ugcinc 4.22.0 → 4.23.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -528,7 +528,7 @@ export type ScenesResult = z.infer<typeof scenesResultSchema>;
528
528
  export type MediaAnalysisResult = z.infer<typeof mediaAnalysisResultSchema>;
529
529
  export interface MediaAnalysisJob {
530
530
  jobId: string;
531
- operation: 'extract_frames' | 'detect_scenes' | 'read_image_text' | 'read_video_text';
531
+ operation: 'extract_frames' | 'detect_scenes' | 'read_image_text' | 'read_video_text' | 'transcribe_audio' | 'measure_audio_levels';
532
532
  status: 'pending' | 'processing' | 'completed' | 'failed';
533
533
  progress: number;
534
534
  result?: MediaAnalysisResult;
@@ -1684,6 +1684,812 @@ export declare const videoTextResultSchema: z.ZodObject<{
1684
1684
  }>;
1685
1685
  export type ImageTextResult = z.infer<typeof imageTextResultSchema>;
1686
1686
  export type VideoTextResult = z.infer<typeof videoTextResultSchema>;
1687
+ export declare const transcribeAudioSchema: z.ZodObject<{
1688
+ url: z.ZodEffects<z.ZodString, string, string>;
1689
+ range: z.ZodOptional<z.ZodEffects<z.ZodObject<{
1690
+ startMs: z.ZodNumber;
1691
+ endMs: z.ZodNumber;
1692
+ }, "strict", z.ZodTypeAny, {
1693
+ startMs: number;
1694
+ endMs: number;
1695
+ }, {
1696
+ startMs: number;
1697
+ endMs: number;
1698
+ }>, {
1699
+ startMs: number;
1700
+ endMs: number;
1701
+ }, {
1702
+ startMs: number;
1703
+ endMs: number;
1704
+ }>>;
1705
+ language: z.ZodDefault<z.ZodEnum<["auto", "en", "es", "fr", "de", "it", "pt", "ja", "ko", "zh", "ar", "hi", "ru"]>>;
1706
+ timing: z.ZodDefault<z.ZodEnum<["word", "segment"]>>;
1707
+ channel: z.ZodOptional<z.ZodNumber>;
1708
+ }, "strict", z.ZodTypeAny, {
1709
+ url: string;
1710
+ timing: "word" | "segment";
1711
+ language: "auto" | "en" | "es" | "fr" | "de" | "it" | "pt" | "ja" | "ko" | "zh" | "ar" | "hi" | "ru";
1712
+ range?: {
1713
+ startMs: number;
1714
+ endMs: number;
1715
+ } | undefined;
1716
+ channel?: number | undefined;
1717
+ }, {
1718
+ url: string;
1719
+ range?: {
1720
+ startMs: number;
1721
+ endMs: number;
1722
+ } | undefined;
1723
+ timing?: "word" | "segment" | undefined;
1724
+ language?: "auto" | "en" | "es" | "fr" | "de" | "it" | "pt" | "ja" | "ko" | "zh" | "ar" | "hi" | "ru" | undefined;
1725
+ channel?: number | undefined;
1726
+ }>;
1727
+ export declare const measureAudioLevelsSchema: z.ZodObject<{
1728
+ url: z.ZodEffects<z.ZodString, string, string>;
1729
+ range: z.ZodOptional<z.ZodEffects<z.ZodObject<{
1730
+ startMs: z.ZodNumber;
1731
+ endMs: z.ZodNumber;
1732
+ }, "strict", z.ZodTypeAny, {
1733
+ startMs: number;
1734
+ endMs: number;
1735
+ }, {
1736
+ startMs: number;
1737
+ endMs: number;
1738
+ }>, {
1739
+ startMs: number;
1740
+ endMs: number;
1741
+ }, {
1742
+ startMs: number;
1743
+ endMs: number;
1744
+ }>>;
1745
+ windowMs: z.ZodDefault<z.ZodNumber>;
1746
+ silenceThresholdDbfs: z.ZodDefault<z.ZodNumber>;
1747
+ }, "strict", z.ZodTypeAny, {
1748
+ url: string;
1749
+ windowMs: number;
1750
+ silenceThresholdDbfs: number;
1751
+ range?: {
1752
+ startMs: number;
1753
+ endMs: number;
1754
+ } | undefined;
1755
+ }, {
1756
+ url: string;
1757
+ range?: {
1758
+ startMs: number;
1759
+ endMs: number;
1760
+ } | undefined;
1761
+ windowMs?: number | undefined;
1762
+ silenceThresholdDbfs?: number | undefined;
1763
+ }>;
1764
+ export type TranscribeAudioParams = z.input<typeof transcribeAudioSchema>;
1765
+ export type MeasureAudioLevelsParams = z.input<typeof measureAudioLevelsSchema>;
1766
+ export declare const audioLevelsResultSchema: z.ZodObject<{
1767
+ version: z.ZodLiteral<1>;
1768
+ operation: z.ZodLiteral<"measure_audio_levels">;
1769
+ source: z.ZodObject<{
1770
+ sha256: z.ZodString;
1771
+ bytes: z.ZodNumber;
1772
+ hasAudio: z.ZodBoolean;
1773
+ durationMs: z.ZodNumber;
1774
+ originStartMs: z.ZodNumber;
1775
+ audioStartPts: z.ZodNullable<z.ZodString>;
1776
+ audioTimeBase: z.ZodNullable<z.ZodString>;
1777
+ sampleRate: z.ZodNullable<z.ZodNumber>;
1778
+ channels: z.ZodNullable<z.ZodNumber>;
1779
+ channelLayout: z.ZodNullable<z.ZodString>;
1780
+ audioOffsetMs: z.ZodNullable<z.ZodNumber>;
1781
+ }, "strip", z.ZodTypeAny, {
1782
+ durationMs: number;
1783
+ sha256: string;
1784
+ bytes: number;
1785
+ hasAudio: boolean;
1786
+ originStartMs: number;
1787
+ audioStartPts: string | null;
1788
+ audioTimeBase: string | null;
1789
+ sampleRate: number | null;
1790
+ channels: number | null;
1791
+ channelLayout: string | null;
1792
+ audioOffsetMs: number | null;
1793
+ }, {
1794
+ durationMs: number;
1795
+ sha256: string;
1796
+ bytes: number;
1797
+ hasAudio: boolean;
1798
+ originStartMs: number;
1799
+ audioStartPts: string | null;
1800
+ audioTimeBase: string | null;
1801
+ sampleRate: number | null;
1802
+ channels: number | null;
1803
+ channelLayout: string | null;
1804
+ audioOffsetMs: number | null;
1805
+ }>;
1806
+ status: z.ZodEnum<["no_audio", "measured"]>;
1807
+ coverage: z.ZodObject<{
1808
+ requestedRange: z.ZodObject<{
1809
+ startMs: z.ZodNumber;
1810
+ endMs: z.ZodNumber;
1811
+ }, "strip", z.ZodTypeAny, {
1812
+ startMs: number;
1813
+ endMs: number;
1814
+ }, {
1815
+ startMs: number;
1816
+ endMs: number;
1817
+ }>;
1818
+ analyzed: z.ZodArray<z.ZodObject<{
1819
+ startMs: z.ZodNumber;
1820
+ endMs: z.ZodNumber;
1821
+ }, "strip", z.ZodTypeAny, {
1822
+ startMs: number;
1823
+ endMs: number;
1824
+ }, {
1825
+ startMs: number;
1826
+ endMs: number;
1827
+ }>, "many">;
1828
+ withoutAudio: z.ZodArray<z.ZodObject<{
1829
+ startMs: z.ZodNumber;
1830
+ endMs: z.ZodNumber;
1831
+ }, "strip", z.ZodTypeAny, {
1832
+ startMs: number;
1833
+ endMs: number;
1834
+ }, {
1835
+ startMs: number;
1836
+ endMs: number;
1837
+ }>, "many">;
1838
+ sampleRoundingToleranceMs: z.ZodNumber;
1839
+ }, "strip", z.ZodTypeAny, {
1840
+ requestedRange: {
1841
+ startMs: number;
1842
+ endMs: number;
1843
+ };
1844
+ analyzed: {
1845
+ startMs: number;
1846
+ endMs: number;
1847
+ }[];
1848
+ withoutAudio: {
1849
+ startMs: number;
1850
+ endMs: number;
1851
+ }[];
1852
+ sampleRoundingToleranceMs: number;
1853
+ }, {
1854
+ requestedRange: {
1855
+ startMs: number;
1856
+ endMs: number;
1857
+ };
1858
+ analyzed: {
1859
+ startMs: number;
1860
+ endMs: number;
1861
+ }[];
1862
+ withoutAudio: {
1863
+ startMs: number;
1864
+ endMs: number;
1865
+ }[];
1866
+ sampleRoundingToleranceMs: number;
1867
+ }>;
1868
+ channels: z.ZodArray<z.ZodObject<{
1869
+ index: z.ZodNumber;
1870
+ rmsDbfs: z.ZodNullable<z.ZodNumber>;
1871
+ samplePeakDbfs: z.ZodNullable<z.ZodNumber>;
1872
+ potentialClippedSamples: z.ZodNumber;
1873
+ } & {
1874
+ truePeakDbtp: z.ZodNullable<z.ZodNumber>;
1875
+ sampleCount: z.ZodNumber;
1876
+ intersampleOvershoot: z.ZodBoolean;
1877
+ }, "strip", z.ZodTypeAny, {
1878
+ index: number;
1879
+ rmsDbfs: number | null;
1880
+ samplePeakDbfs: number | null;
1881
+ potentialClippedSamples: number;
1882
+ truePeakDbtp: number | null;
1883
+ sampleCount: number;
1884
+ intersampleOvershoot: boolean;
1885
+ }, {
1886
+ index: number;
1887
+ rmsDbfs: number | null;
1888
+ samplePeakDbfs: number | null;
1889
+ potentialClippedSamples: number;
1890
+ truePeakDbtp: number | null;
1891
+ sampleCount: number;
1892
+ intersampleOvershoot: boolean;
1893
+ }>, "many">;
