| { |
| "file": "references/0007.wav", |
| "sample_rate_for_analysis": 16000, |
| "duration_sec": 5.438, |
| "voiced_fraction": 0.665, |
| "pause_map": [ |
| { |
| "start": 0.86, |
| "end": 1.7, |
| "duration": 0.84 |
| } |
| ], |
| "f0_hz": { |
| "median": 179.131, |
| "p10_p90": [ |
| 137.73, |
| 235.683 |
| ], |
| "contour_20_points": [ |
| 207.56, |
| 183.64, |
| 137.73, |
| 78.94, |
| 84.14, |
| 172.44, |
| 255.27, |
| 178.21, |
| 184.91, |
| 186.56, |
| 178.92, |
| 201.65, |
| 154.6, |
| 185.98, |
| 180.05, |
| 196.68, |
| 171.27, |
| 224.87, |
| 227.9, |
| 228.97 |
| ] |
| }, |
| "energy_db": { |
| "p10_p50_p90": [ |
| -61.677, |
| -14.105, |
| -7.75 |
| ], |
| "dynamic_range_db": 53.927, |
| "contour_20_points": [ |
| -66.05, |
| -26.37, |
| -11.38, |
| -40.81, |
| -63.92, |
| -67.19, |
| -13.76, |
| -14.77, |
| -6.73, |
| -20.79, |
| -29.62, |
| -10.4, |
| -12.51, |
| -10.93, |
| -18.11, |
| -7.98, |
| -12.72, |
| -26.29, |
| -21.2, |
| -29.78 |
| ] |
| }, |
| "timbral_proxies": { |
| "spectral_centroid_hz_median": 1699.258, |
| "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": "Отлично. После каждого теста проходите в шлюзовую камеру.", |
| "text_chars_per_sec": 10.482, |
| "source": { |
| "file": "sources_silero/0007.wav", |
| "sample_rate_for_analysis": 16000, |
| "duration_sec": 4.125, |
| "voiced_fraction": 0.69, |
| "pause_map": [ |
| { |
| "start": 0.69, |
| "end": 1.0, |
| "duration": 0.31 |
| }, |
| { |
| "start": 2.2, |
| "end": 2.5, |
| "duration": 0.3 |
| } |
| ], |
| "f0_hz": { |
| "median": 208.76, |
| "p10_p90": [ |
| 172.53, |
| 259.709 |
| ], |
| "contour_20_points": [ |
| 232.98, |
| 247.63, |
| 196.53, |
| 189.82, |
| 208.65, |
| 215.56, |
| 252.48, |
| 231.46, |
| 207.63, |
| 173.49, |
| 182.9, |
| 219.51, |
| 209.91, |
| 275.84, |
| 227.65, |
| 197.39, |
| 211.96, |
| 176.4, |
| 171.24, |
| 200.49 |
| ] |
| }, |
| "energy_db": { |
| "p10_p50_p90": [ |
| -65.927, |
| -24.735, |
| -15.671 |
| ], |
| "dynamic_range_db": 50.256, |
| "contour_20_points": [ |
| -85.74, |
| -27.73, |
| -19.09, |
| -29.61, |
| -63.76, |
| -19.11, |
| -27.93, |
| -24.32, |
| -35.39, |
| -30.89, |
| -34.38, |
| -73.77, |
| -13.86, |
| -19.52, |
| -26.67, |
| -24.29, |
| -17.99, |
| -21.04, |
| -26.7, |
| -77.69 |
| ] |
| }, |
| "timbral_proxies": { |
| "spectral_centroid_hz_median": 1607.59, |
| "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": "Отлично. После каждого теста. Проходите в шлюзовую камеру.", |
| "text_chars_per_sec": 14.061 |
| }, |
| "pipeline_control": { |
| "tts_text": "Отлично. После каждого теста. Проходите в шлюзовую камеру.", |
| "duration_stage": "pre", |
| "speaker_reference": "0007" |
| } |
| } |
|
|
