-
Notifications
You must be signed in to change notification settings - Fork 961
/
main.cpp
1240 lines (1021 loc) · 52.6 KB
/
main.cpp
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
#include "common.h"
#include "whisper.h"
#include "grammar-parser.h"
#include <cmath>
#include <fstream>
#include <cstdio>
#include <regex>
#include <string>
#include <thread>
#include <vector>
#include <cstring>
#if defined(_MSC_VER)
#pragma warning(disable: 4244 4267) // possible loss of data
#endif
// helper function to replace substrings
void replace_all(std::string & s, const std::string & search, const std::string & replace) {
for (size_t pos = 0; ; pos += replace.length()) {
pos = s.find(search, pos);
if (pos == std::string::npos) break;
s.erase(pos, search.length());
s.insert(pos, replace);
}
}
// command-line parameters
struct whisper_params {
int32_t n_threads = std::min(4, (int32_t) std::thread::hardware_concurrency());
int32_t n_processors = 1;
int32_t offset_t_ms = 0;
int32_t offset_n = 0;
int32_t duration_ms = 0;
int32_t progress_step = 5;
int32_t max_context = -1;
int32_t max_len = 0;
int32_t best_of = whisper_full_default_params(WHISPER_SAMPLING_GREEDY).greedy.best_of;
int32_t beam_size = whisper_full_default_params(WHISPER_SAMPLING_BEAM_SEARCH).beam_search.beam_size;
int32_t audio_ctx = 0;
float word_thold = 0.01f;
float entropy_thold = 2.40f;
float logprob_thold = -1.00f;
float grammar_penalty = 100.0f;
float temperature = 0.0f;
float temperature_inc = 0.2f;
bool speed_up = false;
bool debug_mode = false;
bool translate = false;
bool detect_language = false;
bool diarize = false;
bool tinydiarize = false;
bool split_on_word = false;
bool no_fallback = false;
bool output_txt = false;
bool output_vtt = false;
bool output_srt = false;
bool output_wts = false;
bool output_csv = false;
bool output_jsn = false;
bool output_jsn_full = false;
bool output_lrc = false;
bool no_prints = false;
bool print_special = false;
bool print_colors = false;
bool print_progress = false;
bool no_timestamps = false;
bool log_score = false;
bool use_gpu = true;
std::string language = "en";
std::string prompt;
std::string font_path = "/System/Library/Fonts/Supplemental/Courier New Bold.ttf";
std::string model = "models/ggml-base.en.bin";
std::string grammar;
std::string grammar_rule;
// [TDRZ] speaker turn string
std::string tdrz_speaker_turn = " [SPEAKER_TURN]"; // TODO: set from command line
// A regular expression that matches tokens to suppress
std::string suppress_regex;
std::string openvino_encode_device = "CPU";
std::string dtw = "";
std::vector<std::string> fname_inp = {};
std::vector<std::string> fname_out = {};
grammar_parser::parse_state grammar_parsed;
};
void whisper_print_usage(int argc, char ** argv, const whisper_params & params);
char* whisper_param_turn_lowercase(char* in){
int string_len = strlen(in);
for(int i = 0; i < string_len; i++){
*(in+i) = tolower((unsigned char)*(in+i));
}
return in;
}
bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
for (int i = 1; i < argc; i++) {
std::string arg = argv[i];
if (arg == "-"){
params.fname_inp.push_back(arg);
continue;
}
if (arg[0] != '-') {
params.fname_inp.push_back(arg);
continue;
}
if (arg == "-h" || arg == "--help") {
whisper_print_usage(argc, argv, params);
exit(0);
}
else if (arg == "-t" || arg == "--threads") { params.n_threads = std::stoi(argv[++i]); }
else if (arg == "-p" || arg == "--processors") { params.n_processors = std::stoi(argv[++i]); }
else if (arg == "-ot" || arg == "--offset-t") { params.offset_t_ms = std::stoi(argv[++i]); }
else if (arg == "-on" || arg == "--offset-n") { params.offset_n = std::stoi(argv[++i]); }
else if (arg == "-d" || arg == "--duration") { params.duration_ms = std::stoi(argv[++i]); }
else if (arg == "-mc" || arg == "--max-context") { params.max_context = std::stoi(argv[++i]); }
else if (arg == "-ml" || arg == "--max-len") { params.max_len = std::stoi(argv[++i]); }
else if (arg == "-bo" || arg == "--best-of") { params.best_of = std::stoi(argv[++i]); }
else if (arg == "-bs" || arg == "--beam-size") { params.beam_size = std::stoi(argv[++i]); }
else if (arg == "-ac" || arg == "--audio-ctx") { params.audio_ctx = std::stoi(argv[++i]); }
else if (arg == "-wt" || arg == "--word-thold") { params.word_thold = std::stof(argv[++i]); }