1894
+ windows: z.ZodArray<z.ZodObject<{
1895
+ startMs: z.ZodNumber;
1896
+ endMs: z.ZodNumber;
1897
+ } & {
1898
+ channels: z.ZodArray<z.ZodObject<{
1899
+ index: z.ZodNumber;
1900
+ rmsDbfs: z.ZodNullable<z.ZodNumber>;
1901
+ samplePeakDbfs: z.ZodNullable<z.ZodNumber>;
1902
+ potentialClippedSamples: z.ZodNumber;
1903
+ }, "strip", z.ZodTypeAny, {
1904
+ index: number;
1905
+ rmsDbfs: number | null;
1906
+ samplePeakDbfs: number | null;
1907
+ potentialClippedSamples: number;
1908
+ }, {
1909
+ index: number;
1910
+ rmsDbfs: number | null;
1911
+ samplePeakDbfs: number | null;
1912
+ potentialClippedSamples: number;
1913
+ }>, "many">;
1914
+ }, "strip", z.ZodTypeAny, {
1915
+ startMs: number;
1916
+ endMs: number;
1917
+ channels: {
1918
+ index: number;
1919
+ rmsDbfs: number | null;
1920
+ samplePeakDbfs: number | null;
1921
+ potentialClippedSamples: number;
1922
+ }[];
1923
+ }, {
1924
+ startMs: number;
1925
+ endMs: number;
1926
+ channels: {
1927
+ index: number;
1928
+ rmsDbfs: number | null;
1929
+ samplePeakDbfs: number | null;
1930
+ potentialClippedSamples: number;
1931
+ }[];
1932
+ }>, "many">;
1933
+ silence: z.ZodArray<z.ZodObject<{
1934
+ startMs: z.ZodNumber;
1935
+ endMs: z.ZodNumber;
1936
+ }, "strip", z.ZodTypeAny, {
1937
+ startMs: number;
1938
+ endMs: number;
1939
+ }, {
1940
+ startMs: number;
1941
+ endMs: number;
1942
+ }>, "many">;
1943
+ integratedLufs: z.ZodNullable<z.ZodNumber>;
1944
+ truePeakDbtp: z.ZodNullable<z.ZodNumber>;
1945
+ loudnessTimeline: z.ZodArray<z.ZodObject<{
1946
+ startMs: z.ZodNumber;
1947
+ endMs: z.ZodNumber;
1948
+ } & {
1949
+ momentaryLufs: z.ZodNullable<z.ZodNumber>;
1950
+ shortTermLufs: z.ZodNullable<z.ZodNumber>;
1951
+ }, "strip", z.ZodTypeAny, {
1952
+ startMs: number;
1953
+ endMs: number;
1954
+ momentaryLufs: number | null;
1955
+ shortTermLufs: number | null;
1956
+ }, {
1957
+ startMs: number;
1958
+ endMs: number;
1959
+ momentaryLufs: number | null;
1960
+ shortTermLufs: number | null;
1961
+ }>, "many">;
1962
+ settings: z.ZodObject<{
1963
+ windowMs: z.ZodNumber;
1964
+ silenceThresholdDbfs: z.ZodNumber;
1965
+ silenceMethod: z.ZodLiteral<"all-channel-window-rms">;
1966
+ potentialClippingThreshold: z.ZodNumber;
1967
+ channelPolicy: z.ZodLiteral<"native-separate">;
1968
+ loudnessMethod: z.ZodLiteral<"FFmpeg-ebur128-native-channels">;
1969
+ ffmpegVersion: z.ZodString;
1970
+ }, "strip", z.ZodTypeAny, {
1971
+ channelPolicy: "native-separate";
1972
+ windowMs: number;
1973
+ silenceThresholdDbfs: number;
1974
+ silenceMethod: "all-channel-window-rms";
1975
+ potentialClippingThreshold: number;
1976
+ loudnessMethod: "FFmpeg-ebur128-native-channels";
1977
+ ffmpegVersion: string;
1978
+ }, {
1979
+ channelPolicy: "native-separate";
1980
+ windowMs: number;
1981
+ silenceThresholdDbfs: number;
1982
+ silenceMethod: "all-channel-window-rms";
1983
+ potentialClippingThreshold: number;
1984
+ loudnessMethod: "FFmpeg-ebur128-native-channels";
1985
+ ffmpegVersion: string;
1986
+ }>;
1987
+ warnings: z.ZodArray<z.ZodString, "many">;
1988
+ previewUrl: z.ZodOptional<z.ZodString>;
1989
+ manifestUrl: z.ZodString;
1990
+ }, "strip", z.ZodTypeAny, {
1991
+ status: "no_audio" | "measured";
1992
+ version: 1;
1993
+ operation: "measure_audio_levels";
1994
+ source: {
1995
+ durationMs: number;
1996
+ sha256: string;
1997
+ bytes: number;
1998
+ hasAudio: boolean;
1999
+ originStartMs: number;
2000
+ audioStartPts: string | null;
2001
+ audioTimeBase: string | null;
2002
+ sampleRate: number | null;
2003
+ channels: number | null;
2004
+ channelLayout: string | null;
2005
+ audioOffsetMs: number | null;
2006
+ };
2007
+ manifestUrl: string;
2008
+ settings: {
2009
+ channelPolicy: "native-separate";
2010
+ windowMs: number;
2011
+ silenceThresholdDbfs: number;
2012
+ silenceMethod: "all-channel-window-rms";
2013
+ potentialClippingThreshold: number;
2014
+ loudnessMethod: "FFmpeg-ebur128-native-channels";
2015
+ ffmpegVersion: string;
2016
+ };
2017
+ warnings: string[];
2018
+ coverage: {
2019
+ requestedRange: {
2020
+ startMs: number;
2021
+ endMs: number;
2022
+ };
2023
+ analyzed: {
2024
+ startMs: number;
2025
+ endMs: number;
2026
+ }[];
2027
+ withoutAudio: {
2028
+ startMs: number;
2029
+ endMs: number;
2030
+ }[];
2031
+ sampleRoundingToleranceMs: number;
2032
+ };
2033
+ channels: {
2034
+ index: number;
2035
+ rmsDbfs: number | null;
2036
+ samplePeakDbfs: number | null;
2037
+ potentialClippedSamples: number;
2038
+ truePeakDbtp: number | null;
2039
+ sampleCount: number;
2040
+ intersampleOvershoot: boolean;
2041
+ }[];
2042
+ truePeakDbtp: number | null;
2043
+ windows: {
2044
+ startMs: number;
2045
+ endMs: number;
2046
+ channels: {
2047
+ index: number;
2048
+ rmsDbfs: number | null;
2049
+ samplePeakDbfs: number | null;
2050
+ potentialClippedSamples: number;
2051
+ }[];
2052
+ }[];
2053
+ silence: {
2054
+ startMs: number;
2055
+ endMs: number;
2056
+ }[];
2057
+ integratedLufs: number | null;
2058
+ loudnessTimeline: {
2059
+ startMs: number;
2060
+ endMs: number;
2061
+ momentaryLufs: number | null;
2062
+ shortTermLufs: number | null;
2063
+ }[];
2064
+ previewUrl?: string | undefined;
2065
+ }, {
2066
+ status: "no_audio" | "measured";
2067
+ version: 1;
2068
+ operation: "measure_audio_levels";
2069
+ source: {
2070
+ durationMs: number;
2071
+ sha256: string;
2072
+ bytes: number;
2073
+ hasAudio: boolean;
2074
+ originStartMs: number;
2075
+ audioStartPts: string | null;
2076
+ audioTimeBase: string | null;
2077
+ sampleRate: number | null;
2078
+ channels: number | null;
2079
+ channelLayout: string | null;
2080
+ audioOffsetMs: number | null;
2081
+ };
2082
+ manifestUrl: string;
2083
+ settings: {
2084
+ channelPolicy: "native-separate";
2085
+ windowMs: number;
2086
+ silenceThresholdDbfs: number;
2087
+ silenceMethod: "all-channel-window-rms";
2088
+ potentialClippingThreshold: number;
2089
+ loudnessMethod: "FFmpeg-ebur128-native-channels";
2090
+ ffmpegVersion: string;
2091
+ };
2092
+ warnings: string[];
2093
+ coverage: {
2094
+ requestedRange: {
2095
+ startMs: number;
2096
+ endMs: number;
2097
+ };
2098
+ analyzed: {
2099
+ startMs: number;
2100
+ endMs: number;
2101
+ }[];
2102
+ withoutAudio: {
2103
+ startMs: number;
2104
+ endMs: number;
2105
+ }[];
2106
+ sampleRoundingToleranceMs: number;
2107
+ };
2108
+ channels: {
2109
+ index: number;
2110
+ rmsDbfs: number | null;
2111
+ samplePeakDbfs: number | null;
2112
+ potentialClippedSamples: number;
2113
+ truePeakDbtp: number | null;
2114
+ sampleCount: number;
2115
+ intersampleOvershoot: boolean;
2116
+ }[];
2117
+ truePeakDbtp: number | null;
2118
+ windows: {
2119
+ startMs: number;
2120
+ endMs: number;
2121
+ channels: {
2122
+ index: number;
2123
+ rmsDbfs: number | null;
2124
+ samplePeakDbfs: number | null;
2125
+ potentialClippedSamples: number;
2126
+ }[];
2127
+ }[];
2128
+ silence: {
2129
+ startMs: number;
2130
+ endMs: number;
2131
+ }[];
2132
+ integratedLufs: number | null;
2133
+ loudnessTimeline: {
2134
+ startMs: number;
2135
+ endMs: number;
2136
+ momentaryLufs: number | null;
2137
+ shortTermLufs: number | null;
2138
+ }[];
2139
+ previewUrl?: string | undefined;
2140
+ }>;
2141
+ export declare const audioTranscriptResultSchema: z.ZodObject<{
2142
+ version: z.ZodLiteral<1>;
