| { |
| "file": "references/0000.wav", |
| "sample_rate_for_analysis": 16000, |
| "duration_sec": 5.905, |
| "voiced_fraction": 0.802, |
| "pause_map": [ |
| { |
| "start": 0.9, |
| "end": 1.02, |
| "duration": 0.12 |
| }, |
| { |
| "start": 1.39, |
| "end": 1.57, |
| "duration": 0.18 |
| } |
| ], |
| "f0_hz": { |
| "median": 176.562, |
| "p10_p90": [ |
| 147.616, |
| 208.76 |
| ], |
| "contour_20_points": [ |
| 160.98, |
| 162.05, |
| 155.49, |
| 163.16, |
| 174.22, |
| 195.22, |
| 146.77, |
| 175.1, |
| 209.2, |
| 177.59, |
| 184.91, |
| 179.69, |
| 208.76, |
| 183.59, |
| 185.98, |
| 175.55, |
| 184.13, |
| 174.53, |
| 147.62, |
| 139.33 |
| ] |
| }, |
| "energy_db": { |
| "p10_p50_p90": [ |
| -41.24, |
| -12.628, |
| -8.596 |
| ], |
| "dynamic_range_db": 32.645, |
| "contour_20_points": [ |
| -63.95, |
| -37.65, |
| -10.4, |
| -51.71, |
| -8.75, |
| -49.89, |
| -10.58, |
| -19.8, |
| -13.29, |
| -10.03, |
| -10.99, |
| -16.55, |
| -8.74, |
| -13.35, |
| -13.51, |
| -14.64, |
| -6.86, |
| -7.47, |
| -11.27, |
| -28.92 |
| ] |
| }, |
| "timbral_proxies": { |
| "spectral_centroid_hz_median": 1468.627, |
| "spectral_flatness_median": 0.002 |
| }, |
| "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": "И снова здравствуйте. Добро пожаловать в компьютеризированный центр обогащения «Апертур Наука».", |
| "text_chars_per_sec": 16.088, |
| "source": { |
| "file": "sources_silero/0000.wav", |
| "sample_rate_for_analysis": 16000, |
| "duration_sec": 5.0, |
| "voiced_fraction": 0.737, |
| "pause_map": [ |
| { |
| "start": 1.2, |
| "end": 1.45, |
| "duration": 0.25 |
| } |
| ], |
| "f0_hz": { |
| "median": 205.174, |
| "p10_p90": [ |
| 179.649, |
| 243.994 |
| ], |
| "contour_20_points": [ |
| 249.7, |
| 264.14, |
| 228.79, |
| 174.61, |
| 231.21, |
| 227.45, |
| 208.76, |
| 221.92, |
| 196.94, |
| 200.31, |
| 212.22, |
| 229.22, |
| 209.98, |
| 192.54, |
| 182.19, |
| 190.91, |
| 197.21, |
| 186.95, |
| 170.55, |
| 170.55 |
| ] |
| }, |
| "energy_db": { |
| "p10_p50_p90": [ |
| -44.987, |
| -21.942, |
| -15.888 |
| ], |
| "dynamic_range_db": 29.099, |
| "contour_20_points": [ |
| -86.81, |
| -32.16, |
| -15.78, |
| -30.65, |
| -32.34, |
| -75.88, |
| -14.97, |
| -14.96, |
| -19.04, |
| -24.73, |
| -17.52, |
| -17.45, |
| -33.13, |
| -20.73, |
| -22.71, |
| -18.76, |
| -40.35, |
| -18.17, |
| -34.33, |
| -83.28 |
| ] |
| }, |
| "timbral_proxies": { |
| "spectral_centroid_hz_median": 1564.514, |
| "spectral_flatness_median": 0.007 |
| }, |
| "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": "И снова здравствуйте. Добро пожаловать в компьютеризированный центр обогащения «Апертур Наука».", |
| "text_chars_per_sec": 19.0 |
| }, |
| "pipeline_control": { |
| "tts_text": "И снова здравствуйте. Добро пожаловать в компьютеризированный центр обогащения «Апертур Наука».", |
| "duration_stage": "pre", |
| "speaker_reference": "0000" |
| } |
| } |
|
|
