
parent e010828e5e
commit 03e65fb7f7
+17 -52
View File
@@ -32,10 +32,6 @@
#define MAX_TITLE 128 #define MAX_TITLE 128
#define MAX_AUTHOR 128 #define MAX_AUTHOR 128
// Fade-in/out applied at the start/end of each page's narration to avoid an
// amplitude step (audible "pop") at page boundaries. In frames per channel.
#define FADE_FRAMES 4096
typedef enum { typedef enum {
STATE_IDLE, STATE_IDLE,
STATE_PLAYING, STATE_PLAYING,
@@ -93,6 +89,10 @@ typedef struct {
uint8_t stream_channels; uint8_t stream_channels;
uint8_t stream_bits; uint8_t stream_bits;
// Page navigation stops the decoder task but keeps a compatible output
// stream open for the next page.
bool keep_stream_on_stop;
TaskHandle_t playback_task_handle; TaskHandle_t playback_task_handle;
} AppCtx; } AppCtx;
@@ -116,7 +116,7 @@ static AppCtx g_ctx;
/* ─── Forward Declarations ─── */ /* ─── Forward Declarations ─── */
static void update_ui(AppCtx* ctx); static void update_ui(AppCtx* ctx);
static void wait_for_playback_task_to_exit(AppCtx* ctx); static void wait_for_playback_task_to_exit(AppCtx* ctx, bool keep_stream_open);
static void load_page(AppCtx* ctx, int page_index, bool start_audio); static void load_page(AppCtx* ctx, int page_index, bool start_audio);
static void return_to_picker(AppCtx* ctx); static void return_to_picker(AppCtx* ctx);
static void audio_playback_task(void* arg); static void audio_playback_task(void* arg);
@@ -266,20 +266,22 @@ static void handle_audio_finished(AppCtx* ctx) {
} }
/* ─── Helper to wait for playback thread to terminate safely ─── */ /* ─── Helper to wait for playback thread to terminate safely ─── */
static void wait_for_playback_task_to_exit(AppCtx* ctx) { static void wait_for_playback_task_to_exit(AppCtx* ctx, bool keep_stream_open) {
if (ctx->playback_task_handle != NULL) { if (ctx->playback_task_handle != NULL) {
ctx->keep_stream_on_stop = keep_stream_open;
ctx->state = STATE_IDLE; ctx->state = STATE_IDLE;
while (ctx->playback_task_handle != NULL) { while (ctx->playback_task_handle != NULL) {
tt_lvgl_unlock(); tt_lvgl_unlock();
vTaskDelay(pdMS_TO_TICKS(10)); vTaskDelay(pdMS_TO_TICKS(10));
tt_lvgl_lock(portMAX_DELAY); tt_lvgl_lock(portMAX_DELAY);
} }
ctx->keep_stream_on_stop = false;
} }
} }
/* ─── Load Page Content (Image, Caption, Audio path) ─── */ /* ─── Load Page Content (Image, Caption, Audio path) ─── */
static void load_page(AppCtx* ctx, int page_index, bool start_audio) { static void load_page(AppCtx* ctx, int page_index, bool start_audio) {
wait_for_playback_task_to_exit(ctx); wait_for_playback_task_to_exit(ctx, true);
if (page_index < 0 || page_index >= ctx->page_count) return; if (page_index < 0 || page_index >= ctx->page_count) return;
ctx->current_page = page_index; ctx->current_page = page_index;
@@ -351,7 +353,7 @@ static void load_page(AppCtx* ctx, int page_index, bool start_audio) {
/* ─── Return to Book Picker Screen ─── */ /* ─── Return to Book Picker Screen ─── */
static void return_to_picker(AppCtx* ctx) { static void return_to_picker(AppCtx* ctx) {
wait_for_playback_task_to_exit(ctx); wait_for_playback_task_to_exit(ctx, false);
close_stream_if_open(ctx); close_stream_if_open(ctx);
if (ctx->manifest_root) { if (ctx->manifest_root) {
@@ -419,7 +421,6 @@ static void play_mp3(AppCtx* ctx) {
bool eof = false; bool eof = false;
int sample_rate = 0; int sample_rate = 0;
int channels = 0; int channels = 0;
size_t page_frames_written = 0;
ESP_LOGI(TAG, "Starting MP3 playback via audio-stream: %s (%d bytes)", ctx->current_audio_path, file_size); ESP_LOGI(TAG, "Starting MP3 playback via audio-stream: %s (%d bytes)", ctx->current_audio_path, file_size);
@@ -474,30 +475,14 @@ static void play_mp3(AppCtx* ctx) {
channels = info.channels; channels = info.channels;
} }
// Adjust Volume and apply a fade-in/out at the start/end of the page // Adjust volume without altering the start or end of the narration.
// so the transition to the next page doesn't click.
