#include "voice_recorder.h" #include "audio_compressor.h" #include "instance_lite.h" #include "../../../lib/opus_codec.h" #include "../../../lib/debug_config.h" #include "../../../lib/mem.h" #include "../../../lib/platform_compat.h" #include "../../../lib/u_async.h" #include "../../../src/chat/chat_core.h" #include "../../../src/media_async/attachment_send.h" #include #include #include #include #include #include #include #include #include #define FRAME_MS 20 struct voice_recorder { pthread_mutex_t mtx; int active; char channel_id[64]; uint8_t reply_to[32]; uint64_t request_id; int sample_rate; int channels; int frame_samples; int16_t* pcm_buffer; size_t pcm_count; size_t pcm_cap; struct audio_compressor* compressor; int compressor_enabled; char db_path[512]; int preset; float peak_level; int64_t start_time_ms; }; static struct voice_recorder* g_rec = NULL; static pthread_mutex_t g_init_mtx = PTHREAD_MUTEX_INITIALIZER; static int64_t now_ms(void) { struct timespec ts; clock_gettime(CLOCK_MONOTONIC, &ts); return ts.tv_sec * 1000LL + ts.tv_nsec / 1000000LL; } int voice_recorder_init(const char* db_path) { pthread_mutex_lock(&g_init_mtx); if (g_rec) { pthread_mutex_unlock(&g_init_mtx); return 0; } g_rec = u_calloc(1, sizeof(*g_rec)); if (!g_rec) { DEBUG_ERROR(DEBUG_CATEGORY_GENERAL, "voice_recorder_init: OOM"); pthread_mutex_unlock(&g_init_mtx); return -1; } pthread_mutex_init(&g_rec->mtx, NULL); snprintf(g_rec->db_path, sizeof(g_rec->db_path), "%s", db_path ? db_path : ""); g_rec->preset = 1; g_rec->compressor_enabled = 1; g_rec->compressor = audio_compressor_create(); if (g_rec->compressor) { audio_compressor_config_t cfg = {0}; cfg.sample_rate = 48000; cfg.channels = 1; cfg.block_duration_ms = 20; cfg.lookback_ms = 200; cfg.lookahead_ms = 100; cfg.max_gain_db = 15.0f; cfg.rise_rate_per_sec = 6.0f; cfg.target_level = powf(10.0f, -3.0f / 20.0f); /* -3 dBFS — едино с Android UI */ audio_compressor_configure(g_rec->compressor, &cfg); } DEBUG_INFO(DEBUG_CATEGORY_GENERAL, "voice_recorder_init: db=%s preset=%d compressor=%d", g_rec->db_path, g_rec->preset, g_rec->compressor_enabled); debug_set_category_level(DEBUG_CATEGORY_GENERAL, DEBUG_LEVEL_INFO); pthread_mutex_unlock(&g_init_mtx); return 0; } void voice_recorder_deinit(void) { pthread_mutex_lock(&g_init_mtx); if (!g_rec) { pthread_mutex_unlock(&g_init_mtx); return; } if (g_rec->active) { u_free(g_rec->pcm_buffer); g_rec->pcm_buffer = NULL; g_rec->pcm_count = 0; g_rec->pcm_cap = 0; g_rec->active = 0; } audio_compressor_destroy(g_rec->compressor); pthread_mutex_destroy(&g_rec->mtx); u_free(g_rec); g_rec = NULL; DEBUG_INFO(DEBUG_CATEGORY_GENERAL, "voice_recorder_deinit"); pthread_mutex_unlock(&g_init_mtx); } int voice_recorder_start(const char* channel_id, int sample_rate, int channels, const uint8_t reply_to[32], uint64_t request_id) { if (!channel_id || !channel_id[0] || strlen(channel_id) >= 64 || sample_rate != 48000 || channels != 1) { DEBUG_ERROR(DEBUG_CATEGORY_MEDIA, "voice start: invalid target or capture format rate=%d channels=%d", sample_rate, channels); return -1; } pthread_mutex_lock(&g_init_mtx); if (!g_rec) { DEBUG_ERROR(DEBUG_CATEGORY_GENERAL, "voice_recorder_start: not initialized (g_rec=NULL)"); pthread_mutex_unlock(&g_init_mtx); return -1; } pthread_mutex_lock(&g_rec->mtx); if (g_rec->active) { DEBUG_WARN(DEBUG_CATEGORY_GENERAL, "voice_recorder_start: already active"); pthread_mutex_unlock(&g_rec->mtx); pthread_mutex_unlock(&g_init_mtx); return -1; } g_rec->sample_rate = sample_rate; g_rec->channels = channels; g_rec->frame_samples = sample_rate * FRAME_MS / 1000; snprintf(g_rec->channel_id, sizeof(g_rec->channel_id), "%s", channel_id ? channel_id : ""); memcpy(g_rec->reply_to, reply_to, sizeof(g_rec->reply_to)); g_rec->request_id = request_id; g_rec->pcm_count = 0; g_rec->active = 1; struct timespec ts; clock_gettime(CLOCK_MONOTONIC, &ts); g_rec->start_time_ms = ts.tv_sec * 1000LL + ts.tv_nsec / 1000000LL; if (g_rec->compressor && g_rec->compressor_enabled) { audio_compressor_reset(g_rec->compressor); audio_compressor_set_enabled(g_rec->compressor, 1); } DEBUG_INFO(DEBUG_CATEGORY_GENERAL, "voice_recorder_start: ch=%s rate=%d chans=%d frame=%d", g_rec->channel_id, g_rec->sample_rate, g_rec->channels, g_rec->frame_samples); pthread_mutex_unlock(&g_rec->mtx); pthread_mutex_unlock(&g_init_mtx); return 0; } int voice_recorder_feed(const int16_t* samples, int count) { if (!samples || count <= 0) return -1; pthread_mutex_lock(&g_init_mtx); if (!g_rec) { pthread_mutex_unlock(&g_init_mtx); return -1; } pthread_mutex_lock(&g_rec->mtx); if (!g_rec->active) { pthread_mutex_unlock(&g_rec->mtx); pthread_mutex_unlock(&g_init_mtx); return -1; } /* compute peak level for UI */ { float sum_sq = 0.0f; float peak = 0.0f; for (int i = 0; i < count; i++) { float v = (float)samples[i] / 32768.0f; sum_sq += v * v; float av = fabsf(v); if (av > peak) peak = av; } g_rec->peak_level = (sqrtf(sum_sq / (float)count) * 2.0f + peak) / 2.0f; } if (g_rec->compressor && g_rec->compressor_enabled) { audio_compressor_push(g_rec->compressor, samples, (size_t)count); } else { size_t new_cnt = g_rec->pcm_count + (size_t)count; if (new_cnt > g_rec->pcm_cap) { size_t new_cap = g_rec->pcm_cap ? g_rec->pcm_cap * 2 : (size_t)count * 2; while (new_cap < new_cnt) new_cap *= 2; int16_t* tmp = u_realloc(g_rec->pcm_buffer, new_cap * sizeof(int16_t)); if (!tmp) { pthread_mutex_unlock(&g_rec->mtx); pthread_mutex_unlock(&g_init_mtx); return -1; } g_rec->pcm_buffer = tmp; g_rec->pcm_cap = new_cap; } memcpy(g_rec->pcm_buffer + g_rec->pcm_count, samples, (size_t)count * sizeof(int16_t)); g_rec->pcm_count = new_cnt; } pthread_mutex_unlock(&g_rec->mtx); pthread_mutex_unlock(&g_init_mtx); return 0; } static void voice_cleanup_locked(struct voice_recorder* rec) { u_free(rec->pcm_buffer); rec->pcm_buffer = NULL; rec->pcm_count = 0; rec->pcm_cap = 0; rec->active = 0; } /* Захват отдаёт собственные PCM/compressor; кодирование и отправка — в worker/uasync. */ int voice_recorder_stop(int* out_duration_ms) { pthread_mutex_lock(&g_init_mtx); if (!g_rec) { DEBUG_WARN(DEBUG_CATEGORY_MEDIA, "voice stop: recorder missing"); pthread_mutex_unlock(&g_init_mtx); return -1; } pthread_mutex_lock(&g_rec->mtx); if (!g_rec->active) { DEBUG_WARN(DEBUG_CATEGORY_MEDIA, "voice stop: recorder inactive"); pthread_mutex_unlock(&g_rec->mtx); pthread_mutex_unlock(&g_init_mtx); return -1; } int duration = (int)(now_ms() - g_rec->start_time_ms); if (out_duration_ms) *out_duration_ms = duration; struct UASYNC* ua = instance_lite_get_uasync(); struct attachment_send_req* req = u_calloc(1, sizeof(*req)); struct posted_task* post = u_calloc(1, sizeof(*post)); if (!uasync_is_running(ua) || !req || !post || duration < 500) { DEBUG_WARN(DEBUG_CATEGORY_MEDIA, "voice stop: discarded duration=%d core=%d allocated=%d", duration, uasync_is_running(ua), !!req && !!post); u_free(req); u_free(post); voice_cleanup_locked(g_rec); pthread_mutex_unlock(&g_rec->mtx); pthread_mutex_unlock(&g_init_mtx); return duration < 500 ? 0 : -1; } req->inst = instance_lite_get_instance(); req->info.kind = ATTACHMENT_VOICE; req->preset = g_rec->preset; req->is_dm = !strncmp(g_rec->channel_id, "dm:", 3); if (!req->is_dm) { memcpy(req->reply_to, g_rec->reply_to, sizeof(req->reply_to)); req->request_id = g_rec->request_id; } snprintf(req->target, sizeof(req->target), "%s", g_rec->channel_id + (req->is_dm ? 