else if (arg == "-et" || arg == "--entropy-thold") { params.entropy_thold = std::stof(argv[++i]); }
else if (arg == "-lpt" || arg == "--logprob-thold") { params.logprob_thold = std::stof(argv[++i]); }
else if (arg == "-tp" || arg == "--temperature") { params.temperature = std::stof(argv[++i]); }
else if (arg == "-tpi" || arg == "--temperature-inc") { params.temperature_inc = std::stof(argv[++i]); }
// else if (arg == "-su" || arg == "--speed-up") { params.speed_up = true; }
else if (arg == "-debug"|| arg == "--debug-mode") { params.debug_mode = true; }
else if (arg == "-tr" || arg == "--translate") { params.translate = true; }
else if (arg == "-di" || arg == "--diarize") { params.diarize = true; }
else if (arg == "-tdrz" || arg == "--tinydiarize") { params.tinydiarize = true; }
else if (arg == "-sow" || arg == "--split-on-word") { params.split_on_word = true; }
else if (arg == "-nf" || arg == "--no-fallback") { params.no_fallback = true; }
else if (arg == "-otxt" || arg == "--output-txt") { params.output_txt = true; }
else if (arg == "-ovtt" || arg == "--output-vtt") { params.output_vtt = true; }
else if (arg == "-osrt" || arg == "--output-srt") { params.output_srt = true; }
else if (arg == "-owts" || arg == "--output-words") { params.output_wts = true; }
else if (arg == "-olrc" || arg == "--output-lrc") { params.output_lrc = true; }
else if (arg == "-fp" || arg == "--font-path") { params.font_path = argv[++i]; }
else if (arg == "-ocsv" || arg == "--output-csv") { params.output_csv = true; }
else if (arg == "-oj" || arg == "--output-json") { params.output_jsn = true; }
else if (arg == "-ojf" || arg == "--output-json-full"){ params.output_jsn_full = params.output_jsn = true; }
else if (arg == "-of" || arg == "--output-file") { params.fname_out.emplace_back(argv[++i]); }
else if (arg == "-np" || arg == "--no-prints") { params.no_prints = true; }
else if (arg == "-ps" || arg == "--print-special") { params.print_special = true; }
else if (arg == "-pc" || arg == "--print-colors") { params.print_colors = true; }
else if (arg == "-pp" || arg == "--print-progress") { params.print_progress = true; }
else if (arg == "-nt" || arg == "--no-timestamps") { params.no_timestamps = true; }
else if (arg == "-l" || arg == "--language") { params.language = whisper_param_turn_lowercase(argv[++i]); }
else if (arg == "-dl" || arg == "--detect-language") { params.detect_language = true; }
else if ( arg == "--prompt") { params.prompt = argv[++i]; }
else if (arg == "-m" || arg == "--model") { params.model = argv[++i]; }
else if (arg == "-f" || arg == "--file") { params.fname_inp.emplace_back(argv[++i]); }
else if (arg == "-oved" || arg == "--ov-e-device") { params.openvino_encode_device = argv[++i]; }
else if (arg == "-dtw" || arg == "--dtw") { params.dtw = argv[++i]; }
else if (arg == "-ls" || arg == "--log-score") { params.log_score = true; }
else if (arg == "-ng" || arg == "--no-gpu") { params.use_gpu = false; }
else if ( arg == "--suppress-regex") { params.suppress_regex = argv[++i]; }
else if ( arg == "--grammar") { params.grammar = argv[++i]; }
else if ( arg == "--grammar-rule") { params.grammar_rule = argv[++i]; }
else if ( arg == "--grammar-penalty") { params.grammar_penalty = std::stof(argv[++i]); }
else {
fprintf(stderr, "error: unknown argument: %s\n", arg.c_str());
whisper_print_usage(argc, argv, params);
exit(0);
}
}
return true;
}
void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params & params) {
fprintf(stderr, "\n");
fprintf(stderr, "usage: %s [options] file0.wav file1.wav ...\n", argv[0]);
fprintf(stderr, "\n");
fprintf(stderr, "options:\n");
fprintf(stderr, " -h, --help [default] show this help message and exit\n");
fprintf(stderr, " -t N, --threads N [%-7d] number of threads to use during computation\n", params.n_threads);
fprintf(stderr, " -p N, --processors N [%-7d] number of processors to use during computation\n", params.n_processors);
fprintf(stderr, " -ot N, --offset-t N [%-7d] time offset in milliseconds\n", params.offset_t_ms);
fprintf(stderr, " -on N, --offset-n N [%-7d] segment index offset\n", params.offset_n);
fprintf(stderr, " -d N, --duration N [%-7d] duration of audio to process in milliseconds\n", params.duration_ms);
fprintf(stderr, " -mc N, --max-context N [%-7d] maximum number of text context tokens to store\n", params.max_context);