2143
+ operation: z.ZodLiteral<"transcribe_audio">;
2144
+ source: z.ZodObject<{
2145
+ sha256: z.ZodString;
2146
+ bytes: z.ZodNumber;
2147
+ hasAudio: z.ZodBoolean;
2148
+ durationMs: z.ZodNumber;
2149
+ originStartMs: z.ZodNumber;
2150
+ audioStartPts: z.ZodNullable<z.ZodString>;
2151
+ audioTimeBase: z.ZodNullable<z.ZodString>;
2152
+ sampleRate: z.ZodNullable<z.ZodNumber>;
2153
+ channels: z.ZodNullable<z.ZodNumber>;
2154
+ channelLayout: z.ZodNullable<z.ZodString>;
2155
+ audioOffsetMs: z.ZodNullable<z.ZodNumber>;
2156
+ }, "strip", z.ZodTypeAny, {
2157
+ durationMs: number;
2158
+ sha256: string;
2159
+ bytes: number;
2160
+ hasAudio: boolean;
2161
+ originStartMs: number;
2162
+ audioStartPts: string | null;
2163
+ audioTimeBase: string | null;
2164
+ sampleRate: number | null;
2165
+ channels: number | null;
2166
+ channelLayout: string | null;
2167
+ audioOffsetMs: number | null;
2168
+ }, {
2169
+ durationMs: number;
2170
+ sha256: string;
2171
+ bytes: number;
2172
+ hasAudio: boolean;
2173
+ originStartMs: number;
2174
+ audioStartPts: string | null;
2175
+ audioTimeBase: string | null;
2176
+ sampleRate: number | null;
2177
+ channels: number | null;
2178
+ channelLayout: string | null;
2179
+ audioOffsetMs: number | null;
2180
+ }>;
2181
+ status: z.ZodEnum<["no_audio", "no_speech_detected", "transcribed"]>;
2182
+ text: z.ZodString;
2183
+ coverage: z.ZodObject<{
2184
+ requestedRange: z.ZodObject<{
2185
+ startMs: z.ZodNumber;
2186
+ endMs: z.ZodNumber;
2187
+ }, "strip", z.ZodTypeAny, {
2188
+ startMs: number;
2189
+ endMs: number;
2190
+ }, {
2191
+ startMs: number;
2192
+ endMs: number;
2193
+ }>;
2194
+ analyzed: z.ZodArray<z.ZodObject<{
2195
+ startMs: z.ZodNumber;
2196
+ endMs: z.ZodNumber;
2197
+ }, "strip", z.ZodTypeAny, {
2198
+ startMs: number;
2199
+ endMs: number;
2200
+ }, {
2201
+ startMs: number;
2202
+ endMs: number;
2203
+ }>, "many">;
2204
+ withoutAudio: z.ZodArray<z.ZodObject<{
2205
+ startMs: z.ZodNumber;
2206
+ endMs: z.ZodNumber;
2207
+ }, "strip", z.ZodTypeAny, {
2208
+ startMs: number;
2209
+ endMs: number;
2210
+ }, {
2211
+ startMs: number;
2212
+ endMs: number;
2213
+ }>, "many">;
2214
+ sampleRoundingToleranceMs: z.ZodNumber;
2215
+ }, "strip", z.ZodTypeAny, {
2216
+ requestedRange: {
2217
+ startMs: number;
2218
+ endMs: number;
2219
+ };
2220
+ analyzed: {
2221
+ startMs: number;
2222
+ endMs: number;
2223
+ }[];
2224
+ withoutAudio: {
2225
+ startMs: number;
2226
+ endMs: number;
2227
+ }[];
2228
+ sampleRoundingToleranceMs: number;
2229
+ }, {
2230
+ requestedRange: {
2231
+ startMs: number;
2232
+ endMs: number;
2233
+ };
2234
+ analyzed: {
2235
+ startMs: number;
2236
+ endMs: number;
2237
+ }[];
2238
+ withoutAudio: {
2239
+ startMs: number;
2240
+ endMs: number;
2241
+ }[];
2242
+ sampleRoundingToleranceMs: number;
2243
+ }>;
2244
+ model: z.ZodObject<{
2245
+ name: z.ZodString;
2246
+ repository: z.ZodString;
2247
+ revision: z.ZodString;
2248
+ sha256: z.ZodRecord<z.ZodString, z.ZodString>;
2249
+ runtime: z.ZodString;
2250
+ device: z.ZodLiteral<"cpu">;
2251
+ computeType: z.ZodLiteral<"int8">;
2252
+ vad: z.ZodString;
2253
+ beamSize: z.ZodNumber;
2254
+ temperature: z.ZodNumber;
2255
+ conditionOnPreviousText: z.ZodBoolean;
2256
+ }, "strip", z.ZodTypeAny, {
2257
+ sha256: Record<string, string>;
2258
+ name: string;
2259
+ runtime: string;
2260
+ device: "cpu";
2261
+ repository: string;
2262
+ revision: string;
2263
+ computeType: "int8";
2264
+ vad: string;
2265
+ beamSize: number;
2266
+ temperature: number;
2267
+ conditionOnPreviousText: boolean;
2268
+ }, {
2269
+ sha256: Record<string, string>;
2270
+ name: string;
2271
+ runtime: string;
2272
+ device: "cpu";
2273
+ repository: string;
2274
+ revision: string;
2275
+ computeType: "int8";
2276
+ vad: string;
2277
+ beamSize: number;
2278
+ temperature: number;
2279
+ conditionOnPreviousText: boolean;
2280
+ }>;
2281
+ segments: z.ZodArray<z.ZodObject<{
2282
+ index: z.ZodNumber;
2283
+ text: z.ZodString;
2284
+ startMs: z.ZodNumber;
2285
+ endMs: z.ZodNumber;
2286
+ timing: z.ZodLiteral<"estimated">;
2287
+ avgLogProbability: z.ZodNumber;
2288
+ noSpeechProbability: z.ZodNumber;
2289
+ words: z.ZodArray<z.ZodObject<{
2290
+ text: z.ZodString;
2291
+ startMs: z.ZodNumber;
2292
+ endMs: z.ZodNumber;
2293
+ timing: z.ZodLiteral<"estimated">;
2294
+ probability: z.ZodNumber;
2295
+ }, "strip", z.ZodTypeAny, {
2296
+ startMs: number;
2297
+ endMs: number;
2298
+ text: string;
2299
+ timing: "estimated";
2300
+ probability: number;
2301
+ }, {
2302
+ startMs: number;
2303
+ endMs: number;
2304
+ text: string;
2305
+ timing: "estimated";
2306
+ probability: number;
2307
+ }>, "many">;
2308
+ }, "strip", z.ZodTypeAny, {
2309
+ startMs: number;
2310
+ endMs: number;
2311
+ index: number;
2312
+ text: string;
2313
+ timing: "estimated";
2314
+ avgLogProbability: number;
2315
+ noSpeechProbability: number;
2316
+ words: {
2317
+ startMs: number;
2318
+ endMs: number;
2319
+ text: string;
2320
+ timing: "estimated";
2321
+ probability: number;
2322
+ }[];
2323
+ }, {
2324
+ startMs: number;
2325
+ endMs: number;
2326
+ index: number;
2327
+ text: string;
2328
+ timing: "estimated";
2329
+ avgLogProbability: number;
2330
+ noSpeechProbability: number;
2331
+ words: {
2332
+ startMs: number;
2333
+ endMs: number;
2334
+ text: string;
2335
+ timing: "estimated";
2336
+ probability: number;
2337
+ }[];
2338
+ }>, "many">;
2339
+ language: z.ZodNullable<z.ZodString>;
2340
+ languageProbability: z.ZodNullable<z.ZodNumber>;
2341
+ segmentTiming: z.ZodEnum<["estimated", "unavailable"]>;
2342
+ wordTiming: z.ZodEnum<["estimated", "unavailable"]>;
2343
+ speakerLabels: z.ZodLiteral<"unavailable">;
2344
+ channelPolicy: z.ZodString;
2345
+ warnings: z.ZodArray<z.ZodString, "many">;
2346
+ audioPreviewUrl: z.ZodOptional<z.ZodString>;
2347
+ manifestUrl: z.ZodString;
2348
+ }, "strip", z.ZodTypeAny, {
2349
+ status: "no_audio" | "no_speech_detected" | "transcribed";
2350
+ version: 1;
2351
+ operation: "transcribe_audio";
2352
+ source: {
2353
+ durationMs: number;
2354
+ sha256: string;
2355
+ bytes: number;
2356
+ hasAudio: boolean;
2357
+ originStartMs: number;
2358
+ audioStartPts: string | null;
2359
+ audioTimeBase: string | null;
2360
+ sampleRate: number | null;
2361
+ channels: number | null;
2362
+ channelLayout: string | null;
2363
+ audioOffsetMs: number | null;
2364
+ };
2365
+ manifestUrl: string;
2366
+ warnings: string[];
2367
+ text: string;
2368
+ model: {
2369
+ sha256: Record<string, string>;
2370
+ name: string;
2371
+ runtime: string;
2372
+ device: "cpu";
2373
+ repository: string;
2374
+ revision: string;
2375
+ computeType: "int8";
2376
+ vad: string;
2377
+ beamSize: number;
2378
+ temperature: number;
2379
+ conditionOnPreviousText: boolean;
2380
+ };
2381
+ coverage: {
2382
+ requestedRange: {
2383
+ startMs: number;
2384
+ endMs: number;
2385
+ };
2386
+ analyzed: {
2387
+ startMs: number;
2388
+ endMs: number;
2389
+ }[];
2390
+ withoutAudio: {
2391
+ startMs: number;
2392
+ endMs: number;
2393
+ }[];
2394
+ sampleRoundingToleranceMs: number;
2395
+ };
2396
+ segments: {
2397
+ startMs: number;
2398
+ endMs: number;
2399
+ index: number;
2400
+ text: string;
2401
+ timing: "estimated";
2402
+ avgLogProbability: number;
