@@ -26,7 +26,6 @@ struct whisper_params {
26
26
float vad_thold = 0 .6f ;
27
27
float freq_thold = 100 .0f ;
28
28
29
- bool speed_up = false ;
30
29
bool translate = false ;
31
30
bool print_special = false ;
32
31
bool print_energy = false ;
@@ -70,7 +69,6 @@ bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
70
69
else if (arg == " -ac" || arg == " --audio-ctx" ) { params.audio_ctx = std::stoi (argv[++i]); }
71
70
else if (arg == " -vth" || arg == " --vad-thold" ) { params.vad_thold = std::stof (argv[++i]); }
72
71
else if (arg == " -fth" || arg == " --freq-thold" ) { params.freq_thold = std::stof (argv[++i]); }
73
- else if (arg == " -su" || arg == " --speed-up" ) { params.speed_up = true ; }
74
72
else if (arg == " -tr" || arg == " --translate" ) { params.translate = true ; }
75
73
else if (arg == " -ps" || arg == " --print-special" ) { params.print_special = true ; }
76
74
else if (arg == " -pe" || arg == " --print-energy" ) { params.print_energy = true ; }
@@ -102,7 +100,6 @@ void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params & para
102
100
fprintf (stderr, " -ac N, --audio-ctx N [%-7d] audio context size (0 - all)\n " , params.audio_ctx );
103
101
fprintf (stderr, " -vth N, --vad-thold N [%-7.2f] voice activity detection threshold\n " , params.vad_thold );
104
102
fprintf (stderr, " -fth N, --freq-thold N [%-7.2f] high-pass frequency cutoff\n " , params.freq_thold );
105
- fprintf (stderr, " -su, --speed-up [%-7s] speed up audio by x2 (reduced accuracy)\n " , params.speed_up ? " true" : " false" );
106
103
fprintf (stderr, " -tr, --translate [%-7s] translate from source language to english\n " , params.translate ? " true" : " false" );
107
104
fprintf (stderr, " -ps, --print-special [%-7s] print special tokens\n " , params.print_special ? " true" : " false" );
108
105
fprintf (stderr, " -pe, --print-energy [%-7s] print sound energy (for debugging)\n " , params.print_energy ? " true" : " false" );
@@ -184,7 +181,6 @@ json unguided_transcription(struct whisper_context * ctx, audio_async &audio, js
184
181
wparams.n_threads = params.n_threads ;
185
182
186
183
wparams.audio_ctx = params.audio_ctx ;
187
- wparams.speed_up = params.speed_up ;
188
184
wparams.suppress_non_speech_tokens = true ;
189
185
// run the transformer and a single decoding pass
190
186
if (whisper_full (ctx, wparams, pcmf32.data (), pcmf32.size ()) != 0 ) {
@@ -223,7 +219,6 @@ json guided_transcription(struct whisper_context * ctx, audio_async &audio, cons
223
219
wparams.n_threads = params.n_threads ;
224
220
225
221
wparams.audio_ctx = params.audio_ctx ;
226
- wparams.speed_up = params.speed_up ;
227
222
228
223
// TODO: Do some time testing. Does an overly long prompt slow down processing?
229
224
// Set up command sets/precompute prompts
0 commit comments