fprintf(stderr, " -ml N, --max-len N [%-7d] maximum segment length in characters\n", params.max_len);
fprintf(stderr, " -sow, --split-on-word [%-7s] split on word rather than on token\n", params.split_on_word ? "true" : "false");
fprintf(stderr, " -bo N, --best-of N [%-7d] number of best candidates to keep\n", params.best_of);
fprintf(stderr, " -bs N, --beam-size N [%-7d] beam size for beam search\n", params.beam_size);
fprintf(stderr, " -ac N, --audio-ctx N [%-7d] audio context size (0 - all)\n", params.audio_ctx);
fprintf(stderr, " -wt N, --word-thold N [%-7.2f] word timestamp probability threshold\n", params.word_thold);
fprintf(stderr, " -et N, --entropy-thold N [%-7.2f] entropy threshold for decoder fail\n", params.entropy_thold);
fprintf(stderr, " -lpt N, --logprob-thold N [%-7.2f] log probability threshold for decoder fail\n", params.logprob_thold);
fprintf(stderr, " -tp, --temperature N [%-7.2f] The sampling temperature, between 0 and 1\n", params.temperature);
fprintf(stderr, " -tpi, --temperature-inc N [%-7.2f] The increment of temperature, between 0 and 1\n",params.temperature_inc);
// fprintf(stderr, " -su, --speed-up [%-7s] speed up audio by x2 (reduced accuracy)\n", params.speed_up ? "true" : "false");
fprintf(stderr, " -debug, --debug-mode [%-7s] enable debug mode (eg. dump log_mel)\n", params.debug_mode ? "true" : "false");
fprintf(stderr, " -tr, --translate [%-7s] translate from source language to english\n", params.translate ? "true" : "false");
fprintf(stderr, " -di, --diarize [%-7s] stereo audio diarization\n", params.diarize ? "true" : "false");
fprintf(stderr, " -tdrz, --tinydiarize [%-7s] enable tinydiarize (requires a tdrz model)\n", params.tinydiarize ? "true" : "false");
fprintf(stderr, " -nf, --no-fallback [%-7s] do not use temperature fallback while decoding\n", params.no_fallback ? "true" : "false");
fprintf(stderr, " -otxt, --output-txt [%-7s] output result in a text file\n", params.output_txt ? "true" : "false");
fprintf(stderr, " -ovtt, --output-vtt [%-7s] output result in a vtt file\n", params.output_vtt ? "true" : "false");
fprintf(stderr, " -osrt, --output-srt [%-7s] output result in a srt file\n", params.output_srt ? "true" : "false");
fprintf(stderr, " -olrc, --output-lrc [%-7s] output result in a lrc file\n", params.output_lrc ? "true" : "false");
fprintf(stderr, " -owts, --output-words [%-7s] output script for generating karaoke video\n", params.output_wts ? "true" : "false");
fprintf(stderr, " -fp, --font-path [%-7s] path to a monospace font for karaoke video\n", params.font_path.c_str());
fprintf(stderr, " -ocsv, --output-csv [%-7s] output result in a CSV file\n", params.output_csv ? "true" : "false");
fprintf(stderr, " -oj, --output-json [%-7s] output result in a JSON file\n", params.output_jsn ? "true" : "false");
fprintf(stderr, " -ojf, --output-json-full [%-7s] include more information in the JSON file\n", params.output_jsn_full ? "true" : "false");
fprintf(stderr, " -of FNAME, --output-file FNAME [%-7s] output file path (without file extension)\n", "");
fprintf(stderr, " -np, --no-prints [%-7s] do not print anything other than the results\n", params.no_prints ? "true" : "false");
fprintf(stderr, " -ps, --print-special [%-7s] print special tokens\n", params.print_special ? "true" : "false");
fprintf(stderr, " -pc, --print-colors [%-7s] print colors\n", params.print_colors ? "true" : "false");
fprintf(stderr, " -pp, --print-progress [%-7s] print progress\n", params.print_progress ? "true" : "false");
fprintf(stderr, " -nt, --no-timestamps [%-7s] do not print timestamps\n", params.no_timestamps ? "true" : "false");
fprintf(stderr, " -l LANG, --language LANG [%-7s] spoken language ('auto' for auto-detect)\n", params.language.c_str());
fprintf(stderr, " -dl, --detect-language [%-7s] exit after automatically detecting language\n", params.detect_language ? "true" : "false");
fprintf(stderr, " --prompt PROMPT [%-7s] initial prompt (max n_text_ctx/2 tokens)\n", params.prompt.c_str());
fprintf(stderr, " -m FNAME, --model FNAME [%-7s] model path\n", params.model.c_str());
fprintf(stderr, " -f FNAME, --file FNAME [%-7s] input WAV file path\n", "");
fprintf(stderr, " -oved D, --ov-e-device DNAME [%-7s] the OpenVINO device used for encode inference\n", params.openvino_encode_device.c_str());
