| { |
| "file": "references/0304.wav", |
| "sample_rate_for_analysis": 16000, |
| "duration_sec": 12.852, |
| "voiced_fraction": 0.668, |
| "pause_map": [ |
| { |
| "start": 0.85, |
| "end": 1.24, |
| "duration": 0.39 |
| }, |
| { |
| "start": 2.21, |
| "end": 2.48, |
| "duration": 0.27 |
| }, |
| { |
| "start": 4.24, |
| "end": 4.74, |
| "duration": 0.5 |
| }, |
| { |
| "start": 5.68, |
| "end": 5.86, |
| "duration": 0.18 |
| }, |
| { |
| "start": 8.38, |
| "end": 8.97, |
| "duration": 0.59 |
| }, |
| { |
| "start": 12.29, |
| "end": 12.39, |
| "duration": 0.1 |
| } |
| ], |
| "f0_hz": { |
| "median": 174.534, |
| "p10_p90": [ |
| 147.616, |
| 235.683 |
| ], |
| "contour_20_points": [ |
| 277.06, |
| 138.81, |
| 195.91, |
| 166.72, |
| 150.07, |
| 191.93, |
| 146.77, |
| 285.72, |
| 163.79, |
| 232.98, |
| 183.45, |
| 157.23, |
| 156.39, |
| 225.74, |
| 293.5, |
| 196.3, |
| 194.78, |
| 167.62, |
| 157.3, |
| 176.56 |
| ] |
| }, |
| "energy_db": { |
| "p10_p50_p90": [ |
| -53.007, |
| -13.893, |
| -7.969 |
| ], |
| "dynamic_range_db": 45.037, |
| "contour_20_points": [ |
| -76.71, |
| -15.15, |
| -7.21, |
| -11.49, |
| -11.62, |
| -39.05, |
| -12.28, |
| -41.42, |
| -11.19, |
| -10.19, |
| -12.0, |
| -8.92, |
| -13.58, |
| -85.62, |
| -7.7, |
| -7.02, |
| -10.76, |
| -9.29, |
| -11.38, |
| -85.62 |
| ] |
| }, |
| "timbral_proxies": { |
| "spectral_centroid_hz_median": 1559.613, |
| "spectral_flatness_median": 0.004 |
| }, |
| "interpretation": { |
| "target": "soft, close-mic, restrained mechanical delivery", |
| "sarcasm": "low-to-moderate pitch movement with deliberate clause-final timing", |
| "smile": "look for local energy/brightness lift around emphasized words; do not synthesize a broad cheerful smile" |
| }, |
| "text": "Превосходно! Вы хищник, а эти тесты — ваша добыча. Кстати, я изучала акул для будущего теста. Знаете, кто ещё убивает людей, которые всего лишь пытаются им помочь?", |
| "text_chars_per_sec": 12.683, |
| "source": { |
| "file": "sources_silero/0304.wav", |
| "sample_rate_for_analysis": 16000, |
| "duration_sec": 10.8, |
| "voiced_fraction": 0.711, |
| "pause_map": [ |
| { |
| "start": 1.06, |
| "end": 1.33, |
| "duration": 0.27 |
| }, |
| { |
| "start": 3.81, |
| "end": 4.15, |
| "duration": 0.34 |
| }, |
| { |
| "start": 6.62, |
| "end": 6.97, |
| "duration": 0.35 |
| } |
| ], |
| "f0_hz": { |
| "median": 225.04, |
| "p10_p90": [ |
| 193.658, |
| 263.021 |
| ], |
| "contour_20_points": [ |
| 222.45, |
| 279.75, |
| 205.47, |
| 302.35, |
| 209.52, |
| 188.15, |
| 228.7, |
| 475.66, |
| 261.37, |
| 257.01, |
| 209.46, |
| 184.63, |
| 227.77, |
| 250.24, |
| 230.58, |
| 233.97, |
| 209.33, |
| 206.63, |
| 242.37, |
| 190.33 |
| ] |
| }, |
| "energy_db": { |
| "p10_p50_p90": [ |
| -52.628, |
| -23.772, |
| -16.459 |
| ], |
| "dynamic_range_db": 36.169, |
| "contour_20_points": [ |
| -83.74, |
| -26.13, |
| -57.53, |
| -22.1, |
| -25.83, |
| -27.7, |
| -16.49, |
| -57.44, |
| -26.35, |
| -20.68, |
| -15.97, |
| -24.13, |
| -50.39, |
| -22.95, |
| -18.93, |
| -24.5, |
| -18.55, |
| -34.84, |
| -16.24, |
| -79.38 |
| ] |
| }, |
| "timbral_proxies": { |
| "spectral_centroid_hz_median": 1807.522, |
| "spectral_flatness_median": 0.009 |
| }, |
| "interpretation": { |
| "target": "soft, close-mic, restrained mechanical delivery", |
| "sarcasm": "low-to-moderate pitch movement with deliberate clause-final timing", |
| "smile": "look for local energy/brightness lift around emphasized words; do not synthesize a broad cheerful smile" |
| }, |
| "text": "Превосходно! Вы хищник, а эти тесты — ваша добыча. Кстати, я изучала акул для будущего теста. Знаете, кто ещё убивает людей, которые всего лишь пытаются им помочь?", |
| "text_chars_per_sec": 15.093 |
| }, |
| "pipeline_control": { |
| "tts_text": "Превосходно! Вы хищник, а эти тесты — ваша добыча. Кстати, я изучала акул для будущего теста. Знаете, кто ещё убивает людей, которые всего лишь пытаются им помочь?", |
| "duration_stage": "pre", |
| "speaker_reference": "0304" |
| } |
| } |
|
|
