“I can’t find the hellpick! Should I build another or climb all the way to Jimmy’s office?”
Speaker embeddings are first class. I averaged with random() and the accent changed but still high-pitched and such, I wonder which of these numbers is gender.
speaker = tensor([ 0.1066, 0.0259, 0.5086, 0.3959, 0.0881, -0.0490, 0.0223, 0.8254,
0.3045, 0.4574, 0.0989, 0.6380, 0.6526, 0.0638, 0.3042, 0.4008,
0.2482, 0.2559, 0.3023, 0.0516, 0.7169, -0.0523, 0.5802, -0.3008,
0.3744, 0.2430, 0.6635, 0.2458, 0.2221, -0.3828, 0.7512, 0.3135,
0.0860, 0.2542, 0.2730, 0.2849, 0.3722, 0.7822, 0.0410, -0.0560,
0.1284, 0.6072, -0.0794, -0.0866, 0.5913, 0.3724, 0.1841, 0.1808,
0.3498, 0.2965, 0.4538, 0.0776, 0.2383, 1.1943, 0.3907, 0.2430,
0.1141, 0.2758, -0.1908, 0.2067, 0.3293, 0.2710, 0.1114, 0.7877,
0.1905, 0.3969, 0.4095, 0.4954, -0.0683, 0.4314, 0.4397, 0.8943,
0.3643, 0.4579, 0.4021, 0.4536, 0.1490, 0.1193, -0.0533, 0.0506,
0.1852, 0.5002, 0.5305, 0.0351, 0.2539, 0.2448, 0.1736, 0.1557,
0.4788, 0.1384, -0.0533, 0.9108, 0.0489, 0.4524, 0.1497, 0.4435,
0.1382, 0.1280, -0.3168, 0.2600, 0.1717, 0.1125, 0.6440, 0.1540,
0.1440, 0.0202, 0.3108, 0.2440, 0.0066, 0.0082, 0.4545, 0.3424,
0.3098, 0.5659, 0.0037, 0.2884, 0.2409, 0.3569, 0.6281, 0.5068,
0.4341, 0.6150, 0.6827, 0.5059, 0.7445, 0.4291, 0.6599, 0.7472,
-0.0086, 0.0688, 0.5493, 0.1909, 0.2528, 0.0396, 0.2264, -0.2069,
0.3067, 0.2213, 0.5432, 0.6556, 0.3711, 0.0612, 0.0288, -0.3123,
0.0370, 0.1657, 0.1064, 0.3140, -0.1349, 0.0481, 0.3594, 0.4289,
0.3617, 0.5209, 0.3829, 0.3175, 0.3119, 0.1147, 0.1306, 0.4851,
0.4073, -0.1298, 0.4898, 0.1434, 0.3283, 0.0168, -0.0554, 0.1544,
-0.0052, 0.1071, 0.5624, 0.2370, 0.5639, 0.2203, 0.4902, 0.1865,
-0.2933, 0.1848, 0.4513, 0.6684, 0.1123, -0.1688, 0.3635, 0.1639,
0.4584, 0.3960, 0.0929, -0.1089, 0.2814, 0.5176, 0.4132, 0.2479])