fprintf(stderr, " -dtw MODEL --dtw MODEL [%-7s] compute token-level timestamps\n", params.dtw.c_str());
fprintf(stderr, " -ls, --log-score [%-7s] log best decoder scores of tokens\n", params.log_score?"true":"false");
fprintf(stderr, " -ng, --no-gpu [%-7s] disable GPU\n", params.use_gpu ? "false" : "true");
fprintf(stderr, " --suppress-regex REGEX [%-7s] regular expression matching tokens to suppress\n", params.suppress_regex.c_str());
fprintf(stderr, " --grammar GRAMMAR [%-7s] GBNF grammar to guide decoding\n", params.grammar.c_str());
fprintf(stderr, " --grammar-rule RULE [%-7s] top-level GBNF grammar rule name\n", params.grammar_rule.c_str());
fprintf(stderr, " --grammar-penalty N [%-7.1f] scales down logits of nongrammar tokens\n", params.grammar_penalty);
fprintf(stderr, "\n");
}
struct whisper_print_user_data {
const whisper_params * params;
const std::vector<std::vector<float>> * pcmf32s;
int progress_prev;
};
std::string estimate_diarization_speaker(std::vector<std::vector<float>> pcmf32s, int64_t t0, int64_t t1, bool id_only = false) {
std::string speaker = "";
const int64_t n_samples = pcmf32s[0].size();
const int64_t is0 = timestamp_to_sample(t0, n_samples, WHISPER_SAMPLE_RATE);
const int64_t is1 = timestamp_to_sample(t1, n_samples, WHISPER_SAMPLE_RATE);
double energy0 = 0.0f;
double energy1 = 0.0f;
for (int64_t j = is0; j < is1; j++) {
energy0 += fabs(pcmf32s[0][j]);
energy1 += fabs(pcmf32s[1][j]);
}
if (energy0 > 1.1*energy1) {
speaker = "0";
} else if (energy1 > 1.1*energy0) {
speaker = "1";
} else {
speaker = "?";
}
//printf("is0 = %lld, is1 = %lld, energy0 = %f, energy1 = %f, speaker = %s\n", is0, is1, energy0, energy1, speaker.c_str());
if (!id_only) {
speaker.insert(0, "(speaker ");
speaker.append(")");
}
return speaker;
}
void whisper_print_progress_callback(struct whisper_context * /*ctx*/, struct whisper_state * /*state*/, int progress, void * user_data) {
int progress_step = ((whisper_print_user_data *) user_data)->params->progress_step;
int * progress_prev = &(((whisper_print_user_data *) user_data)->progress_prev);
if (progress >= *progress_prev + progress_step) {
*progress_prev += progress_step;
fprintf(stderr, "%s: progress = %3d%%\n", __func__, progress);
}
}
void whisper_print_segment_callback(struct whisper_context * ctx, struct whisper_state * /*state*/, int n_new, void * user_data) {
const auto & params = *((whisper_print_user_data *) user_data)->params;
const auto & pcmf32s = *((whisper_print_user_data *) user_data)->pcmf32s;
const int n_segments = whisper_full_n_segments(ctx);
std::string speaker = "";
int64_t t0 = 0;
int64_t t1 = 0;
// print the last n_new segments
const int s0 = n_segments - n_new;
if (s0 == 0) {
printf("\n");
}
for (int i = s0; i < n_segments; i++) {
if (!params.no_timestamps || params.diarize) {
t0 = whisper_full_get_segment_t0(ctx, i);
t1 = whisper_full_get_segment_t1(ctx, i);
}
if (!params.no_timestamps) {
printf("[%s --> %s] ", to_timestamp(t0).c_str(), to_timestamp(t1).c_str());
}
if (params.diarize && pcmf32s.size() == 2) {
speaker = estimate_diarization_speaker(pcmf32s, t0, t1);
}
if (params.print_colors) {
for (int j = 0; j < whisper_full_n_tokens(ctx, i); ++j) {
if (params.print_special == false) {
const whisper_token id = whisper_full_get_token_id(ctx, i, j);
if (id >= whisper_token_eot(ctx)) {
continue;
}
}
const char * text = whisper_full_get_token_text(ctx, i, j);
const float p = whisper_full_get_token_p (ctx, i, j);
const int col = std::max(0, std::min((int) k_colors.size() - 1, (int) (std::pow(p, 3)*float(k_colors.size()))));
printf("%s%s%s%s", speaker.c_str(), k_colors[col].c_str(), text, "\033[0m");
}
} else {
const char * text = whisper_full_get_segment_text(ctx, i);
printf("%s%s", speaker.c_str(), text);
}
if (params.tinydiarize) {
if (whisper_full_get_segment_speaker_turn_next(ctx, i)) {
printf("%s", params.tdrz_speaker_turn.c_str());
}
}
// with timestamps or speakers: each segment on new line
if (!params.no_timestamps || params.diarize) {
printf("\n");
}
fflush(stdout);
}
}
bool output_txt(struct whisper_context * ctx, const char * fname, const whisper_params & params, std::vector<std::vector<float>> pcmf32s) {
std::ofstream fout(fname);
if (!fout.is_open()) {
fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname);
return false;
}
fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname);
const int n_segments = whisper_full_n_segments(ctx);
for (int i = 0; i < n_segments; ++i) {
const char * text = whisper_full_get_segment_text(ctx, i);
std::string speaker = "";
if (params.diarize && pcmf32s.size() == 2)
{
const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
const int64_t t1 = whisper_full_get_segment_t1(ctx, i);
speaker = estimate_diarization_speaker(pcmf32s, t0, t1);
}
fout << speaker << text << "\n";
}
return true;
}
bool output_vtt(struct whisper_context * ctx, const char * fname, const whisper_params & params, std::vector<std::vector<float>> pcmf32s) {
std::ofstream fout(fname);
if (!fout.is_open()) {
fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname);
return false;
}
fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname);
fout << "WEBVTT\n\n";
const int n_segments = whisper_full_n_segments(ctx);
for (int i = 0; i < n_segments; ++i) {
const char * text = whisper_full_get_segment_text(ctx, i);
const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
const int64_t t1 = whisper_full_get_segment_t1(ctx, i);
std::string speaker = "";
if (params.diarize && pcmf32s.size() == 2)
{
speaker = estimate_diarization_speaker(pcmf32s, t0, t1, true);
speaker.insert(0, "<v Speaker");
speaker.append(">");
}
fout << to_timestamp(t0) << " --> " << to_timestamp(t1) << "\n";
fout << speaker << text << "\n\n";
}
return true;
}
bool output_srt(struct whisper_context * ctx, const char * fname, const whisper_params & params, std::vector<std::vector<float>> pcmf32s) {
std::ofstream fout(fname);
if (!fout.is_open()) {
fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname);
return false;
}
fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname);
const int n_segments = whisper_full_n_segments(ctx);
for (int i = 0; i < n_segments; ++i) {
const char * text = whisper_full_get_segment_text(ctx, i);
const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
const int64_t t1 = whisper_full_get_segment_t1(ctx, i);
std::string speaker = "";
if (params.diarize && pcmf32s.size() == 2)
{
speaker = estimate_diarization_speaker(pcmf32s, t0, t1);
}
fout << i + 1 + params.offset_n << "\n";
fout << to_timestamp(t0, true) << " --> " << to_timestamp(t1, true) << "\n";
fout << speaker << text << "\n\n";
}
return true;
}
char *escape_double_quotes_and_backslashes(const char *str) {
if (str == NULL) {
return NULL;
}
size_t escaped_length = strlen(str) + 1;
for (size_t i = 0; str[i] != '\0'; i++) {
if (str[i] == '"' || str[i] == '\\') {
escaped_length++;
}
}
char *escaped = (char *)calloc(escaped_length, 1); // pre-zeroed
if (escaped == NULL) {
return NULL;
}
size_t pos = 0;
for (size_t i = 0; str[i] != '\0'; i++) {
if (str[i] == '"' || str[i] == '\\') {
escaped[pos++] = '\\';
}
escaped[pos++] = str[i];
}
// no need to set zero due to calloc() being used prior
return escaped;
}
// double quote should be escaped by another double quote. (rfc4180)
char *escape_double_quotes_in_csv(const char *str) {
if (str == NULL) {
return NULL;
}
size_t escaped_length = strlen(str) + 1;
for (size_t i = 0; str[i] != '\0'; i++) {
if (str[i] == '"') {
escaped_length++;
}
}
char *escaped = (char *)calloc(escaped_length, 1); // pre-zeroed
if (escaped == NULL) {
return NULL;
}
size_t pos = 0;
for (size_t i = 0; str[i] != '\0'; i++) {
if (str[i] == '"') {
escaped[pos++] = '"';
}
escaped[pos++] = str[i];
}
// no need to set zero due to calloc() being used prior
return escaped;
}
bool output_csv(struct whisper_context * ctx, const char * fname, const whisper_params & params, std::vector<std::vector<float>> pcmf32s) {
std::ofstream fout(fname);
if (!fout.is_open()) {
fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname);
return false;
}
fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname);
const int n_segments = whisper_full_n_segments(ctx);
fout << "start,end,";
if (params.diarize && pcmf32s.size() == 2)
{
fout << "speaker,";
}
fout << "text\n";
for (int i = 0; i < n_segments; ++i) {
const char * text = whisper_full_get_segment_text(ctx, i);
const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
const int64_t t1 = whisper_full_get_segment_t1(ctx, i);
char * text_escaped = escape_double_quotes_in_csv(text);
//need to multiply times returned from whisper_full_get_segment_t{0,1}() by 10 to get milliseconds.
