mirrors/obs-localvocal
1
0
Fork 0
mirror of https://github.com/occ-ai/obs-localvocal synced 2024-11-07 18:57:14 +00:00
Code Issues Packages Projects Releases Wiki Activity

remove unneeded stuff

Browse Source
This commit is contained in:
Roy Shilkrot 2023-08-13 16:31:40 +03:00
parent e246e80448
commit 86e719150d
3 changed files with 14 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
src/transcription-filter.cpp
View File

@ -6,7 +6,6 @@
#include "whisper-processing.h" #include "whisper-processing.h"
#include "whisper-language.h" #include "whisper-language.h"
inline enum speaker_layout convert_speaker_layout(uint8_t channels) inline enum speaker_layout convert_speaker_layout(uint8_t channels)
{ {
switch (channels) { switch (channels) {
@ -167,9 +166,7 @@ void transcription_filter_update(void *data, obs_data_t *s)
struct transcription_filter_data *gf = struct transcription_filter_data *gf =
static_cast<struct transcription_filter_data *>(data); static_cast<struct transcription_filter_data *>(data);
gf->filler_p_threshold = (float)obs_data_get_double(s, "filler_p_threshold");
gf->log_level = (int)obs_data_get_int(s, "log_level"); gf->log_level = (int)obs_data_get_int(s, "log_level");
gf->do_silence = obs_data_get_bool(s, "do_silence");
gf->vad_enabled = obs_data_get_bool(s, "vad_enabled"); gf->vad_enabled = obs_data_get_bool(s, "vad_enabled");
gf->log_words = obs_data_get_bool(s, "log_words"); gf->log_words = obs_data_get_bool(s, "log_words");
@ -381,8 +378,6 @@ void transcription_filter_deactivate(void *data)
void transcription_filter_defaults(obs_data_t *s) void transcription_filter_defaults(obs_data_t *s)
{ {
obs_data_set_default_double(s, "filler_p_threshold", 0.75);
obs_data_set_default_bool(s, "do_silence", true);
obs_data_set_default_bool(s, "vad_enabled", true); obs_data_set_default_bool(s, "vad_enabled", true);
obs_data_set_default_int(s, "log_level", LOG_DEBUG); obs_data_set_default_int(s, "log_level", LOG_DEBUG);
obs_data_set_default_bool(s, "log_words", true); obs_data_set_default_bool(s, "log_words", true);
@ -419,16 +414,13 @@ obs_properties_t *transcription_filter_properties(void *data)
{ {
obs_properties_t *ppts = obs_properties_create(); obs_properties_t *ppts = obs_properties_create();
obs_properties_add_float_slider(ppts, "filler_p_threshold", "filler_p_threshold", 0.0f, obs_properties_add_bool(ppts, "vad_enabled", "VAD Enabled");
1.0f, 0.05f); obs_property_t *list = obs_properties_add_list(ppts, "log_level", "Log level",
obs_properties_add_bool(ppts, "do_silence", "do_silence");
obs_properties_add_bool(ppts, "vad_enabled", "vad_enabled");
obs_property_t *list = obs_properties_add_list(ppts, "log_level", "log_level",
OBS_COMBO_TYPE_LIST, OBS_COMBO_FORMAT_INT); OBS_COMBO_TYPE_LIST, OBS_COMBO_FORMAT_INT);
obs_property_list_add_int(list, "DEBUG", LOG_DEBUG); obs_property_list_add_int(list, "DEBUG", LOG_DEBUG);
obs_property_list_add_int(list, "INFO", LOG_INFO); obs_property_list_add_int(list, "INFO", LOG_INFO);
obs_property_list_add_int(list, "WARNING", LOG_WARNING); obs_property_list_add_int(list, "WARNING", LOG_WARNING);
obs_properties_add_bool(ppts, "log_words", "log_words"); obs_properties_add_bool(ppts, "log_words", "Log output words");
obs_property_t *sources = obs_properties_add_list(ppts, "subtitle_sources", obs_property_t *sources = obs_properties_add_list(ppts, "subtitle_sources",
"subtitle_sources", OBS_COMBO_TYPE_LIST, "subtitle_sources", OBS_COMBO_TYPE_LIST,

16
src/whisper-processing.cpp
View File

@ -4,6 +4,7 @@
#include "plugin-support.h" #include "plugin-support.h"
#include "transcription-filter-data.h" #include "transcription-filter-data.h"
#include "whisper-processing.h"
#include <algorithm> #include <algorithm>
#include <cctype> #include <cctype>
@ -286,12 +287,15 @@ void process_audio_from_buffer(struct transcription_filter_data *gf)
"audio processing took too long (%d ms), reducing overlap to %lu ms", "audio processing took too long (%d ms), reducing overlap to %lu ms",
(int)duration, gf->overlap_ms); (int)duration, gf->overlap_ms);
} else if (!skipped_inference) { } else if (!skipped_inference) {
// try to increase overlap up to 75% of the segment if (gf->overlap_ms < OVERLAP_SIZE_MSEC) {
gf->overlap_ms = std::min((uint64_t)gf->overlap_ms + 10, // try to increase overlap up to OVERLAP_SIZE_MSEC
(uint64_t)((float)new_frames_from_infos_ms * 0.75f)); gf->overlap_ms = std::min((uint64_t)gf->overlap_ms + 10,
gf->overlap_frames = gf->overlap_ms * gf->sample_rate / 1000; (uint64_t)OVERLAP_SIZE_MSEC);
obs_log(gf->log_level, "audio processing took %d ms, increasing overlap to %lu ms", gf->overlap_frames = gf->overlap_ms * gf->sample_rate / 1000;
(int)duration, gf->overlap_ms); obs_log(gf->log_level,
"audio processing took %d ms, increasing overlap to %lu ms",
(int)duration, gf->overlap_ms);
}
} }
} }

2
src/whisper-processing.h
View File

@ -6,7 +6,7 @@
// at 16Khz, 3000 msec is 48000 samples // at 16Khz, 3000 msec is 48000 samples
#define WHISPER_FRAME_SIZE 48000 #define WHISPER_FRAME_SIZE 48000
// overlap in msec // overlap in msec
#define OVERLAP_SIZE_MSEC 340 #define OVERLAP_SIZE_MSEC 200
void whisper_loop(void *data); void whisper_loop(void *data);
struct whisper_context *init_whisper_context(const std::string &model_path); struct whisper_context *init_whisper_context(const std::string &model_path);