2403
+ noSpeechProbability: number;
2404
+ words: {
2405
+ startMs: number;
2406
+ endMs: number;
2407
+ text: string;
2408
+ timing: "estimated";
2409
+ probability: number;
2410
+ }[];
2411
+ }[];
2412
+ language: string | null;
2413
+ languageProbability: number | null;
2414
+ segmentTiming: "estimated" | "unavailable";
2415
+ wordTiming: "estimated" | "unavailable";
2416
+ speakerLabels: "unavailable";
2417
+ channelPolicy: string;
2418
+ audioPreviewUrl?: string | undefined;
2419
+ }, {
2420
+ status: "no_audio" | "no_speech_detected" | "transcribed";
2421
+ version: 1;
2422
+ operation: "transcribe_audio";
2423
+ source: {
2424
+ durationMs: number;
2425
+ sha256: string;
2426
+ bytes: number;
2427
+ hasAudio: boolean;
2428
+ originStartMs: number;
2429
+ audioStartPts: string | null;
2430
+ audioTimeBase: string | null;
2431
+ sampleRate: number | null;
2432
+ channels: number | null;
2433
+ channelLayout: string | null;
2434
+ audioOffsetMs: number | null;
2435
+ };
2436
+ manifestUrl: string;
2437
+ warnings: string[];
2438
+ text: string;
2439
+ model: {
2440
+ sha256: Record<string, string>;
2441
+ name: string;
2442
+ runtime: string;
2443
+ device: "cpu";
2444
+ repository: string;
2445
+ revision: string;
2446
+ computeType: "int8";
2447
+ vad: string;
2448
+ beamSize: number;
2449
+ temperature: number;
2450
+ conditionOnPreviousText: boolean;
2451
+ };
2452
+ coverage: {
2453
+ requestedRange: {
2454
+ startMs: number;
2455
+ endMs: number;
2456
+ };
2457
+ analyzed: {
2458
+ startMs: number;
2459
+ endMs: number;
2460
+ }[];
2461
+ withoutAudio: {
2462
+ startMs: number;
2463
+ endMs: number;
2464
+ }[];
2465
+ sampleRoundingToleranceMs: number;
2466
+ };
2467
+ segments: {
2468
+ startMs: number;
2469
+ endMs: number;
2470
+ index: number;
2471
+ text: string;
2472
+ timing: "estimated";
2473
+ avgLogProbability: number;
2474
+ noSpeechProbability: number;
2475
+ words: {
2476
+ startMs: number;
2477
+ endMs: number;
2478
+ text: string;
2479
+ timing: "estimated";
2480
+ probability: number;
2481
+ }[];
2482
+ }[];
2483
+ language: string | null;
2484
+ languageProbability: number | null;
2485
+ segmentTiming: "estimated" | "unavailable";
2486
+ wordTiming: "estimated" | "unavailable";
2487
+ speakerLabels: "unavailable";
2488
+ channelPolicy: string;
2489
+ audioPreviewUrl?: string | undefined;
2490
+ }>;
2491
+ export type AudioTranscriptResult = z.infer<typeof audioTranscriptResultSchema>;
2492
+ export type AudioLevelsResult = z.infer<typeof audioLevelsResultSchema>;
1687
2493
  export declare const mediaAnalysisResultSchema: z.ZodDiscriminatedUnion<"operation", [z.ZodObject<{
1688
2494
  version: z.ZodLiteral<1>;
1689
2495
  operation: z.ZodLiteral<"extract_frames">;
@@ -2733,10 +3539,287 @@ export declare const mediaAnalysisResultSchema: z.ZodDiscriminatedUnion<"operati
2733
3539
  y: number;
2734
3540
  };
2735
3541
  confidence: number;
2736
- sampleIndex: number;
2737
- lineId: string;
3542
+ sampleIndex: number;
3543
+ lineId: string;
3544
+ }[];
3545
+ }, {
3546
+ id: string;
3547
+ text: string | null;
3548
+ rawText: string;
3549
+ firstObservedMs: number;
3550
+ lastObservedEndMs: number;
3551
+ observations: {
3552
+ endMs: number;
3553
+ timeMs: number;
3554
+ sourcePts: string;
3555
+ frameIndex: number;
3556
+ box: {
3557
+ width: number;
3558
+ height: number;
3559
+ x: number;
3560
+ y: number;
3561
+ };
3562
+ confidence: number;
3563
+ sampleIndex: number;
3564
+ lineId: string;
3565
+ }[];
3566
+ }>, "many">;
3567
+ coverage: z.ZodObject<{
3568
+ mode: z.ZodEnum<["fast", "frame_scan"]>;
3569
+ range: z.ZodObject<{
3570
+ startMs: z.ZodNumber;
3571
+ endMs: z.ZodNumber;
3572
+ }, "strip", z.ZodTypeAny, {
3573
+ startMs: number;
3574
+ endMs: number;
3575
+ }, {
3576
+ startMs: number;
3577
+ endMs: number;
3578
+ }>;
3579
+ complete: z.ZodBoolean;
3580
+ uninspected: z.ZodArray<z.ZodObject<{
3581
+ startMs: z.ZodNumber;
3582
+ endMs: z.ZodNumber;
3583
+ }, "strip", z.ZodTypeAny, {
3584
+ startMs: number;
3585
+ endMs: number;
3586
+ }, {
3587
+ startMs: number;
3588
+ endMs: number;
3589
+ }>, "many">;
3590
+ inspected: z.ZodArray<z.ZodObject<{
3591
+ startMs: z.ZodNumber;
3592
+ endMs: z.ZodNumber;
3593
+ } & {
3594
+ frameIndex: z.ZodNumber;
3595
+ sourcePts: z.ZodString;
3596
+ }, "strip", z.ZodTypeAny, {
3597
+ startMs: number;
3598
+ endMs: number;
3599
+ sourcePts: string;
3600
+ frameIndex: number;
3601
+ }, {
3602
+ startMs: number;
3603
+ endMs: number;
3604
+ sourcePts: string;
3605
+ frameIndex: number;
3606
+ }>, "many">;
3607
+ }, "strip", z.ZodTypeAny, {
3608
+ complete: boolean;
3609
+ range: {
3610
+ startMs: number;
3611
+ endMs: number;
3612
+ };
3613
+ mode: "fast" | "frame_scan";
3614
+ uninspected: {
3615
+ startMs: number;
3616
+ endMs: number;
3617
+ }[];
3618
+ inspected: {
3619
+ startMs: number;
3620
+ endMs: number;
3621
+ sourcePts: string;
3622
+ frameIndex: number;
3623
+ }[];
3624
+ }, {
3625
+ complete: boolean;
3626
+ range: {
3627
+ startMs: number;
3628
+ endMs: number;
3629
+ };
3630
+ mode: "fast" | "frame_scan";
3631
+ uninspected: {
3632
+ startMs: number;
3633
+ endMs: number;
3634
+ }[];
3635
+ inspected: {
3636
+ startMs: number;
3637
+ endMs: number;
3638
+ sourcePts: string;
3639
+ frameIndex: number;
3640
+ }[];
3641
+ }>;
3642
+ }, "strip", z.ZodTypeAny, {
3643
+ version: 1;
3644
+ operation: "read_video_text";
3645
+ source: {
3646
+ encodedWidth: number;
3647
+ encodedHeight: number;
3648
+ displayWidth: number;
3649
+ displayHeight: number;
3650
+ rotationDegrees: number;
3651
+ sampleAspectRatio: string;
3652
+ durationMs: number;
3653
+ timeBase: string;
3654
+ startPts: string;
3655
+ startTimeMs: number;
3656
+ frameCount: number;
3657
+ variableFrameRate: boolean;
3658
+ sha256: string;
3659
+ bytes: number;
3660
+ };
3661
+ frames: {
3662
+ endMs: number;
3663
+ timeMs: number;
3664
+ sourcePts: string;
3665
+ frameIndex: number;
3666
+ regions: {
3667
+ status: "readable" | "unreadable";
3668
+ box: {
3669
+ width: number;
3670
+ height: number;
3671
+ x: number;
3672
+ y: number;
3673
+ };
3674
+ id: string;
3675
+ polygon: {
3676
+ x: number;
3677
+ y: number;
3678
+ }[];
3679
+ text: string | null;
3680
+ rawText: string;
3681
+ confidence: number;
3682
+ }[];
3683
+ lines: {
3684
+ status: "readable" | "unreadable";
3685
+ box: {
3686
+ width: number;
3687
+ height: number;
3688
+ x: number;
3689
+ y: number;
3690
+ };
3691
+ id: string;
3692
+ text: string | null;
3693
+ rawText: string;
3694
+ confidence: number;
3695
+ readingOrder: number;
3696
+ regionIds: string[];
3697
+ }[];
3698
+ sceneChange: boolean;
3699
+ previewUrl?: string | undefined;
3700
+ }[];
3701
+ manifestUrl: string;
3702
+ warnings: string[];
3703
+ model: {
3704
+ name: string;
3705
+ runtime: string;
3706
+ device: "cpu";
3707
+ modelHashes: Record<string, string>;
3708
+ orientationClassifier: string;
3709
+ recognitionThreshold: number;
3710
+ coordinates: "display-oriented-source-pixels";
3711
+ };
3712
+ tracks: {
3713
+ id: string;
3714
+ text: string | null;
3715
+ rawText: string;
3716
+ firstObservedMs: number;
3717
+ lastObservedEndMs: number;
3718
+ observations: {
3719
+ endMs: number;
3720
+ timeMs: number;
3721
+ sourcePts: string;
3722
+ frameIndex: number;