3 : 0)); if (g_rec->compressor && g_rec->compressor_enabled) { struct audio_compressor* replacement = audio_compressor_clone_config(g_rec->compressor); if (!replacement) { DEBUG_ERROR(DEBUG_CATEGORY_MEDIA, "voice stop: replacement compressor allocation failed"); u_free(req); u_free(post); voice_cleanup_locked(g_rec); pthread_mutex_unlock(&g_rec->mtx); pthread_mutex_unlock(&g_init_mtx); return -1; } req->compressor = g_rec->compressor; g_rec->compressor = replacement; } else { req->pcm = g_rec->pcm_buffer; req->pcm_count = g_rec->pcm_count; g_rec->pcm_buffer = NULL; } voice_cleanup_locked(g_rec); pthread_mutex_unlock(&g_rec->mtx); pthread_mutex_unlock(&g_init_mtx); DEBUG_INFO(DEBUG_CATEGORY_MEDIA, "voice stop: detached target=%s pm=%d duration=%d", req->target, req->is_dm, duration); post->callback = attachment_send_trampoline; post->arg = req; uasync_post_reserved(ua, post); return 0; } void voice_recorder_cancel(void) { pthread_mutex_lock(&g_init_mtx); if (!g_rec) { pthread_mutex_unlock(&g_init_mtx); return; } pthread_mutex_lock(&g_rec->mtx); if (!g_rec->active) { pthread_mutex_unlock(&g_rec->mtx); pthread_mutex_unlock(&g_init_mtx); return; } DEBUG_INFO(DEBUG_CATEGORY_GENERAL, "voice_recorder_cancel"); voice_cleanup_locked(g_rec); pthread_mutex_unlock(&g_rec->mtx); pthread_mutex_unlock(&g_init_mtx); } int voice_recorder_is_active(void) { int active = 0; pthread_mutex_lock(&g_init_mtx); if (g_rec) { pthread_mutex_lock(&g_rec->mtx); active = g_rec->active; pthread_mutex_unlock(&g_rec->mtx); } pthread_mutex_unlock(&g_init_mtx); return active; } void voice_recorder_set_preset(int preset) { if (preset < 0) preset = 0; if (preset > 2) preset = 2; pthread_mutex_lock(&g_init_mtx); if (g_rec) { pthread_mutex_lock(&g_rec->mtx); g_rec->preset = preset; DEBUG_INFO(DEBUG_CATEGORY_GENERAL, "voice_recorder_set_preset: %d", preset); pthread_mutex_unlock(&g_rec->mtx); } pthread_mutex_unlock(&g_init_mtx); } int voice_recorder_get_preset(void) { int p = 1; pthread_mutex_lock(&g_init_mtx); if (g_rec) { pthread_mutex_lock(&g_rec->mtx); p = g_rec->preset; pthread_mutex_unlock(&g_rec->mtx); } pthread_mutex_unlock(&g_init_mtx); return p; } void voice_recorder_set_compressor_enabled(int enabled) { pthread_mutex_lock(&g_init_mtx); if (g_rec) { pthread_mutex_lock(&g_rec->mtx); g_rec->compressor_enabled = enabled ? 1 : 0; DEBUG_INFO(DEBUG_CATEGORY_GENERAL, "voice_recorder_set_compressor: %s", g_rec->compressor_enabled ? "on" : "off"); pthread_mutex_unlock(&g_rec->mtx); } pthread_mutex_unlock(&g_init_mtx); } int voice_recorder_is_compressor_enabled(void) { int e = 1; pthread_mutex_lock(&g_init_mtx); if (g_rec) { pthread_mutex_lock(&g_rec->mtx); e = g_rec->compressor_enabled; pthread_mutex_unlock(&g_rec->mtx); } pthread_mutex_unlock(&g_init_mtx); return e; } void voice_recorder_set_compressor_config(int max_gain_db, int lookback_ms, int lookahead_ms, float rise_rate_per_sec, float target_level_db) { pthread_mutex_lock(&g_init_mtx); if (g_rec && g_rec->compressor) { pthread_mutex_lock(&g_rec->mtx); audio_compressor_config_t cfg = {0}; cfg.sample_rate = 48000; cfg.channels = 1; cfg.block_duration_ms = 20; cfg.lookback_ms = lookback_ms; cfg.lookahead_ms = lookahead_ms; cfg.max_gain_db = (float)max_gain_db; cfg.rise_rate_per_sec = rise_rate_per_sec; cfg.target_level = powf(10.0f, target_level_db / 20.0f); audio_compressor_configure(g_rec->compressor, &cfg); DEBUG_INFO(DEBUG_CATEGORY_GENERAL, "voice_recorder compressor config: gain=%ddB lookback=%d lookahead=%d rise=%.1f target=%.0fdBFS", max_gain_db, lookback_ms, lookahead_ms, rise_rate_per_sec, target_level_db); pthread_mutex_unlock(&g_rec->mtx); } pthread_mutex_unlock(&g_init_mtx); } float voice_recorder_get_peak_level(void) { float v = 0.0f; pthread_mutex_lock(&g_init_mtx); if (g_rec) { pthread_mutex_lock(&g_rec->mtx); v = g_rec->peak_level; pthread_mutex_unlock(&g_rec->mtx); } pthread_mutex_unlock(&g_init_mtx); return v; }