Random118's picture
Add Russian CosyVoice 3 dubbing pack and audio gallery
8c2913a verified
Raw History Blame Contribute Delete
4.18 kB
{
"file": "references/0025.wav",
"sample_rate_for_analysis": 16000,
"duration_sec": 6.101,
"voiced_fraction": 0.81,
"pause_map": [
{
"start": 0.98,
"end": 1.07,
"duration": 0.09
},
{
"start": 1.64,
"end": 1.72,
"duration": 0.08
}
],
"f0_hz": {
"median": 169.566,
"p10_p90": [
139.331,
185.984
],
"contour_20_points": [
216.12,
216.12,
142.4,
138.02,
185.98,
157.02,
175.55,
173.74,
178.72,
165.69,
164.86,
164.51,
152.56,
137.15,
174.96,
175.55,
164.93,
179.65,
131.51,
130.0
]
},
"energy_db": {
"p10_p50_p90": [
-43.413,
-11.373,
-7.799
],
"dynamic_range_db": 35.614,
"contour_20_points": [
-73.8,
-54.09,
-10.86,
-28.03,
-8.46,
-17.55,
-10.84,
-10.41,
-9.4,
-5.91,
-9.3,
-8.26,
-14.27,
-26.77,
-17.65,
-8.2,
-12.01,
-10.17,
-11.02,
-60.44
]
},
"timbral_proxies": {
"spectral_centroid_hz_median": 1699.878,
"spectral_flatness_median": 0.005
},
"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! Запомните: день «Приведи дочь на работу» в «Апертур Наука» — идеальное время, чтобы её протестировали.",
"text_chars_per_sec": 18.193,
"source": {
"file": "sources_silero/0025.wav",
"sample_rate_for_analysis": 16000,
"duration_sec": 7.338,
"voiced_fraction": 0.749,
"pause_map": [
{
"start": 1.0,
"end": 1.28,
"duration": 0.28
},
{
"start": 2.01,
"end": 2.22,
"duration": 0.21
},
{
"start": 3.15,
"end": 3.56,
"duration": 0.41
}
],
"f0_hz": {
"median": 216.122,
"p10_p90": [
185.877,
286.826
],
"contour_20_points": [
208.76,
237.05,
247.65,
214.44,
247.58,
187.16,
274.77,
291.86,
205.75,
241.4,
288.37,
213.14,
266.08,
200.65,
219.47,
208.76,
203.87,
202.29,
173.27,
216.12
]
},
"energy_db": {
"p10_p50_p90": [
-56.4,
-24.917,
-16.008
],
"dynamic_range_db": 40.393,
"contour_20_points": [
-82.3,
-13.67,
-27.86,
-58.67,
-14.5,
-32.17,
-17.59,
-25.22,
-31.82,
-54.08,
-15.02,
-18.37,
-15.62,
-37.23,
-13.95,
-18.09,
-23.13,
-31.1,
-27.38,
-69.21
]
},
"timbral_proxies": {
"spectral_centroid_hz_median": 1641.73,
"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": "Отлично! Запомните: день «Приведи дочь... на работу» в «Апертур Наука» — идеальное время, чтобы её протестировали.",
"text_chars_per_sec": 15.537
},
"pipeline_control": {
"tts_text": "Отлично! Запомните: день «Приведи дочь... на работу» в «Апертур Наука» — идеальное время, чтобы её протестировали.",
"duration_stage": "post",
"speaker_reference": "0025"
}
}