3723
+ box: {
3724
+ width: number;
3725
+ height: number;
3726
+ x: number;
3727
+ y: number;
3728
+ };
3729
+ confidence: number;
3730
+ sampleIndex: number;
3731
+ lineId: string;
3732
+ }[];
3733
+ }[];
3734
+ coverage: {
3735
+ complete: boolean;
3736
+ range: {
3737
+ startMs: number;
3738
+ endMs: number;
3739
+ };
3740
+ mode: "fast" | "frame_scan";
3741
+ uninspected: {
3742
+ startMs: number;
3743
+ endMs: number;
3744
+ }[];
3745
+ inspected: {
3746
+ startMs: number;
3747
+ endMs: number;
3748
+ sourcePts: string;
3749
+ frameIndex: number;
3750
+ }[];
3751
+ };
3752
+ }, {
3753
+ version: 1;
3754
+ operation: "read_video_text";
3755
+ source: {
3756
+ encodedWidth: number;
3757
+ encodedHeight: number;
3758
+ displayWidth: number;
3759
+ displayHeight: number;
3760
+ rotationDegrees: number;
3761
+ sampleAspectRatio: string;
3762
+ durationMs: number;
3763
+ timeBase: string;
3764
+ startPts: string;
3765
+ startTimeMs: number;
3766
+ frameCount: number;
3767
+ variableFrameRate: boolean;
3768
+ sha256: string;
3769
+ bytes: number;
3770
+ };
3771
+ frames: {
3772
+ endMs: number;
3773
+ timeMs: number;
3774
+ sourcePts: string;
3775
+ frameIndex: number;
3776
+ regions: {
3777
+ status: "readable" | "unreadable";
3778
+ box: {
3779
+ width: number;
3780
+ height: number;
3781
+ x: number;
3782
+ y: number;
3783
+ };
3784
+ id: string;
3785
+ polygon: {
3786
+ x: number;
3787
+ y: number;
3788
+ }[];
3789
+ text: string | null;
3790
+ rawText: string;
3791
+ confidence: number;
3792
+ }[];
3793
+ lines: {
3794
+ status: "readable" | "unreadable";
3795
+ box: {
3796
+ width: number;
3797
+ height: number;
3798
+ x: number;
3799
+ y: number;
3800
+ };
3801
+ id: string;
3802
+ text: string | null;
3803
+ rawText: string;
3804
+ confidence: number;
3805
+ readingOrder: number;
3806
+ regionIds: string[];
2738
3807
  }[];
2739
- }, {
3808
+ sceneChange: boolean;
3809
+ previewUrl?: string | undefined;
3810
+ }[];
3811
+ manifestUrl: string;
3812
+ warnings: string[];
3813
+ model: {
3814
+ name: string;
3815
+ runtime: string;
3816
+ device: "cpu";
3817
+ modelHashes: Record<string, string>;
3818
+ orientationClassifier: string;
3819
+ recognitionThreshold: number;
3820
+ coordinates: "display-oriented-source-pixels";
3821
+ };
3822
+ tracks: {
2740
3823
  id: string;
2741
3824
  text: string | null;
2742
3825
  rawText: string;
@@ -2757,10 +3840,417 @@ export declare const mediaAnalysisResultSchema: z.ZodDiscriminatedUnion<"operati
2757
3840
  sampleIndex: number;
2758
3841
  lineId: string;
2759
3842
  }[];
3843
+ }[];
3844
+ coverage: {
3845
+ complete: boolean;
3846
+ range: {
3847
+ startMs: number;
3848
+ endMs: number;
3849
+ };
3850
+ mode: "fast" | "frame_scan";
3851
+ uninspected: {
3852
+ startMs: number;
3853
+ endMs: number;
3854
+ }[];
3855
+ inspected: {
3856
+ startMs: number;
3857
+ endMs: number;
3858
+ sourcePts: string;
3859
+ frameIndex: number;
3860
+ }[];
3861
+ };
3862
+ }>, z.ZodObject<{
3863
+ version: z.ZodLiteral<1>;
3864
+ operation: z.ZodLiteral<"transcribe_audio">;
3865
+ source: z.ZodObject<{
3866
+ sha256: z.ZodString;
3867
+ bytes: z.ZodNumber;
3868
+ hasAudio: z.ZodBoolean;
3869
+ durationMs: z.ZodNumber;
3870
+ originStartMs: z.ZodNumber;
3871
+ audioStartPts: z.ZodNullable<z.ZodString>;
3872
+ audioTimeBase: z.ZodNullable<z.ZodString>;
3873
+ sampleRate: z.ZodNullable<z.ZodNumber>;
3874
+ channels: z.ZodNullable<z.ZodNumber>;
3875
+ channelLayout: z.ZodNullable<z.ZodString>;
3876
+ audioOffsetMs: z.ZodNullable<z.ZodNumber>;
3877
+ }, "strip", z.ZodTypeAny, {
3878
+ durationMs: number;
3879
+ sha256: string;
3880
+ bytes: number;
3881
+ hasAudio: boolean;
3882
+ originStartMs: number;
3883
+ audioStartPts: string | null;
3884
+ audioTimeBase: string | null;
3885
+ sampleRate: number | null;
3886
+ channels: number | null;
3887
+ channelLayout: string | null;
3888
+ audioOffsetMs: number | null;
3889
+ }, {
3890
+ durationMs: number;
3891
+ sha256: string;
3892
+ bytes: number;
3893
+ hasAudio: boolean;
3894
+ originStartMs: number;
3895
+ audioStartPts: string | null;
3896
+ audioTimeBase: string | null;
3897
+ sampleRate: number | null;
3898
+ channels: number | null;
3899
+ channelLayout: string | null;
3900
+ audioOffsetMs: number | null;
3901
+ }>;
3902
+ status: z.ZodEnum<["no_audio", "no_speech_detected", "transcribed"]>;
3903
+ text: z.ZodString;
3904
+ coverage: z.ZodObject<{
3905
+ requestedRange: z.ZodObject<{
3906
+ startMs: z.ZodNumber;
3907
+ endMs: z.ZodNumber;
3908
+ }, "strip", z.ZodTypeAny, {
3909
+ startMs: number;
3910
+ endMs: number;
3911
+ }, {
3912
+ startMs: number;
3913
+ endMs: number;
3914
+ }>;
3915
+ analyzed: z.ZodArray<z.ZodObject<{
3916
+ startMs: z.ZodNumber;
3917
+ endMs: z.ZodNumber;
3918
+ }, "strip", z.ZodTypeAny, {
3919
+ startMs: number;
3920
+ endMs: number;
3921
+ }, {
3922
+ startMs: number;
3923
+ endMs: number;
3924
+ }>, "many">;
3925
+ withoutAudio: z.ZodArray<z.ZodObject<{
3926
+ startMs: z.ZodNumber;
3927
+ endMs: z.ZodNumber;
3928
+ }, "strip", z.ZodTypeAny, {
3929
+ startMs: number;
3930
+ endMs: number;
3931
+ }, {
3932
+ startMs: number;
3933
+ endMs: number;
3934
+ }>, "many">;
3935
+ sampleRoundingToleranceMs: z.ZodNumber;
3936
+ }, "strip", z.ZodTypeAny, {
3937
+ requestedRange: {
3938
+ startMs: number;
3939
+ endMs: number;
3940
+ };
3941
+ analyzed: {
3942
+ startMs: number;
3943
+ endMs: number;
3944
+ }[];
3945
+ withoutAudio: {
3946
+ startMs: number;
3947
+ endMs: number;
3948
+ }[];
3949
+ sampleRoundingToleranceMs: number;
3950
+ }, {
3951
+ requestedRange: {
3952
+ startMs: number;
3953
+ endMs: number;
3954
+ };
3955
+ analyzed: {
3956
+ startMs: number;
3957
+ endMs: number;
3958
+ }[];
3959
+ withoutAudio: {
3960
+ startMs: number;
3961
+ endMs: number;
3962
+ }[];
3963
+ sampleRoundingToleranceMs: number;
3964
+ }>;
3965
+ model: z.ZodObject<{
3966
+ name: z.ZodString;
3967
+ repository: z.ZodString;
3968
+ revision: z.ZodString;
3969
+ sha256: z.ZodRecord<z.ZodString, z.ZodString>;
3970
+ runtime: z.ZodString;
3971
+ device: z.ZodLiteral<"cpu">;
3972
+ computeType: z.ZodLiteral<"int8">;
3973
+ vad: z.ZodString;
3974
+ beamSize: z.ZodNumber;
3975
+ temperature: z.ZodNumber;
3976
+ conditionOnPreviousText: z.ZodBoolean;
3977
+ }, "strip", z.ZodTypeAny, {
3978
+ sha256: Record<string, string>;
3979
+ name: string;
3980
+ runtime: string;
3981
+ device: "cpu";
3982
+ repository: string;
3983
+ revision: string;
3984
+ computeType: "int8";
3985
+ vad: string;
3986
+ beamSize: number;
3987
+ temperature: number;
3988
+ conditionOnPreviousText: boolean;
3989
+ }, {
3990
+ sha256: Record<string, string>;
3991
+ name: string;
3992
+ runtime: string;
3993
+ device: "cpu";
3994
+ repository: string;
3995
+ revision: string;
3996
+ computeType: "int8";
3997
+ vad: string;
3998
+ beamSize: number;
3999
+ temperature: number;
4000
+ conditionOnPreviousText: boolean;
4001
+ }>;
4002
+ segments: z.ZodArray<z.ZodObject<{
4003
+ index: z.ZodNumber;
4004
+ text: z.ZodString;
4005
+ startMs: z.ZodNumber;
4006
+ endMs: z.ZodNumber;
4007
+ timing: z.ZodLiteral<"estimated">;
4008
+ avgLogProbability: z.ZodNumber;
4009
+ noSpeechProbability: z.ZodNumber;
4010
+ words: z.ZodArray<z.ZodObject<{
4011
+ text: z.ZodString;
4012
+ startMs: z.ZodNumber;
4013
