Duration Issue with the generated audio

[Aliasgarsaifee]

Whenever I do inference with VITS model it produces relatively fast speaking audio as compared to GT. Using these parameters in inference -

• length_scale=1.1
• noise_scale=0.667
• noise_scale_w=0.8

Is there something I am missing?

Model Configs - "model": { "inter_channels": 192, "hidden_channels": 192, "filter_channels": 768, "n_heads": 2, "n_layers": 6, "kernel_size": 3, "p_dropout": 0.1, "resblock": "1", "resblock_kernel_sizes": [3,7,11], "resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]], "upsample_rates": [8,8,2,2], "upsample_initial_channel": 512, "upsample_kernel_sizes": [16,16,4,4], "n_layers_q": 3, "use_spectral_norm": false, "gin_channels": 375, "lin_channels" : 192 }

[bjkite]

The same problem. Why?

[rishav-bose0]

You can change length_scale to alter the relative speed of the audio. Also, you can play around with noise_scale and noise_scale_w for audio quality. Increasing noise_scale will make the speech incomprehensible

[phamkhactu]

HI @Aliasgarsaifee @rishav-bose0

I've just change length_scale=2. But speed don't change with length_scale=1. Could you share your exp about it?