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

adding min sub duration

Browse Source
This commit is contained in:
Roy Shilkrot 2023-10-29 00:26:36 -04:00
parent a39979e9d7
commit 465193a12b
5 changed files with 31 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
data/locale/en-US.ini
View File

@ -41,3 +41,4 @@ save_srt="Save in SRT format (no file truncation)"
only_while_recording="Write output only while recording"
process_while_muted="Process speech while source is muted"
rename_file_to_match_recording="Rename file to match recording"
min_sub_duration="Minimal subtitle duration (msec)"

1
data/locale/pt_BR.ini
View File

@ -42,3 +42,4 @@ save_srt="Salvar no formato SRT"
only_while_recording="Escreva durante a gravação"
process_while_muted="Processar enquanto está silenciada"
rename_file_to_match_recording="Renomear arquivo para corresponder à gravação"
min_sub_duration="Duração mínima da legenda (msec)"

1
data/locale/ru_RU.ini
View File

@ -41,3 +41,4 @@ save_srt="Сохранить в формате SRT"
only_while_recording="Записывать вывод только во время записи"
process_while_muted="Обрабатывать речь, пока источник отключен"
rename_file_to_match_recording="Переименовать файл, чтобы соответствовать записи"
min_sub_duration="Минимальная длительность субтитров (мс)"

4
src/transcription-filter-data.h
View File

@ -49,6 +49,10 @@ struct transcription_filter_data {
uint64_t start_timestamp_ms;
// Sentence counter for srt
size_t sentence_number;
// Minimal subtitle duration in ms
size_t min_sub_duration;
// Last time a subtitle was rendered
uint64_t last_sub_render_time;
/* PCM buffers */
float *copy_buffers[MAX_PREPROC_CHANNELS];

26
src/transcription-filter.cpp
View File

@ -193,8 +193,23 @@ void acquire_weak_text_source_ref(struct transcription_filter_data *gf)
}
}
void set_text_callback(struct transcription_filter_data *gf, const DetectionResultWithText &result)
void set_text_callback(struct transcription_filter_data *gf,
const DetectionResultWithText &resultIn)
{
DetectionResultWithText result = resultIn;
uint64_t now = now_ms();
if (result.text.empty() || result.result != DETECTION_RESULT_SPEECH) {
// check if we should clear the current sub depending on the minimum subtitle duration
if ((now - gf->last_sub_render_time) > gf->min_sub_duration) {
// clear the current sub, run an empty sub
result.text = "";
} else {
// nothing to do, the incoming sub is empty
return;
}
}
gf->last_sub_render_time = now;
#ifdef _WIN32
// Some UTF8 charsets on Windows output have a bug, instead of 0xd? it outputs
// 0xf?, and 0xc? becomes 0xe?, so we need to replace it.
@ -356,6 +371,8 @@ void transcription_filter_update(void *data, obs_data_t *s)
gf->start_timestamp_ms = now_ms();
gf->sentence_number = 1;
gf->process_while_muted = obs_data_get_bool(s, "process_while_muted");
gf->min_sub_duration = (int)obs_data_get_int(s, "min_sub_duration");
gf->last_sub_render_time = 0;
obs_log(gf->log_level, "transcription_filter: update text source");
// update the text source
@ -475,7 +492,7 @@ void transcription_filter_update(void *data, obs_data_t *s)
}
} else {
// model path did not change
obs_log(LOG_INFO, "model path did not change: %s == %s", gf->whisper_model_path,
obs_log(LOG_DEBUG, "model path did not change: %s == %s", gf->whisper_model_path,
new_model_path.c_str());
}
@ -531,6 +548,8 @@ void *transcription_filter_create(obs_data_t *settings, obs_source_t *filter)
gf->step_size_msec = step_by_step_processing
? (int)obs_data_get_int(settings, "step_size_msec")
: BUFFER_SIZE_MSEC;
gf->min_sub_duration = (int)obs_data_get_int(settings, "min_sub_duration");
gf->last_sub_render_time = 0;
gf->log_level = (int)obs_data_get_int(settings, "log_level");
gf->save_srt = obs_data_get_bool(settings, "subtitle_save_srt");
gf->save_only_while_recording = obs_data_get_bool(settings, "only_while_recording");
@ -673,6 +692,7 @@ void transcription_filter_defaults(obs_data_t *s)
obs_data_set_default_bool(s, "only_while_recording", false);
obs_data_set_default_bool(s, "rename_file_to_match_recording", true);
obs_data_set_default_int(s, "step_size_msec", 1000);
obs_data_set_default_int(s, "min_sub_duration", 3000);
// Whisper parameters
obs_data_set_default_int(s, "whisper_sampling_method", WHISPER_SAMPLING_BEAM_SEARCH);
@ -721,6 +741,8 @@ obs_properties_t *transcription_filter_properties(void *data)
ppts, "step_by_step_processing", MT_("step_by_step_processing"));
obs_properties_add_int_slider(ppts, "step_size_msec", MT_("step_size_msec"), 1000,
BUFFER_SIZE_MSEC, 50);
obs_properties_add_int_slider(ppts, "min_sub_duration", MT_("min_sub_duration"), 1000, 5000,
50);
obs_property_set_modified_callback(step_by_step_processing, [](obs_properties_t *props,
obs_property_t *property,