| { |
| "file": "references/0133.wav", |
| "sample_rate_for_analysis": 16000, |
| "duration_sec": 4.215, |
| "voiced_fraction": 0.602, |
| "pause_map": [ |
| { |
| "start": 1.42, |
| "end": 1.69, |
| "duration": 0.27 |
| }, |
| { |
| "start": 2.65, |
| "end": 2.78, |
| "duration": 0.13 |
| } |
| ], |
| "f0_hz": { |
| "median": 183.848, |
| "p10_p90": [ |
| 145.92, |
| 217.374 |
| ], |
| "contour_20_points": [ |
| 198.18, |
| 198.18, |
| 203.25, |
| 207.56, |
| 164.74, |
| 171.45, |
| 147.62, |
| 175.89, |
| 217.37, |
| 205.36, |
| 185.98, |
| 208.76, |
| 192.44, |
| 178.61, |
| 213.85, |
| 209.75, |
| 162.41, |
| 146.77, |
| 145.08, |
| 145.92 |
| ] |
| }, |
| "energy_db": { |
| "p10_p50_p90": [ |
| -66.622, |
| -23.924, |
| -12.704 |
| ], |
| "dynamic_range_db": 53.918, |
| "contour_20_points": [ |
| -74.92, |
| -58.69, |
| -13.72, |
| -15.71, |
| -20.07, |
| -33.74, |
| -18.28, |
| -67.47, |
| -12.21, |
| -22.11, |
| -20.83, |
| -13.17, |
| -52.53, |
| -24.95, |
| -21.48, |
| -25.96, |
| -23.08, |
| -22.93, |
| -27.0, |
| -72.99 |
| ] |
| }, |
| "timbral_proxies": { |
| "spectral_centroid_hz_median": 2968.606, |
| "spectral_flatness_median": 0.011 |
| }, |
| "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": "Я не злюсь. Просто возвращайтесь в тестовую зону.", |
| "text_chars_per_sec": 11.626, |
| "source": { |
| "file": "sources_silero/0133.wav", |
| "sample_rate_for_analysis": 16000, |
| "duration_sec": 3.112, |
| "voiced_fraction": 0.644, |
| "pause_map": [ |
| { |
| "start": 0.9, |
| "end": 1.19, |
| "duration": 0.29 |
| } |
| ], |
| "f0_hz": { |
| "median": 212.409, |
| "p10_p90": [ |
| 165.693, |
| 248.259 |
| ], |
| "contour_20_points": [ |
| 254.06, |
| 254.06, |
| 240.86, |
| 223.37, |
| 184.81, |
| 176.4, |
| 206.07, |
| 235.75, |
| 255.85, |
| 241.98, |
| 212.41, |
| 213.31, |
| 237.05, |
| 228.58, |
| 206.36, |
| 217.33, |
| 185.87, |
| 169.59, |
| 160.98, |
| 197.04 |
| ] |
| }, |
| "energy_db": { |
| "p10_p50_p90": [ |
| -73.928, |
| -23.387, |
| -14.01 |
| ], |
| "dynamic_range_db": 59.917, |
| "contour_20_points": [ |
| -79.17, |
| -68.08, |
| -11.01, |
| -17.98, |
| -16.14, |
| -31.13, |
| -71.26, |
| -75.95, |
| -13.01, |
| -17.0, |
| -16.34, |
| -18.33, |
| -17.35, |
| -26.52, |
| -16.08, |
| -24.46, |
| -18.43, |
| -21.71, |
| -27.6, |
| -64.8 |
| ] |
| }, |
| "timbral_proxies": { |
| "spectral_centroid_hz_median": 1621.386, |
| "spectral_flatness_median": 0.011 |
| }, |
| "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": "Я не злюсь. Просто возвращайтесь в тестовую зону.", |
| "text_chars_per_sec": 15.743 |
| }, |
| "pipeline_control": { |
| "tts_text": "Я не злюсь. Просто возвращайтесь в тестовую зону.", |
| "duration_stage": "pre", |
| "speaker_reference": "0133" |
| } |
| } |
|
|
