Random118's picture
Add Russian CosyVoice 3 dubbing pack and audio gallery
8c2913a verified
Raw History Blame Contribute Delete
4.51 kB
{
"file": "references/0109.wav",
"sample_rate_for_analysis": 16000,
"duration_sec": 11.724,
"voiced_fraction": 0.782,
"pause_map": [
{
"start": 1.16,
"end": 1.68,
"duration": 0.52
},
{
"start": 2.22,
"end": 2.86,
"duration": 0.64
},
{
"start": 3.19,
"end": 3.32,
"duration": 0.13
},
{
"start": 5.13,
"end": 5.49,
"duration": 0.36
},
{
"start": 6.66,
"end": 6.76,
"duration": 0.1
}
],
"f0_hz": {
"median": 194.78,
"p10_p90": [
148.471,
258.503
],
"contour_20_points": [
270.73,
201.52,
162.66,
263.02,
131.28,
217.37,
219.9,
227.58,
285.17,
193.35,
207.26,
173.08,
198.18,
194.78,
174.53,
219.9,
184.91,
174.53,
166.24,
145.08
]
},
"energy_db": {
"p10_p50_p90": [
-55.729,
-15.73,
-9.733
],
"dynamic_range_db": 45.996,
"contour_20_points": [
-72.95,
-9.65,
-57.44,
-7.82,
-57.31,
-12.03,
-18.53,
-22.83,
-17.24,
-18.07,
-37.39,
-16.64,
-12.29,
-9.31,
-11.76,
-10.33,
-11.75,
-9.75,
-18.17,
-69.14
]
},
"timbral_proxies": {
"spectral_centroid_hz_median": 1729.692,
"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": "Что вы делаете? Прекратите! Я... я... Мы рады, что вы прошли последнее испытание, в котором мы притворялись, будто собираемся вас убить.",
"text_chars_per_sec": 11.6,
"source": {
"file": "sources_silero/0109.wav",
"sample_rate_for_analysis": 16000,
"duration_sec": 8.025,
"voiced_fraction": 0.733,
"pause_map": [
{
"start": 1.18,
"end": 1.41,
"duration": 0.23
},
{
"start": 2.11,
"end": 2.2,
"duration": 0.09
},
{
"start": 2.67,
"end": 3.11,
"duration": 0.44
}
],
"f0_hz": {
"median": 221.173,
"p10_p90": [
189.235,
267.619
],
"contour_20_points": [
298.66,
296.94,
211.9,
203.29,
280.53,
202.18,
222.66,
242.51,
227.79,
244.93,
302.78,
209.72,
211.77,
213.45,
212.02,
170.65,
228.7,
241.15,
197.88,
165.69
]
},
"energy_db": {
"p10_p50_p90": [
-54.094,
-25.398,
-16.436
],
"dynamic_range_db": 37.657,
"contour_20_points": [
-77.79,
-16.71,
-16.67,
-63.98,
-28.9,
-46.69,
-18.08,
-54.21,
-14.4,
-19.76,
-22.33,
-19.26,
-18.9,
-24.18,
-21.29,
-26.47,
-33.58,
-16.27,
-25.21,
-67.27
]
},
"timbral_proxies": {
"spectral_centroid_hz_median": 1800.431,
"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": "Что вы делаете? Прекратите! Я... я... Мы рады, что вы прошли последнее испытание, в котором мы притворялись, будто собираемся вас убить.",
"text_chars_per_sec": 16.947
},
"pipeline_control": {
"tts_text": "Что вы делаете? Прекратите! Я... я... Мы рады, что вы прошли последнее испытание, в котором мы притворялись, будто собираемся вас убить.",
"duration_stage": "pre",
"speaker_reference": "0109"
}
}