+ endMs: z.ZodNumber;
4014
+ timing: z.ZodLiteral<"estimated">;
4015
+ probability: z.ZodNumber;
4016
+ }, "strip", z.ZodTypeAny, {
4017
+ startMs: number;
4018
+ endMs: number;
4019
+ text: string;
4020
+ timing: "estimated";
4021
+ probability: number;
4022
+ }, {
4023
+ startMs: number;
4024
+ endMs: number;
4025
+ text: string;
4026
+ timing: "estimated";
4027
+ probability: number;
4028
+ }>, "many">;
4029
+ }, "strip", z.ZodTypeAny, {
4030
+ startMs: number;
4031
+ endMs: number;
4032
+ index: number;
4033
+ text: string;
4034
+ timing: "estimated";
4035
+ avgLogProbability: number;
4036
+ noSpeechProbability: number;
4037
+ words: {
4038
+ startMs: number;
4039
+ endMs: number;
4040
+ text: string;
4041
+ timing: "estimated";
4042
+ probability: number;
4043
+ }[];
4044
+ }, {
4045
+ startMs: number;
4046
+ endMs: number;
4047
+ index: number;
4048
+ text: string;
4049
+ timing: "estimated";
4050
+ avgLogProbability: number;
4051
+ noSpeechProbability: number;
4052
+ words: {
4053
+ startMs: number;
4054
+ endMs: number;
4055
+ text: string;
4056
+ timing: "estimated";
4057
+ probability: number;
4058
+ }[];
2760
4059
  }>, "many">;
4060
+ language: z.ZodNullable<z.ZodString>;
4061
+ languageProbability: z.ZodNullable<z.ZodNumber>;
4062
+ segmentTiming: z.ZodEnum<["estimated", "unavailable"]>;
4063
+ wordTiming: z.ZodEnum<["estimated", "unavailable"]>;
4064
+ speakerLabels: z.ZodLiteral<"unavailable">;
4065
+ channelPolicy: z.ZodString;
4066
+ warnings: z.ZodArray<z.ZodString, "many">;
4067
+ audioPreviewUrl: z.ZodOptional<z.ZodString>;
4068
+ manifestUrl: z.ZodString;
4069
+ }, "strip", z.ZodTypeAny, {
4070
+ status: "no_audio" | "no_speech_detected" | "transcribed";
4071
+ version: 1;
4072
+ operation: "transcribe_audio";
4073
+ source: {
4074
+ durationMs: number;
4075
+ sha256: string;
4076
+ bytes: number;
4077
+ hasAudio: boolean;
4078
+ originStartMs: number;
4079
+ audioStartPts: string | null;
4080
+ audioTimeBase: string | null;
4081
+ sampleRate: number | null;
4082
+ channels: number | null;
4083
+ channelLayout: string | null;
4084
+ audioOffsetMs: number | null;
4085
+ };
4086
+ manifestUrl: string;
4087
+ warnings: string[];
4088
+ text: string;
4089
+ model: {
4090
+ sha256: Record<string, string>;
4091
+ name: string;
4092
+ runtime: string;
4093
+ device: "cpu";
4094
+ repository: string;
4095
+ revision: string;
4096
+ computeType: "int8";
4097
+ vad: string;
4098
+ beamSize: number;
4099
+ temperature: number;
4100
+ conditionOnPreviousText: boolean;
4101
+ };
4102
+ coverage: {
4103
+ requestedRange: {
4104
+ startMs: number;
4105
+ endMs: number;
4106
+ };
4107
+ analyzed: {
4108
+ startMs: number;
4109
+ endMs: number;
4110
+ }[];
4111
+ withoutAudio: {
4112
+ startMs: number;
4113
+ endMs: number;
4114
+ }[];
4115
+ sampleRoundingToleranceMs: number;
4116
+ };
4117
+ segments: {
4118
+ startMs: number;
4119
+ endMs: number;
4120
+ index: number;
4121
+ text: string;
4122
+ timing: "estimated";
4123
+ avgLogProbability: number;
4124
+ noSpeechProbability: number;
4125
+ words: {
4126
+ startMs: number;
4127
+ endMs: number;
4128
+ text: string;
4129
+ timing: "estimated";
4130
+ probability: number;
4131
+ }[];
4132
+ }[];
4133
+ language: string | null;
4134
+ languageProbability: number | null;
4135
+ segmentTiming: "estimated" | "unavailable";
4136
+ wordTiming: "estimated" | "unavailable";
4137
+ speakerLabels: "unavailable";
4138
+ channelPolicy: string;
4139
+ audioPreviewUrl?: string | undefined;
4140
+ }, {
4141
+ status: "no_audio" | "no_speech_detected" | "transcribed";
4142
+ version: 1;
4143
+ operation: "transcribe_audio";
4144
+ source: {
4145
+ durationMs: number;
4146
+ sha256: string;
4147
+ bytes: number;
4148
+ hasAudio: boolean;
4149
+ originStartMs: number;
4150
+ audioStartPts: string | null;
4151
+ audioTimeBase: string | null;
4152
+ sampleRate: number | null;
4153
+ channels: number | null;
4154
+ channelLayout: string | null;
4155
+ audioOffsetMs: number | null;
4156
+ };
4157
+ manifestUrl: string;
4158
+ warnings: string[];
4159
+ text: string;
4160
+ model: {
4161
+ sha256: Record<string, string>;
4162
+ name: string;
4163
+ runtime: string;
4164
+ device: "cpu";
4165
+ repository: string;
4166
+ revision: string;
4167
+ computeType: "int8";
4168
+ vad: string;
4169
+ beamSize: number;
4170
+ temperature: number;
4171
+ conditionOnPreviousText: boolean;
4172
+ };
4173
+ coverage: {
4174
+ requestedRange: {
4175
+ startMs: number;
4176
+ endMs: number;
4177
+ };
4178
+ analyzed: {
4179
+ startMs: number;
4180
+ endMs: number;
4181
+ }[];
4182
+ withoutAudio: {
4183
+ startMs: number;
4184
+ endMs: number;
4185
+ }[];
4186
+ sampleRoundingToleranceMs: number;
4187
+ };
4188
+ segments: {
4189
+ startMs: number;
4190
+ endMs: number;
4191
+ index: number;
4192
+ text: string;
4193
+ timing: "estimated";
4194
+ avgLogProbability: number;
4195
+ noSpeechProbability: number;
4196
+ words: {
4197
+ startMs: number;
4198
+ endMs: number;
4199
+ text: string;
4200
+ timing: "estimated";
4201
+ probability: number;
4202
+ }[];
4203
+ }[];
4204
+ language: string | null;
4205
+ languageProbability: number | null;
4206
+ segmentTiming: "estimated" | "unavailable";
4207
+ wordTiming: "estimated" | "unavailable";
4208
+ speakerLabels: "unavailable";
4209
+ channelPolicy: string;
4210
+ audioPreviewUrl?: string | undefined;
4211
+ }>, z.ZodObject<{
4212
+ version: z.ZodLiteral<1>;
4213
+ operation: z.ZodLiteral<"measure_audio_levels">;
4214
+ source: z.ZodObject<{
4215
+ sha256: z.ZodString;
4216
+ bytes: z.ZodNumber;
4217
+ hasAudio: z.ZodBoolean;
4218
+ durationMs: z.ZodNumber;
4219
+ originStartMs: z.ZodNumber;
4220
+ audioStartPts: z.ZodNullable<z.ZodString>;
4221
+ audioTimeBase: z.ZodNullable<z.ZodString>;
4222
+ sampleRate: z.ZodNullable<z.ZodNumber>;
4223
+ channels: z.ZodNullable<z.ZodNumber>;
4224
+ channelLayout: z.ZodNullable<z.ZodString>;
4225
+ audioOffsetMs: z.ZodNullable<z.ZodNumber>;
4226
+ }, "strip", z.ZodTypeAny, {
4227
+ durationMs: number;
4228
+ sha256: string;
4229
+ bytes: number;
4230
+ hasAudio: boolean;
4231
+ originStartMs: number;
4232
+ audioStartPts: string | null;
4233
+ audioTimeBase: string | null;
4234
+ sampleRate: number | null;
4235
+ channels: number | null;
4236
+ channelLayout: string | null;
4237
+ audioOffsetMs: number | null;
4238
+ }, {
4239
+ durationMs: number;
4240
+ sha256: string;
4241
+ bytes: number;
4242
+ hasAudio: boolean;
4243
+ originStartMs: number;
4244
+ audioStartPts: string | null;
4245
+ audioTimeBase: string | null;
4246
+ sampleRate: number | null;
4247
+ channels: number | null;
4248
+ channelLayout: string | null;
4249
+ audioOffsetMs: number | null;
4250
+ }>;
4251
+ status: z.ZodEnum<["no_audio", "measured"]>;
2761
4252
  coverage: z.ZodObject<{
2762
- mode: z.ZodEnum<["fast", "frame_scan"]>;
2763
- range: z.ZodObject<{
4253
+ requestedRange: z.ZodObject<{
2764
4254
  startMs: z.ZodNumber;
2765
4255
  endMs: z.ZodNumber;
2766
4256
  }, "strip", z.ZodTypeAny, {
@@ -2770,8 +4260,7 @@ export declare const mediaAnalysisResultSchema: z.ZodDiscriminatedUnion<"operati
2770
4260
  startMs: number;
2771
4261
  endMs: number;
2772
4262
  }>;
2773
- complete: z.ZodBoolean;
2774
- uninspected: z.ZodArray<z.ZodObject<{
4263
+ analyzed: z.ZodArray<z.ZodObject<{