fout << 10 * t0 << "," << 10 * t1 << ",";
if (params.diarize && pcmf32s.size() == 2)
{
fout << estimate_diarization_speaker(pcmf32s, t0, t1, true) << ",";
}
fout << "\"" << text_escaped << "\"\n";
}
return true;
}
bool output_score(struct whisper_context * ctx, const char * fname, const whisper_params & /*params*/, std::vector<std::vector<float>> /*pcmf32s*/) {
std::ofstream fout(fname);
fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname);
const int n_segments = whisper_full_n_segments(ctx);
// fprintf(stderr,"segments: %d\n",n_segments);
for (int i = 0; i < n_segments; ++i) {
const int n_tokens = whisper_full_n_tokens(ctx, i);
// fprintf(stderr,"tokens: %d\n",n_tokens);
for (int j = 0; j < n_tokens; j++) {
auto token = whisper_full_get_token_text(ctx, i, j);
auto probability = whisper_full_get_token_p(ctx, i, j);
fout << token << '\t' << probability << std::endl;
// fprintf(stderr,"token: %s %f\n",token,probability);
}
}
return true;
}
bool output_json(
struct whisper_context * ctx,
const char * fname,
const whisper_params & params,
std::vector<std::vector<float>> pcmf32s,
bool full) {
std::ofstream fout(fname);
int indent = 0;
auto doindent = [&]() {
for (int i = 0; i < indent; i++) fout << "\t";
};
auto start_arr = [&](const char *name) {
doindent();
fout << "\"" << name << "\": [\n";
indent++;
};
auto end_arr = [&](bool end) {
indent--;
doindent();
fout << (end ? "]\n" : "],\n");
};
auto start_obj = [&](const char *name) {
doindent();
if (name) {
fout << "\"" << name << "\": {\n";
} else {
fout << "{\n";
}
indent++;
};
auto end_obj = [&](bool end) {
indent--;
doindent();
fout << (end ? "}\n" : "},\n");
};
auto start_value = [&](const char *name) {
doindent();
fout << "\"" << name << "\": ";
};
auto value_s = [&](const char *name, const char *val, bool end) {
start_value(name);
char * val_escaped = escape_double_quotes_and_backslashes(val);
fout << "\"" << val_escaped << (end ? "\"\n" : "\",\n");
free(val_escaped);
};
auto end_value = [&](bool end) {
fout << (end ? "\n" : ",\n");
};
auto value_i = [&](const char *name, const int64_t val, bool end) {
start_value(name);
fout << val;
end_value(end);
};
auto value_f = [&](const char *name, const float val, bool end) {
start_value(name);
fout << val;
end_value(end);
};
auto value_b = [&](const char *name, const bool val, bool end) {
start_value(name);
fout << (val ? "true" : "false");
end_value(end);
};
auto times_o = [&](int64_t t0, int64_t t1, bool end) {
start_obj("timestamps");
value_s("from", to_timestamp(t0, true).c_str(), false);
value_s("to", to_timestamp(t1, true).c_str(), true);
end_obj(false);
start_obj("offsets");
value_i("from", t0 * 10, false);
value_i("to", t1 * 10, true);
end_obj(end);
};
if (!fout.is_open()) {
fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname);
return false;
}
fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname);
start_obj(nullptr);
value_s("systeminfo", whisper_print_system_info(), false);
start_obj("model");
value_s("type", whisper_model_type_readable(ctx), false);
value_b("multilingual", whisper_is_multilingual(ctx), false);
value_i("vocab", whisper_model_n_vocab(ctx), false);
start_obj("audio");
value_i("ctx", whisper_model_n_audio_ctx(ctx), false);
value_i("state", whisper_model_n_audio_state(ctx), false);
value_i("head", whisper_model_n_audio_head(ctx), false);
value_i("layer", whisper_model_n_audio_layer(ctx), true);
end_obj(false);
start_obj("text");
value_i("ctx", whisper_model_n_text_ctx(ctx), false);
value_i("state", whisper_model_n_text_state(ctx), false);
value_i("head", whisper_model_n_text_head(ctx), false);
value_i("layer", whisper_model_n_text_layer(ctx), true);
end_obj(false);
value_i("mels", whisper_model_n_mels(ctx), false);
value_i("ftype", whisper_model_ftype(ctx), true);
end_obj(false);
start_obj("params");
value_s("model", params.model.c_str(), false);
value_s("language", params.language.c_str(), false);
value_b("translate", params.translate, true);
end_obj(false);
start_obj("result");
value_s("language", whisper_lang_str(whisper_full_lang_id(ctx)), true);
end_obj(false);
start_arr("transcription");
const int n_segments = whisper_full_n_segments(ctx);
for (int i = 0; i < n_segments; ++i) {
const char * text = whisper_full_get_segment_text(ctx, i);
const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
const int64_t t1 = whisper_full_get_segment_t1(ctx, i);
start_obj(nullptr);
times_o(t0, t1, false);
value_s("text", text, !params.diarize && !params.tinydiarize && !full);
if (full) {
start_arr("tokens");
const int n = whisper_full_n_tokens(ctx, i);
for (int j = 0; j < n; ++j) {
auto token = whisper_full_get_token_data(ctx, i, j);
start_obj(nullptr);
value_s("text", whisper_token_to_str(ctx, token.id), false);
if(token.t0 > -1 && token.t1 > -1) {
// If we have per-token timestamps, write them out
times_o(token.t0, token.t1, false);
}
value_i("id", token.id, false);
value_f("p", token.p, false);
value_f("t_dtw", token.t_dtw, true);
end_obj(j == (n - 1));
}
end_arr(!params.diarize && !params.tinydiarize);
}
if (params.diarize && pcmf32s.size() == 2) {
value_s("speaker", estimate_diarization_speaker(pcmf32s, t0, t1, true).c_str(), true);