int vol = ctx->volume; int vol = ctx->volume;
int16_t* samples_ptr = (int16_t*)ctx->pcm_buf; int16_t* samples_ptr = (int16_t*)ctx->pcm_buf;
size_t sample_count = (size_t)samples * info.channels; size_t sample_count = (size_t)samples * info.channels;
uint8_t ch = (uint8_t)info.channels;
size_t bytes_per_frame = (size_t)ch * sizeof(int16_t);
size_t frames_in_chunk = (size_t)samples;
float remaining_frames = (file_size > bytes_read_total)
? (float)(file_size - bytes_read_total) / (float)bytes_per_frame : 0.0f;
for (size_t i = 0; i < sample_count; ++i) { for (size_t i = 0; i < sample_count; ++i) {
size_t frame = page_frames_written + i / ch;
float fade_in = (frame < FADE_FRAMES) ? (float)frame / (float)FADE_FRAMES : 1.0f;
float fade_out = 1.0f;
float remaining = remaining_frames - (float)(i / ch);
if (remaining < (float)FADE_FRAMES) {
fade_out = (remaining > 0.0f) ? remaining / (float)FADE_FRAMES : 0.0f;
}
float gain = fade_in * fade_out;
int32_t scaled = (int32_t)samples_ptr[i] * vol / 100; int32_t scaled = (int32_t)samples_ptr[i] * vol / 100;
scaled = (int32_t)((float)scaled * gain);
samples_ptr[i] = (int16_t)scaled; samples_ptr[i] = (int16_t)scaled;
} }
page_frames_written += frames_in_chunk;
// Write via audio_stream (resampled to native 44100 internally) // Write via audio_stream (resampled to native 44100 internally)
size_t offset = 0; size_t offset = 0;
@@ -536,12 +521,10 @@ static void play_mp3(AppCtx* ctx) {
bool stopped_externally = (ctx->state == STATE_IDLE); bool stopped_externally = (ctx->state == STATE_IDLE);
if (stopped_externally) { if (stopped_externally && !ctx->keep_stream_on_stop) {
// User-initiated stop (prev/next, back, app close): tear the stream down. // App exit and returning to the picker release the output device.
close_stream_if_open(ctx); close_stream_if_open(ctx);
} }
// On a natural page finish we deliberately LEAVE the stream open so the next
// page reuses it, instead of closing+recreating the codec (which pops).
if (!stopped_externally) { if (!stopped_externally) {
tt_lvgl_lock(portMAX_DELAY); tt_lvgl_lock(portMAX_DELAY);
@@ -621,7 +604,6 @@ static void play_wav(AppCtx* ctx) {
ESP_LOGI(TAG, "Starting WAV via audio-stream: %s (%u Hz, %u ch)", ctx->current_audio_path, (unsigned int)header.sample_rate, (unsigned int)header.channels); ESP_LOGI(TAG, "Starting WAV via audio-stream: %s (%u Hz, %u ch)", ctx->current_audio_path, (unsigned int)header.sample_rate, (unsigned int)header.channels);
size_t total_played = 0; size_t total_played = 0;
size_t page_frames_written = 0;
bool stream_open = true; bool stream_open = true;
while (total_played < data_size && ctx->state != STATE_IDLE) { while (total_played < data_size && ctx->state != STATE_IDLE) {
@@ -643,29 +625,14 @@ static void play_wav(AppCtx* ctx) {
size_t read_bytes = fread(ctx->audio_buf, 1, to_read, file); size_t read_bytes = fread(ctx->audio_buf, 1, to_read, file);
if (read_bytes == 0) break; if (read_bytes == 0) break;
// Scaling Volume + fade-in/out at page boundaries to avoid a click. // Adjust volume without altering the start or end of the narration.
int vol = ctx->volume; int vol = ctx->volume;
int16_t* samples_ptr = (int16_t*)ctx->audio_buf; int16_t* samples_ptr = (int16_t*)ctx->audio_buf;
size_t sample_count = read_bytes / sizeof(int16_t); size_t sample_count = read_bytes / sizeof(int16_t);
uint8_t ch = (header.channels > 0) ? header.channels : 1;
size_t bytes_per_frame = (size_t)ch * sizeof(int16_t);
size_t frames_in_chunk = read_bytes / bytes_per_frame;
float remaining_frames = (data_size > total_played)
? (float)(data_size - total_played) / (float)bytes_per_frame : 0.0f;
for (size_t i = 0; i < sample_count; ++i) { for (size_t i = 0; i < sample_count; ++i) {
size_t frame = page_frames_written + i / ch;
float fade_in = (frame < FADE_FRAMES) ? (float)frame / (float)FADE_FRAMES : 1.0f;
float fade_out = 1.0f;
float remaining = remaining_frames - (float)(i / ch);
if (remaining < (float)FADE_FRAMES) {
fade_out = (remaining > 0.0f) ? remaining / (float)FADE_FRAMES : 0.0f;
}
float gain = fade_in * fade_out;
int32_t scaled = (int32_t)samples_ptr[i] * vol / 100; int32_t scaled = (int32_t)samples_ptr[i] * vol / 100;
scaled = (int32_t)((float)scaled * gain);
samples_ptr[i] = (int16_t)scaled; samples_ptr[i] = (int16_t)scaled;
} }
page_frames_written += frames_in_chunk;
// Write via audio_stream // Write via audio_stream
size_t offset = 0; size_t offset = 0;
@@ -702,12 +669,10 @@ static void play_wav(AppCtx* ctx) {
bool stopped_externally = (ctx->state == STATE_IDLE); bool stopped_externally = (ctx->state == STATE_IDLE);
if (stopped_externally) { if (stopped_externally && !ctx->keep_stream_on_stop) {
// User-initiated stop (prev/next, back, app close): tear the stream down. // App exit and returning to the picker release the output device.
close_stream_if_open(ctx); close_stream_if_open(ctx);
} }
// On a natural page finish we deliberately LEAVE the stream open so the next
// page reuses it, instead of closing+recreating the codec (which pops).
if (!stopped_externally) { if (!stopped_externally) {
tt_lvgl_lock(portMAX_DELAY); tt_lvgl_lock(portMAX_DELAY);
@@ -1063,7 +1028,7 @@ static void onShowApp(AppHandle app, void* data, lv_obj_t* parent) {
} }
static void onHideApp(AppHandle app, void* data) { static void onHideApp(AppHandle app, void* data) {
wait_for_playback_task_to_exit(&g_ctx); wait_for_playback_task_to_exit(&g_ctx, false);
close_stream_if_open(&g_ctx); close_stream_if_open(&g_ctx);
if (g_ctx.manifest_root) { if (g_ctx.manifest_root) {