2775
4264
  startMs: z.ZodNumber;
2776
4265
  endMs: z.ZodNumber;
2777
4266
  }, "strip", z.ZodTypeAny, {
@@ -2781,276 +4270,316 @@ export declare const mediaAnalysisResultSchema: z.ZodDiscriminatedUnion<"operati
2781
4270
  startMs: number;
2782
4271
  endMs: number;
2783
4272
  }>, "many">;
2784
- inspected: z.ZodArray<z.ZodObject<{
4273
+ withoutAudio: z.ZodArray<z.ZodObject<{
2785
4274
  startMs: z.ZodNumber;
2786
4275
  endMs: z.ZodNumber;
2787
- } & {
2788
- frameIndex: z.ZodNumber;
2789
- sourcePts: z.ZodString;
2790
4276
  }, "strip", z.ZodTypeAny, {
2791
4277
  startMs: number;
2792
4278
  endMs: number;
2793
- sourcePts: string;
2794
- frameIndex: number;
2795
4279
  }, {
2796
4280
  startMs: number;
2797
4281
  endMs: number;
2798
- sourcePts: string;
2799
- frameIndex: number;
2800
4282
  }>, "many">;
4283
+ sampleRoundingToleranceMs: z.ZodNumber;
2801
4284
  }, "strip", z.ZodTypeAny, {
2802
- complete: boolean;
2803
- range: {
4285
+ requestedRange: {
2804
4286
  startMs: number;
2805
4287
  endMs: number;
2806
4288
  };
2807
- mode: "fast" | "frame_scan";
2808
- uninspected: {
4289
+ analyzed: {
2809
4290
  startMs: number;
2810
4291
  endMs: number;
2811
4292
  }[];
2812
- inspected: {
4293
+ withoutAudio: {
2813
4294
  startMs: number;
2814
4295
  endMs: number;
2815
- sourcePts: string;
2816
- frameIndex: number;
2817
4296
  }[];
4297
+ sampleRoundingToleranceMs: number;
2818
4298
  }, {
2819
- complete: boolean;
2820
- range: {
4299
+ requestedRange: {
2821
4300
  startMs: number;
2822
4301
  endMs: number;
2823
4302
  };
2824
- mode: "fast" | "frame_scan";
2825
- uninspected: {
4303
+ analyzed: {
2826
4304
  startMs: number;
2827
4305
  endMs: number;
2828
4306
  }[];
2829
- inspected: {
4307
+ withoutAudio: {
2830
4308
  startMs: number;
2831
4309
  endMs: number;
2832
- sourcePts: string;
2833
- frameIndex: number;
2834
4310
  }[];
4311
+ sampleRoundingToleranceMs: number;
4312
+ }>;
4313
+ channels: z.ZodArray<z.ZodObject<{
4314
+ index: z.ZodNumber;
4315
+ rmsDbfs: z.ZodNullable<z.ZodNumber>;
4316
+ samplePeakDbfs: z.ZodNullable<z.ZodNumber>;
4317
+ potentialClippedSamples: z.ZodNumber;
4318
+ } & {
4319
+ truePeakDbtp: z.ZodNullable<z.ZodNumber>;
4320
+ sampleCount: z.ZodNumber;
4321
+ intersampleOvershoot: z.ZodBoolean;
4322
+ }, "strip", z.ZodTypeAny, {
4323
+ index: number;
4324
+ rmsDbfs: number | null;
4325
+ samplePeakDbfs: number | null;
4326
+ potentialClippedSamples: number;
4327
+ truePeakDbtp: number | null;
4328
+ sampleCount: number;
4329
+ intersampleOvershoot: boolean;
4330
+ }, {
4331
+ index: number;
4332
+ rmsDbfs: number | null;
4333
+ samplePeakDbfs: number | null;
4334
+ potentialClippedSamples: number;
4335
+ truePeakDbtp: number | null;
4336
+ sampleCount: number;
4337
+ intersampleOvershoot: boolean;
4338
+ }>, "many">;
4339
+ windows: z.ZodArray<z.ZodObject<{
4340
+ startMs: z.ZodNumber;
4341
+ endMs: z.ZodNumber;
4342
+ } & {
4343
+ channels: z.ZodArray<z.ZodObject<{
4344
+ index: z.ZodNumber;
4345
+ rmsDbfs: z.ZodNullable<z.ZodNumber>;
4346
+ samplePeakDbfs: z.ZodNullable<z.ZodNumber>;
4347
+ potentialClippedSamples: z.ZodNumber;
4348
+ }, "strip", z.ZodTypeAny, {
4349
+ index: number;
4350
+ rmsDbfs: number | null;
4351
+ samplePeakDbfs: number | null;
4352
+ potentialClippedSamples: number;
4353
+ }, {
4354
+ index: number;
4355
+ rmsDbfs: number | null;
4356
+ samplePeakDbfs: number | null;
4357
+ potentialClippedSamples: number;
4358
+ }>, "many">;
4359
+ }, "strip", z.ZodTypeAny, {
4360
+ startMs: number;
4361
+ endMs: number;
4362
+ channels: {
4363
+ index: number;
4364
+ rmsDbfs: number | null;
4365
+ samplePeakDbfs: number | null;
4366
+ potentialClippedSamples: number;
4367
+ }[];
4368
+ }, {
4369
+ startMs: number;
4370
+ endMs: number;
4371
+ channels: {
4372
+ index: number;
4373
+ rmsDbfs: number | null;
4374
+ samplePeakDbfs: number | null;
4375
+ potentialClippedSamples: number;
4376
+ }[];
4377
+ }>, "many">;
4378
+ silence: z.ZodArray<z.ZodObject<{
4379
+ startMs: z.ZodNumber;
4380
+ endMs: z.ZodNumber;
4381
+ }, "strip", z.ZodTypeAny, {
4382
+ startMs: number;
4383
+ endMs: number;
4384
+ }, {
4385
+ startMs: number;
4386
+ endMs: number;
4387
+ }>, "many">;
4388
+ integratedLufs: z.ZodNullable<z.ZodNumber>;
4389
+ truePeakDbtp: z.ZodNullable<z.ZodNumber>;
4390
+ loudnessTimeline: z.ZodArray<z.ZodObject<{
4391
+ startMs: z.ZodNumber;
4392
+ endMs: z.ZodNumber;
4393
+ } & {
4394
+ momentaryLufs: z.ZodNullable<z.ZodNumber>;
4395
+ shortTermLufs: z.ZodNullable<z.ZodNumber>;
4396
+ }, "strip", z.ZodTypeAny, {
4397
+ startMs: number;
4398
+ endMs: number;
4399
+ momentaryLufs: number | null;
4400
+ shortTermLufs: number | null;
4401
+ }, {
4402
+ startMs: number;
4403
+ endMs: number;
4404
+ momentaryLufs: number | null;
4405
+ shortTermLufs: number | null;
4406
+ }>, "many">;
4407
+ settings: z.ZodObject<{
4408
+ windowMs: z.ZodNumber;
4409
+ silenceThresholdDbfs: z.ZodNumber;
4410
+ silenceMethod: z.ZodLiteral<"all-channel-window-rms">;
4411
+ potentialClippingThreshold: z.ZodNumber;
4412
+ channelPolicy: z.ZodLiteral<"native-separate">;
4413
+ loudnessMethod: z.ZodLiteral<"FFmpeg-ebur128-native-channels">;
4414
+ ffmpegVersion: z.ZodString;
4415
+ }, "strip", z.ZodTypeAny, {
4416
+ channelPolicy: "native-separate";
4417
+ windowMs: number;
4418
+ silenceThresholdDbfs: number;
4419
+ silenceMethod: "all-channel-window-rms";
4420
+ potentialClippingThreshold: number;
4421
+ loudnessMethod: "FFmpeg-ebur128-native-channels";
4422
+ ffmpegVersion: string;
4423
+ }, {
4424
+ channelPolicy: "native-separate";
4425
+ windowMs: number;
4426
+ silenceThresholdDbfs: number;
4427
+ silenceMethod: "all-channel-window-rms";
4428
+ potentialClippingThreshold: number;
4429
+ loudnessMethod: "FFmpeg-ebur128-native-channels";
4430
+ ffmpegVersion: string;
2835
4431
  }>;
4432
+ warnings: z.ZodArray<z.ZodString, "many">;
4433
+ previewUrl: z.ZodOptional<z.ZodString>;
4434
+ manifestUrl: z.ZodString;
2836
4435
  }, "strip", z.ZodTypeAny, {
4436
+ status: "no_audio" | "measured";
2837
4437
  version: 1;
2838
- operation: "read_video_text";
4438
+ operation: "measure_audio_levels";
2839
4439
  source: {
2840
- encodedWidth: number;
2841
- encodedHeight: number;
2842
- displayWidth: number;
2843
- displayHeight: number;
2844
- rotationDegrees: number;
2845
- sampleAspectRatio: string;
2846
4440
  durationMs: number;
2847
- timeBase: string;
2848
- startPts: string;
2849
- startTimeMs: number;
2850
- frameCount: number;
2851
- variableFrameRate: boolean;
2852
4441
  sha256: string;
2853
4442
  bytes: number;
4443
+ hasAudio: boolean;
4444
+ originStartMs: number;
4445
+ audioStartPts: string | null;
4446
+ audioTimeBase: string | null;
4447
+ sampleRate: number | null;
4448
+ channels: number | null;
4449
+ channelLayout: string | null;
4450
+ audioOffsetMs: number | null;
2854
4451
  };
2855
- frames: {
2856
- endMs: number;
2857
- timeMs: number;
2858
- sourcePts: string;
2859
- frameIndex: number;
2860
- regions: {
2861
- status: "readable" | "unreadable";
2862
- box: {
2863
- width: number;
2864
- height: number;
2865
- x: number;
2866
- y: number;
2867
- };
2868
- id: string;
2869
- polygon: {
2870
- x: number;
2871
- y: number;
2872
- }[];
2873
- text: string | null;
2874