}
if (params.tinydiarize) {
value_b("speaker_turn_next", whisper_full_get_segment_speaker_turn_next(ctx, i), true);
}
end_obj(i == (n_segments - 1));
}
end_arr(true);
end_obj(true);
return true;
}
// karaoke video generation
// outputs a bash script that uses ffmpeg to generate a video with the subtitles
// TODO: font parameter adjustments
bool output_wts(struct whisper_context * ctx, const char * fname, const char * fname_inp, const whisper_params & params, float t_sec, std::vector<std::vector<float>> pcmf32s) {
std::ofstream fout(fname);
fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname);
static const char * font = params.font_path.c_str();
std::ifstream fin(font);
if (!fin.is_open()) {
fprintf(stderr, "%s: font not found at '%s', please specify a monospace font with -fp\n", __func__, font);
return false;
}
fout << "#!/bin/bash" << "\n";
fout << "\n";
fout << "ffmpeg -i " << fname_inp << " -f lavfi -i color=size=1200x120:duration=" << t_sec << ":rate=25:color=black -vf \"";
for (int i = 0; i < whisper_full_n_segments(ctx); i++) {
const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
const int64_t t1 = whisper_full_get_segment_t1(ctx, i);
const int n = whisper_full_n_tokens(ctx, i);
std::vector<whisper_token_data> tokens(n);
for (int j = 0; j < n; ++j) {
tokens[j] = whisper_full_get_token_data(ctx, i, j);
}
if (i > 0) {
fout << ",";
}
// background text
fout << "drawtext=fontfile='" << font << "':fontsize=24:fontcolor=gray:x=(w-text_w)/2:y=h/2:text='':enable='between(t," << t0/100.0 << "," << t0/100.0 << ")'";
bool is_first = true;
std::string speaker = "";
if (params.diarize && pcmf32s.size() == 2) {
speaker = estimate_diarization_speaker(pcmf32s, t0, t1);
}
for (int j = 0; j < n; ++j) {
const auto & token = tokens[j];
if (tokens[j].id >= whisper_token_eot(ctx)) {
continue;
}
std::string txt_bg = "";
std::string txt_fg = ""; // highlight token
std::string txt_ul = ""; // underline
if (params.diarize && pcmf32s.size() == 2) {
txt_bg = speaker;
txt_fg = speaker;
txt_ul = "\\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ ";
}
txt_bg.append("> ");
txt_fg.append("> ");
txt_ul.append("\\ \\ ");
{
for (int k = 0; k < n; ++k) {
const auto & token2 = tokens[k];
if (tokens[k].id >= whisper_token_eot(ctx)) {
continue;
}
const std::string txt = whisper_token_to_str(ctx, token2.id);
txt_bg += txt;
if (k == j) {
for (int l = 0; l < (int) txt.size(); ++l) {
txt_fg += txt[l];
txt_ul += "_";
}
txt_fg += "|";
} else {
for (int l = 0; l < (int) txt.size(); ++l) {
txt_fg += "\\ ";
txt_ul += "\\ ";
}
}
}
::replace_all(txt_bg, "'", "\u2019");
::replace_all(txt_bg, "\"", "\\\"");
::replace_all(txt_fg, "'", "\u2019");
::replace_all(txt_fg, "\"", "\\\"");
}
if (is_first) {
// background text
fout << ",drawtext=fontfile='" << font << "':fontsize=24:fontcolor=gray:x=(w-text_w)/2:y=h/2:text='" << txt_bg << "':enable='between(t," << t0/100.0 << "," << t1/100.0 << ")'";
is_first = false;
}
// foreground text
fout << ",drawtext=fontfile='" << font << "':fontsize=24:fontcolor=lightgreen:x=(w-text_w)/2+8:y=h/2:text='" << txt_fg << "':enable='between(t," << token.t0/100.0 << "," << token.t1/100.0 << ")'";
// underline
fout << ",drawtext=fontfile='" << font << "':fontsize=24:fontcolor=lightgreen:x=(w-text_w)/2+8:y=h/2+16:text='" << txt_ul << "':enable='between(t," << token.t0/100.0 << "," << token.t1/100.0 << ")'";
}
}
fout << "\" -c:v libx264 -pix_fmt yuv420p -y " << fname_inp << ".mp4" << "\n";
fout << "\n\n";
fout << "echo \"Your video has been saved to " << fname_inp << ".mp4\"" << "\n";
fout << "\n";
fout << "echo \" ffplay " << fname_inp << ".mp4\"\n";
fout << "\n";
fout.close();
fprintf(stderr, "%s: run 'source %s' to generate karaoke video\n", __func__, fname);
return true;
}
bool output_lrc(struct whisper_context * ctx, const char * fname, const whisper_params & params, std::vector<std::vector<float>> pcmf32s) {
std::ofstream fout(fname);
if (!fout.is_open()) {
fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname);
return false;
}
fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname);
fout << "[by:whisper.cpp]\n";
const int n_segments = whisper_full_n_segments(ctx);
for (int i = 0; i < n_segments; ++i) {
const char * text = whisper_full_get_segment_text(ctx, i);
const int64_t t = whisper_full_get_segment_t0(ctx, i);
int64_t msec = t * 10;
int64_t min = msec / (1000 * 60);
msec = msec - min * (1000 * 60);
int64_t sec = msec / 1000;
msec = msec - sec * 1000;
char buf[16];
snprintf(buf, sizeof(buf), "%02d:%02d.%02d", (int) min, (int) sec, (int) ( msec / 10));
std::string timestamp_lrc = std::string(buf);
std::string speaker = "";
if (params.diarize && pcmf32s.size() == 2)
{
const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
const int64_t t1 = whisper_full_get_segment_t1(ctx, i);
speaker = estimate_diarization_speaker(pcmf32s, t0, t1);
}
fout << '[' << timestamp_lrc << ']' << speaker << text << "\n";
}
return true;
}
void cb_log_disable(enum ggml_log_level , const char * , void * ) { }
int main(int argc, char ** argv) {
whisper_params params;
// If the only argument starts with "@", read arguments line-by-line
// from the given file.
