| { |
| "file": "references/0053.wav", |
| "sample_rate_for_analysis": 16000, |
| "duration_sec": 3.975, |
| "voiced_fraction": 0.746, |
| "pause_map": [ |
| { |
| "start": 0.98, |
| "end": 1.07, |
| "duration": 0.09 |
| }, |
| { |
| "start": 2.61, |
| "end": 2.74, |
| "duration": 0.13 |
| } |
| ], |
| "f0_hz": { |
| "median": 184.913, |
| "p10_p90": [ |
| 145.92, |
| 226.343 |
| ], |
| "contour_20_points": [ |
| 211.19, |
| 207.65, |
| 184.01, |
| 167.16, |
| 155.21, |
| 187.86, |
| 200.29, |
| 232.98, |
| 174.53, |
| 174.53, |
| 211.81, |
| 200.53, |
| 139.33, |
| 189.3, |
| 218.63, |
| 220.12, |
| 183.05, |
| 175.55, |
| 146.26, |
| 145.92 |
| ] |
| }, |
| "energy_db": { |
| "p10_p50_p90": [ |
| -38.087, |
| -14.081, |
| -8.765 |
| ], |
| "dynamic_range_db": 29.323, |
| "contour_20_points": [ |
| -47.24, |
| -7.82, |
| -9.88, |
| -16.44, |
| -24.95, |
| -60.16, |
| -32.02, |
| -8.8, |
| -10.47, |
| -9.04, |
| -17.43, |
| -14.03, |
| -17.41, |
| -59.54, |
| -11.28, |
| -7.68, |
| -21.7, |
| -10.56, |
| -17.41, |
| -56.42 |
| ] |
| }, |
| "timbral_proxies": { |
| "spectral_centroid_hz_median": 1612.598, |
| "spectral_flatness_median": 0.003 |
| }, |
| "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": "Например, этот пол вас убьёт. Постарайтесь его избегать.", |
| "text_chars_per_sec": 14.087, |
| "source": { |
| "file": "sources_silero/0053.wav", |
| "sample_rate_for_analysis": 16000, |
| "duration_sec": 4.312, |
| "voiced_fraction": 0.604, |
| "pause_map": [ |
| { |
| "start": 1.42, |
| "end": 1.89, |
| "duration": 0.47 |
| }, |
| { |
| "start": 2.47, |
| "end": 2.88, |
| "duration": 0.41 |
| } |
| ], |
| "f0_hz": { |
| "median": 212.409, |
| "p10_p90": [ |
| 167.618, |
| 243.994 |
| ], |
| "contour_20_points": [ |
| 232.98, |
| 220.18, |
| 224.53, |
| 323.99, |
| 218.9, |
| 221.8, |
| 167.62, |
| 220.89, |
| 267.9, |
| 221.17, |
| 190.36, |
| 156.29, |
| 188.57, |
| 207.56, |
| 232.03, |
| 194.78, |
| 178.78, |
| 176.5, |
| 153.71, |
| 153.71 |
| ] |
| }, |
| "energy_db": { |
| "p10_p50_p90": [ |
| -71.827, |
| -27.349, |
| -15.634 |
| ], |
| "dynamic_range_db": 56.192, |
| "contour_20_points": [ |
| -69.31, |
| -19.04, |
| -14.2, |
| -20.66, |
| -16.41, |
| -43.35, |
| -27.1, |
| -63.79, |
| -76.95, |
| -19.9, |
| -23.11, |
| -45.52, |
| -85.65, |
| -21.58, |
| -12.87, |
| -25.82, |
| -22.73, |
| -26.78, |
| -36.45, |
| -84.07 |
| ] |
| }, |
| "timbral_proxies": { |
| "spectral_centroid_hz_median": 1422.117, |
| "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": "Например, этот пол... вас убьёт. Постарайтесь его избегать.", |
| "text_chars_per_sec": 13.681 |
| }, |
| "pipeline_control": { |
| "tts_text": "Например, этот пол... вас убьёт. Постарайтесь его избегать.", |
| "duration_stage": "pre", |
| "speaker_reference": "0053" |
| } |
| } |
|
|
