Spaces:
Running
Running
main : refactor subtitle output
Browse files
main.cpp
CHANGED
|
@@ -141,6 +141,66 @@ void whisper_print_usage(int argc, char ** argv, const whisper_params & params)
|
|
| 141 |
fprintf(stderr, "\n");
|
| 142 |
}
|
| 143 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 144 |
int main(int argc, char ** argv) {
|
| 145 |
whisper_params params;
|
| 146 |
|
|
@@ -291,66 +351,19 @@ int main(int argc, char ** argv) {
|
|
| 291 |
// output to text file
|
| 292 |
if (params.output_txt) {
|
| 293 |
const auto fname_txt = fname_inp + ".txt";
|
| 294 |
-
|
| 295 |
-
if (!fout_txt.is_open()) {
|
| 296 |
-
fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname_txt.c_str());
|
| 297 |
-
return 8;
|
| 298 |
-
}
|
| 299 |
-
|
| 300 |
-
fprintf(stderr, "%s: saving output to '%s.txt'\n", __func__, fname_inp.c_str());
|
| 301 |
-
|
| 302 |
-
const int n_segments = whisper_full_n_segments(ctx);
|
| 303 |
-
for (int i = 0; i < n_segments; ++i) {
|
| 304 |
-
const char * text = whisper_full_get_segment_text(ctx, i);
|
| 305 |
-
fout_txt << text;
|
| 306 |
-
}
|
| 307 |
}
|
| 308 |
|
| 309 |
// output to VTT file
|
| 310 |
if (params.output_vtt) {
|
| 311 |
const auto fname_vtt = fname_inp + ".vtt";
|
| 312 |
-
|
| 313 |
-
if (!fout_vtt.is_open()) {
|
| 314 |
-
fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname_vtt.c_str());
|
| 315 |
-
return 9;
|
| 316 |
-
}
|
| 317 |
-
|
| 318 |
-
fprintf(stderr, "%s: saving output to '%s.vtt'\n", __func__, fname_inp.c_str());
|
| 319 |
-
|
| 320 |
-
fout_vtt << "WEBVTT\n\n";
|
| 321 |
-
|
| 322 |
-
const int n_segments = whisper_full_n_segments(ctx);
|
| 323 |
-
for (int i = 0; i < n_segments; ++i) {
|
| 324 |
-
const char * text = whisper_full_get_segment_text(ctx, i);
|
| 325 |
-
const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
|
| 326 |
-
const int64_t t1 = whisper_full_get_segment_t1(ctx, i);
|
| 327 |
-
|
| 328 |
-
fout_vtt << to_timestamp(t0) << " --> " << to_timestamp(t1) << "\n";
|
| 329 |
-
fout_vtt << text << "\n\n";
|
| 330 |
-
}
|
| 331 |
}
|
| 332 |
|
| 333 |
// output to SRT file
|
| 334 |
if (params.output_srt) {
|
| 335 |
const auto fname_srt = fname_inp + ".srt";
|
| 336 |
-
|
| 337 |
-
if (!fout_srt.is_open()) {
|
| 338 |
-
fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname_srt.c_str());
|
| 339 |
-
return 10;
|
| 340 |
-
}
|
| 341 |
-
|
| 342 |
-
fprintf(stderr, "%s: saving output to '%s.srt'\n", __func__, fname_inp.c_str());
|
| 343 |
-
|
| 344 |
-
const int n_segments = whisper_full_n_segments(ctx);
|
| 345 |
-
for (int i = 0; i < n_segments; ++i) {
|
| 346 |
-
const char * text = whisper_full_get_segment_text(ctx, i);
|
| 347 |
-
const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
|
| 348 |
-
const int64_t t1 = whisper_full_get_segment_t1(ctx, i);
|
| 349 |
-
|
| 350 |
-
fout_srt << i + 1 + params.offset_n << "\n";
|
| 351 |
-
fout_srt << to_timestamp(t0) << " --> " << to_timestamp(t1) << "\n";
|
| 352 |
-
fout_srt << text << "\n\n";
|
| 353 |
-
}
|
| 354 |
}
|
| 355 |
}
|
| 356 |
}
|
|
|
|
| 141 |
fprintf(stderr, "\n");
|
| 142 |