std::vector<std::string> vec_args;
if (argc == 2 && argv != nullptr && argv[1] != nullptr && argv[1][0] == '@') {
// Save the name of the executable.
vec_args.push_back(argv[0]);
// Open the response file.
char const * rspfile = argv[1] + sizeof(char);
std::ifstream fin(rspfile);
if (fin.is_open() == false) {
fprintf(stderr, "error: response file '%s' not found\n", rspfile);
return 1;
}
// Read the entire response file.
std::string line;
while (std::getline(fin, line)) {
vec_args.push_back(line);
}
// Use the contents of the response file as the command-line arguments.
argc = static_cast<int>(vec_args.size());
argv = static_cast<char **>(alloca(argc * sizeof (char *)));
for (int i = 0; i < argc; ++i) {
argv[i] = const_cast<char *>(vec_args[i].c_str());
}
}
if (whisper_params_parse(argc, argv, params) == false) {
whisper_print_usage(argc, argv, params);
return 1;
}
// remove non-existent files
for (auto it = params.fname_inp.begin(); it != params.fname_inp.end();) {
const auto fname_inp = it->c_str();
if (*it != "-" && !is_file_exist(fname_inp)) {
fprintf(stderr, "error: input file not found '%s'\n", fname_inp);
it = params.fname_inp.erase(it);
continue;
}
it++;
}
if (params.fname_inp.empty()) {
fprintf(stderr, "error: no input files specified\n");
whisper_print_usage(argc, argv, params);
return 2;
}
if (params.language != "auto" && whisper_lang_id(params.language.c_str()) == -1) {
fprintf(stderr, "error: unknown language '%s'\n", params.language.c_str());
whisper_print_usage(argc, argv, params);
exit(0);
}
if (params.diarize && params.tinydiarize) {
fprintf(stderr, "error: cannot use both --diarize and --tinydiarize\n");
whisper_print_usage(argc, argv, params);
exit(0);
}
if (params.no_prints) {
whisper_log_set(cb_log_disable, NULL);
}
// whisper init
struct whisper_context_params cparams = whisper_context_default_params();
cparams.use_gpu = params.use_gpu;
if (!params.dtw.empty()) {
cparams.dtw_token_timestamps = true;
cparams.dtw_aheads_preset = WHISPER_AHEADS_NONE;
if (params.dtw == "tiny") cparams.dtw_aheads_preset = WHISPER_AHEADS_TINY;
if (params.dtw == "tiny.en") cparams.dtw_aheads_preset = WHISPER_AHEADS_TINY_EN;
if (params.dtw == "base") cparams.dtw_aheads_preset = WHISPER_AHEADS_BASE;
if (params.dtw == "base.en") cparams.dtw_aheads_preset = WHISPER_AHEADS_BASE_EN;
if (params.dtw == "small") cparams.dtw_aheads_preset = WHISPER_AHEADS_SMALL;
if (params.dtw == "small.en") cparams.dtw_aheads_preset = WHISPER_AHEADS_SMALL_EN;
if (params.dtw == "medium") cparams.dtw_aheads_preset = WHISPER_AHEADS_MEDIUM;
if (params.dtw == "medium.en") cparams.dtw_aheads_preset = WHISPER_AHEADS_MEDIUM_EN;
if (params.dtw == "large.v1") cparams.dtw_aheads_preset = WHISPER_AHEADS_LARGE_V1;
if (params.dtw == "large.v2") cparams.dtw_aheads_preset = WHISPER_AHEADS_LARGE_V2;
if (params.dtw == "large.v3") cparams.dtw_aheads_preset = WHISPER_AHEADS_LARGE_V3;
if (cparams.dtw_aheads_preset == WHISPER_AHEADS_NONE) {
fprintf(stderr, "error: unknown DTW preset '%s'\n", params.dtw.c_str());
return 3;