- rawText: string;
2875
- confidence: number;
2876
- }[];
2877
- lines: {
2878
- status: "readable" | "unreadable";
2879
- box: {
2880
- width: number;
2881
- height: number;
2882
- x: number;
2883
- y: number;
2884
- };
2885
- id: string;
2886
- text: string | null;
2887
- rawText: string;
2888
- confidence: number;
2889
- readingOrder: number;
2890
- regionIds: string[];
2891
- }[];
2892
- sceneChange: boolean;
2893
- previewUrl?: string | undefined;
2894
- }[];
2895
4452
  manifestUrl: string;
2896
- warnings: string[];
2897
- model: {
2898
- name: string;
2899
- runtime: string;
2900
- device: "cpu";
2901
- modelHashes: Record<string, string>;
2902
- orientationClassifier: string;
2903
- recognitionThreshold: number;
2904
- coordinates: "display-oriented-source-pixels";
4453
+ settings: {
4454
+ channelPolicy: "native-separate";
4455
+ windowMs: number;
4456
+ silenceThresholdDbfs: number;
4457
+ silenceMethod: "all-channel-window-rms";
4458
+ potentialClippingThreshold: number;
4459
+ loudnessMethod: "FFmpeg-ebur128-native-channels";
4460
+ ffmpegVersion: string;
2905
4461
  };
2906
- tracks: {
2907
- id: string;
2908
- text: string | null;
2909
- rawText: string;
2910
- firstObservedMs: number;
2911
- lastObservedEndMs: number;
2912
- observations: {
2913
- endMs: number;
2914
- timeMs: number;
2915
- sourcePts: string;
2916
- frameIndex: number;
2917
- box: {
2918
- width: number;
2919
- height: number;
2920
- x: number;
2921
- y: number;
2922
- };
2923
- confidence: number;
2924
- sampleIndex: number;
2925
- lineId: string;
2926
- }[];
2927
- }[];
4462
+ warnings: string[];
2928
4463
  coverage: {
2929
- complete: boolean;
2930
- range: {
4464
+ requestedRange: {
2931
4465
  startMs: number;
2932
4466
  endMs: number;
2933
4467
  };
2934
- mode: "fast" | "frame_scan";
2935
- uninspected: {
4468
+ analyzed: {
2936
4469
  startMs: number;
2937
4470
  endMs: number;
2938
4471
  }[];
2939
- inspected: {
4472
+ withoutAudio: {
2940
4473
  startMs: number;
2941
4474
  endMs: number;
2942
- sourcePts: string;
2943
- frameIndex: number;
2944
4475
  }[];
4476
+ sampleRoundingToleranceMs: number;
2945
4477
  };
4478
+ channels: {
4479
+ index: number;
4480
+ rmsDbfs: number | null;
4481
+ samplePeakDbfs: number | null;
4482
+ potentialClippedSamples: number;
4483
+ truePeakDbtp: number | null;
4484
+ sampleCount: number;
4485
+ intersampleOvershoot: boolean;
4486
+ }[];
4487
+ truePeakDbtp: number | null;
4488
+ windows: {
4489
+ startMs: number;
4490
+ endMs: number;
4491
+ channels: {
4492
+ index: number;
4493
+ rmsDbfs: number | null;
4494
+ samplePeakDbfs: number | null;
4495
+ potentialClippedSamples: number;
4496
+ }[];
4497
+ }[];
4498
+ silence: {
4499
+ startMs: number;
4500
+ endMs: number;
4501
+ }[];
4502
+ integratedLufs: number | null;
4503
+ loudnessTimeline: {
4504
+ startMs: number;
4505
+ endMs: number;
4506
+ momentaryLufs: number | null;
4507
+ shortTermLufs: number | null;
4508
+ }[];
4509
+ previewUrl?: string | undefined;
2946
4510
  }, {
4511
+ status: "no_audio" | "measured";
2947
4512
  version: 1;
2948
- operation: "read_video_text";
4513
+ operation: "measure_audio_levels";
2949
4514
  source: {
2950
- encodedWidth: number;
2951
- encodedHeight: number;
2952
- displayWidth: number;
2953
- displayHeight: number;
2954
- rotationDegrees: number;
2955
- sampleAspectRatio: string;
2956
4515
  durationMs: number;
2957
- timeBase: string;
2958
- startPts: string;
2959
- startTimeMs: number;
2960
- frameCount: number;
2961
- variableFrameRate: boolean;
2962
4516
  sha256: string;
2963
4517
  bytes: number;
4518
+ hasAudio: boolean;
4519
+ originStartMs: number;
4520
+ audioStartPts: string | null;
4521
+ audioTimeBase: string | null;
4522
+ sampleRate: number | null;
4523
+ channels: number | null;
4524
+ channelLayout: string | null;
4525
+ audioOffsetMs: number | null;
2964
4526
  };
2965
- frames: {
2966
- endMs: number;
2967
- timeMs: number;
2968
- sourcePts: string;
2969
- frameIndex: number;
2970
- regions: {
2971
- status: "readable" | "unreadable";
2972
- box: {
2973
- width: number;
2974
- height: number;
2975
- x: number;
2976
- y: number;
2977
- };
2978
- id: string;
2979
- polygon: {
2980
- x: number;
2981
- y: number;
2982
- }[];
2983
- text: string | null;
2984
- rawText: string;
2985
- confidence: number;
2986
- }[];
2987
- lines: {
2988
- status: "readable" | "unreadable";
2989
- box: {
2990
- width: number;
2991
- height: number;
2992
- x: number;
2993
- y: number;
2994
- };
2995
- id: string;
2996
- text: string | null;
2997
- rawText: string;
2998
- confidence: number;
2999
- readingOrder: number;
3000
- regionIds: string[];
3001
- }[];
3002
- sceneChange: boolean;
3003
- previewUrl?: string | undefined;
3004
- }[];
3005
4527
  manifestUrl: string;
3006
- warnings: string[];
3007
- model: {
3008
- name: string;
3009
- runtime: string;
3010
- device: "cpu";
3011
- modelHashes: Record<string, string>;
3012
- orientationClassifier: string;
3013
- recognitionThreshold: number;
3014
- coordinates: "display-oriented-source-pixels";
4528
+ settings: {
4529
+ channelPolicy: "native-separate";
4530
+ windowMs: number;
4531
+ silenceThresholdDbfs: number;
4532
+ silenceMethod: "all-channel-window-rms";
4533
+ potentialClippingThreshold: number;
4534
+ loudnessMethod: "FFmpeg-ebur128-native-channels";
4535
+ ffmpegVersion: string;
3015
4536
  };
3016
- tracks: {
3017
- id: string;
3018
- text: string | null;
3019
- rawText: string;
3020
- firstObservedMs: number;
3021
- lastObservedEndMs: number;
3022
- observations: {
3023
- endMs: number;
3024
- timeMs: number;
3025
- sourcePts: string;
3026
- frameIndex: number;
3027
- box: {
3028
- width: number;
3029
- height: number;
3030
- x: number;
3031
- y: number;
3032
- };
3033
- confidence: number;
3034
- sampleIndex: number;
3035
- lineId: string;
3036
- }[];
3037
- }[];
4537
+ warnings: string[];
3038
4538
  coverage: {
3039
- complete: boolean;
3040
- range: {
4539
+ requestedRange: {
3041
4540
  startMs: number;
3042
4541
  endMs: number;
3043
4542
  };
3044
- mode: "fast" | "frame_scan";
3045
- uninspected: {
4543
+ analyzed: {
3046
4544
  startMs: number;
3047
4545
  endMs: number;
3048
4546
  }[];
3049
- inspected: {
4547
+ withoutAudio: {
3050
4548
  startMs: number;
3051
4549
  endMs: number;
3052
- sourcePts: string;
3053
- frameIndex: number;
3054
4550
  }[];
4551
+ sampleRoundingToleranceMs: number;
3055
4552
  };
4553
+ channels: {
4554
+ index: number;
4555
+ rmsDbfs: number | null;
4556
+ samplePeakDbfs: number | null;
4557
+ potentialClippedSamples: number;
4558
+ truePeakDbtp: number | null;
4559
+ sampleCount: number;
4560
+ intersampleOvershoot: boolean;
4561
+ }[];
4562
+ truePeakDbtp: number | null;
4563
+ windows: {
4564
+ startMs: number;
4565
+ endMs: number;
4566
+ channels: {
4567
+ index: number;
4568
+ rmsDbfs: number | null;
4569
+ samplePeakDbfs: number | null;
4570
+ potentialClippedSamples: number;
4571
+ }[];
4572
+ }[];
4573
+ silence: {
4574
+ startMs: number;
4575
+ endMs: number;
4576
+ }[];
4577
+ integratedLufs: number | null;
4578
+ loudnessTimeline: {
4579
+ startMs: number;
4580
+ endMs: number;
4581
+ momentaryLufs: number | null;
4582
+ shortTermLufs: number | null;
4583
+ }[];
4584
+ previewUrl?: string | undefined;
3056
4585
  }>]>;