| { |
| "file": "references/0142.wav", |
| "sample_rate_for_analysis": 16000, |
| "duration_sec": 3.691, |
| "voiced_fraction": 0.724, |
| "pause_map": [ |
| { |
| "start": 1.1, |
| "end": 1.3, |
| "duration": 0.2 |
| } |
| ], |
| "f0_hz": { |
| "median": 203.404, |
| "p10_p90": [ |
| 146.765, |
| 261.506 |
| ], |
| "contour_20_points": [ |
| 248.26, |
| 263.9, |
| 208.74, |
| 263.02, |
| 165.98, |
| 146.77, |
| 167.59, |
| 262.88, |
| 253.58, |
| 185.31, |
| 200.26, |
| 214.93, |
| 205.17, |
| 222.51, |
| 218.71, |
| 145.84, |
| 209.44, |
| 195.67, |
| 146.77, |
| 150.2 |
| ] |
| }, |
| "energy_db": { |
| "p10_p50_p90": [ |
| -46.317, |
| -21.486, |
| -11.392 |
| ], |
| "dynamic_range_db": 34.925, |
| "contour_20_points": [ |
| -82.39, |
| -34.69, |
| -23.58, |
| -19.92, |
| -27.79, |
| -18.49, |
| -61.19, |
| -17.76, |
| -31.32, |
| -13.87, |
| -40.37, |
| -17.85, |
| -13.13, |
| -57.6, |
| -10.5, |
| -29.15, |
| -23.95, |
| -21.85, |
| -24.14, |
| -69.52 |
| ] |
| }, |
| "timbral_proxies": { |
| "spectral_centroid_hz_median": 2552.386, |
| "spectral_flatness_median": 0.008 |
| }, |
| "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": "Теперь я не шучу. Поворачивайте назад, иначе я вас убью.", |
| "text_chars_per_sec": 15.174, |
| "source": { |
| "file": "sources_silero/0142.wav", |
| "sample_rate_for_analysis": 16000, |
| "duration_sec": 4.075, |
| "voiced_fraction": 0.667, |
| "pause_map": [ |
| { |
| "start": 0.47, |
| "end": 0.75, |
| "duration": 0.28 |
| }, |
| { |
| "start": 1.44, |
| "end": 1.82, |
| "duration": 0.38 |
| }, |
| { |
| "start": 2.85, |
| "end": 2.99, |
| "duration": 0.14 |
| } |
| ], |
| "f0_hz": { |
| "median": 199.333, |
| "p10_p90": [ |
| 166.653, |
| 230.167 |
| ], |
| "contour_20_points": [ |
| 261.51, |
| 243.71, |
| 187.74, |
| 231.75, |
| 214.62, |
| 228.14, |
| 196.05, |
| 157.3, |
| 200.19, |
| 205.2, |
| 241.3, |
| 210.96, |
| 187.59, |
| 175.33, |
| 212.36, |
| 222.63, |
| 188.15, |
| 196.6, |
| 160.0, |
| 166.65 |
| ] |
| }, |
| "energy_db": { |
| "p10_p50_p90": [ |
| -72.531, |
| -24.355, |
| -15.555 |
| ], |
| "dynamic_range_db": 56.975, |
| "contour_20_points": [ |
| -86.88, |
| -35.91, |
| -33.77, |
| -85.21, |
| -14.52, |
| -23.31, |
| -27.36, |
| -63.51, |
| -61.59, |
| -13.6, |
| -20.96, |
| -33.99, |
| -31.92, |
| -43.96, |
| -26.19, |
| -20.55, |
| -23.28, |
| -25.26, |
| -23.79, |
| -76.02 |
| ] |
| }, |
| "timbral_proxies": { |
| "spectral_centroid_hz_median": 1542.652, |
| "spectral_flatness_median": 0.006 |
| }, |
| "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": "Теперь. Я не шучу. Поворачивайте назад, иначе я вас убью.", |
| "text_chars_per_sec": 13.988 |
| }, |
| "pipeline_control": { |
| "tts_text": "Теперь. Я не шучу. Поворачивайте назад, иначе я вас убью.", |
| "duration_stage": "post", |
| "speaker_reference": "0072" |
| } |
| } |
|
|