}
|
| 143 |
|
| 144 |
+
bool output_txt(struct whisper_context * ctx, const char * fname) {
|
| 145 |
+
std::ofstream fout(fname);
|
| 146 |
+
if (!fout.is_open()) {
|
| 147 |
+
fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname);
|
| 148 |
+
return false;
|
| 149 |
+
}
|
| 150 |
+
|
| 151 |
+
fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname);
|
| 152 |
+
|
| 153 |
+
const int n_segments = whisper_full_n_segments(ctx);
|
| 154 |
+
for (int i = 0; i < n_segments; ++i) {
|
| 155 |
+
const char * text = whisper_full_get_segment_text(ctx, i);
|
| 156 |
+
fout << text;
|
| 157 |
+
}
|
| 158 |
+
|
| 159 |
+
return true;
|
| 160 |
+
}
|
| 161 |
+
|
| 162 |
+
bool output_vtt(struct whisper_context * ctx, const char * fname) {
|
| 163 |
+
std::ofstream fout(fname);
|
| 164 |
+
if (!fout.is_open()) {
|
| 165 |
+
fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname);
|
| 166 |
+
return 9;
|
| 167 |
+
}
|
| 168 |
+
|
| 169 |
+
fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname);
|
| 170 |
+
|
| 171 |
+
fout << "WEBVTT\n\n";
|
| 172 |
+
|
| 173 |
+
const int n_segments = whisper_full_n_segments(ctx);
|
| 174 |
+
for (int i = 0; i < n_segments; ++i) {
|
| 175 |
+
const char * text = whisper_full_get_segment_text(ctx, i);
|
| 176 |
+
const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
|
| 177 |
+
const int64_t t1 = whisper_full_get_segment_t1(ctx, i);
|
| 178 |
+
|
| 179 |
+
fout << to_timestamp(t0) << " --> " << to_timestamp(t1) << "\n";
|
| 180 |
+
fout << text << "\n\n";
|
| 181 |
+
}
|
| 182 |
+
|
| 183 |
+
return true;
|
| 184 |
+
}
|
| 185 |
+
|
| 186 |
+
bool output_srt(struct whisper_context * ctx, const char * fname) {
|
| 187 |
+
std::ofstream fout(fname);
|
| 188 |
+
if (!fout.is_open()) {
|
| 189 |
+
fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname);
|
| 190 |
+
return false;
|
| 191 |
+
}
|
| 192 |
+
|
| 193 |
+
fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname);
|
| 194 |
+
|
| 195 |
+
const int n_segments = whisper_full_n_segments(ctx);
|
| 196 |
+
for (int i = 0; i < n_segments; ++i) {
|
| 197 |
+
const char * text = whisper_full_get_segment_text(ctx, i);
|
| 198 |
+
fout << text;
|
| 199 |
+
}
|
| 200 |
+
|
| 201 |
+
return true;
|
| 202 |
+
}
|
| 203 |
+
|
| 204 |
int main(int argc, char ** argv) {
|
| 205 |
whisper_params params;
|
| 206 |
|
|
|
|
| 351 |
// output to text file
|
| 352 |
if (params.output_txt) {
|
| 353 |
const auto fname_txt = fname_inp + ".txt";
|
| 354 |
+
output_txt(ctx, fname_txt.c_str());
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 355 |
}
|
| 356 |
|
| 357 |
// output to VTT file
|
| 358 |
if (params.output_vtt) {
|
| 359 |
const auto fname_vtt = fname_inp + ".vtt";
|
| 360 |
+
output_vtt(ctx, fname_vtt.c_str());
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 361 |
}
|
| 362 |
|
| 363 |
// output to SRT file
|
| 364 |
if (params.output_srt) {
|
| 365 |
const auto fname_srt = fname_inp + ".srt";
|
| 366 |
+
output_srt(ctx, fname_srt.c_str());
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 367 |
}
|
| 368 |
}
|
| 369 |
}
|