forked from ggerganov/whisper.cpp
-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.cpp
243 lines (204 loc) · 8.84 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
#include "whisper.h"
// third-party utilities
// use your favorite implementations
#define DR_WAV_IMPLEMENTATION
#include "dr_wav.h"
#include <cstdio>
#include <string>
#include <thread>
#include <vector>
// 500 -> 00:05.000
// 6000 -> 01:00.000
std::string to_timestamp(int64_t t) {
int64_t sec = t/100;
int64_t msec = t - sec*100;
int64_t min = sec/60;
sec = sec - min*60;
char buf[32];
snprintf(buf, sizeof(buf), "%02d:%02d.%03d", (int) min, (int) sec, (int) msec);
return std::string(buf);
}
// command-line parameters
struct whisper_params {
int32_t seed = -1; // RNG seed, not used currently
int32_t n_threads = std::min(4, (int32_t) std::thread::hardware_concurrency());
int32_t offset_ms = 0;
bool verbose = false;
bool translate = false;
bool print_special_tokens = false;
bool no_timestamps = false;
std::string language = "en";
std::string model = "models/ggml-base.en.bin";
std::vector<std::string> fname_inp = {};
};
void whisper_print_usage(int argc, char ** argv, const whisper_params & params);
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[0] != '-') {
params.fname_inp.push_back(arg);
continue;
}
if (arg == "-s" || arg == "--seed") {
params.seed = std::stoi(argv[++i]);
} else if (arg == "-t" || arg == "--threads") {
params.n_threads = std::stoi(argv[++i]);
} else if (arg == "-o" || arg == "--offset") {
params.offset_ms = std::stoi(argv[++i]);
} else if (arg == "-v" || arg == "--verbose") {
params.verbose = true;
} else if (arg == "--translate") {
params.translate = true;
} else if (arg == "-l" || arg == "--language") {
params.language = argv[++i];
if (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);
}
} else if (arg == "-ps" || arg == "--print_special") {
params.print_special_tokens = true;
} else if (arg == "-nt" || arg == "--no_timestamps") {
params.no_timestamps = true;
} else if (arg == "-m" || arg == "--model") {
params.model = argv[++i];
} else if (arg == "-f" || arg == "--file") {
params.fname_inp.push_back(argv[++i]);
} else if (arg == "-h" || arg == "--help") {
whisper_print_usage(argc, argv, params);
exit(0);
} 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 show this help message and exit\n");
fprintf(stderr, " -s SEED, --seed SEED RNG seed (default: -1)\n");
fprintf(stderr, " -t N, --threads N number of threads to use during computation (default: %d)\n", params.n_threads);
fprintf(stderr, " -o N, --offset N offset in milliseconds (default: %d)\n", params.offset_ms);
fprintf(stderr, " -v, --verbose verbose output\n");
fprintf(stderr, " --translate translate from source language to english\n");
fprintf(stderr, " -ps, --print_special print special tokens\n");
fprintf(stderr, " -nt, --no_timestamps do not print timestamps\n");
fprintf(stderr, " -l LANG, --language LANG spoken language (default: %s)\n", params.language.c_str());
fprintf(stderr, " -m FNAME, --model FNAME model path (default: %s)\n", params.model.c_str());
fprintf(stderr, " -f FNAME, --file FNAME input WAV file path\n");
fprintf(stderr, "\n");
}
int main(int argc, char ** argv) {
whisper_params params;
if (whisper_params_parse(argc, argv, params) == false) {
return 1;
}
if (params.seed < 0) {
params.seed = time(NULL);
}
if (params.fname_inp.empty()) {
fprintf(stderr, "error: no input files specified\n");
whisper_print_usage(argc, argv, params);
return 1;
}
// whisper init
struct whisper_context * ctx = whisper_init(params.model.c_str());
for (int f = 0; f < (int) params.fname_inp.size(); ++f) {
const auto fname_inp = params.fname_inp[f];
// WAV input
std::vector<float> pcmf32;
{
drwav wav;
if (!drwav_init_file(&wav, fname_inp.c_str(), NULL)) {
fprintf(stderr, "%s: failed to open WAV file '%s' - check your input\n", argv[0], fname_inp.c_str());
whisper_print_usage(argc, argv, {});
return 2;
}
if (wav.channels != 1 && wav.channels != 2) {
fprintf(stderr, "%s: WAV file '%s' must be mono or stereo\n", argv[0], fname_inp.c_str());
return 3;
}
if (wav.sampleRate != WHISPER_SAMPLE_RATE) {
fprintf(stderr, "%s: WAV file '%s' must be 16 kHz\n", argv[0], fname_inp.c_str());
return 4;
}
if (wav.bitsPerSample != 16) {
fprintf(stderr, "%s: WAV file '%s' must be 16-bit\n", argv[0], fname_inp.c_str());
return 5;
}
int n = wav.totalPCMFrameCount;
std::vector<int16_t> pcm16;
pcm16.resize(n*wav.channels);
drwav_read_pcm_frames_s16(&wav, n, pcm16.data());
drwav_uninit(&wav);
// convert to mono, float
pcmf32.resize(n);
if (wav.channels == 1) {
for (int i = 0; i < n; i++) {
pcmf32[i] = float(pcm16[i])/32768.0f;
}
} else {
for (int i = 0; i < n; i++) {
pcmf32[i] = float(pcm16[2*i] + pcm16[2*i + 1])/65536.0f;
}
}
}
// print some info about the processing
{
printf("\n");
if (!whisper_is_multilingual(ctx)) {
if (params.language != "en" || params.translate) {
params.language = "en";
params.translate = false;
printf("%s: WARNING: model is not multilingual, ignoring language and translation options\n", __func__);
}
}
printf("%s: processing '%s' (%d samples, %.1f sec), %d threads, lang = %s, task = %s, timestamps = %d ...\n",
__func__, fname_inp.c_str(), int(pcmf32.size()), float(pcmf32.size())/WHISPER_SAMPLE_RATE, params.n_threads,
params.language.c_str(),
params.translate ? "translate" : "transcribe",
params.no_timestamps ? 0 : 1);
printf("\n");
}
// run the inference
{
whisper_full_params wparams = whisper_full_default_params(WHISPER_DECODE_GREEDY);
wparams.print_realtime = true;
wparams.print_progress = false;
wparams.print_timestamps = !params.no_timestamps;
wparams.print_special_tokens = params.print_special_tokens;
wparams.translate = params.translate;
wparams.language = params.language.c_str();
wparams.n_threads = params.n_threads;
wparams.offset_ms = params.offset_ms;
if (whisper_full(ctx, wparams, pcmf32.data(), pcmf32.size()) != 0) {
fprintf(stderr, "%s: failed to process audio\n", argv[0]);
return 6;
}
// print result;
if (!wparams.print_realtime) {
printf("\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);
if (params.no_timestamps) {
printf ("%s", text);
fflush(stdout);
} else {
const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
const int64_t t1 = whisper_full_get_segment_t1(ctx, i);
printf ("[%s --> %s] %s\n", to_timestamp(t0).c_str(), to_timestamp(t1).c_str(), text);
}
}
}
}
}
whisper_print_timings(ctx);
whisper_free(ctx);
return 0;
}