Prince Canuma comments

Results 572 comments of


                                            Prince Canuma

Model type kokoro not supported

Please install from source :)

Is mixing voices possible?

My pleasure! That's an interesting idea, I will experiment in the upcoming release or right after.

Implement Weight Normalization, addressing issue #1888

Thanks a lot @cavit99, this is great work! One tiny nit: - Could we change the weight naming to `weight_g` and `weight_v`? Makes easier to map from torch and remember.

Implement Weight Normalization, addressing issue #1888

As far as I can tell from intial testing. This PR does address my issues.. The only difference is that I prefered using torch channel first for loading and transposed...

Implement Weight Normalization, addressing issue #1888

Perfect! 🤩 Now we wait for @awni :)

Implement Weight Normalization, addressing issue #1888

@cavit99 I'm porting SparkTTS and this would be a great addition :) Turns out 90% of the audio models we ported in MLX-Audio use WeightNorm layer either in the model...

Add ability to use a custom URL to load Kokoro safetensors

Hey @adrgrondin Thank you very much for this awesome contribution! What do you think if we add support for downloading model info from huggingface using [swift-transformers](https://github.com/huggingface/swift-transformers)? Since we now support...

Add ability to use a custom URL to load Kokoro safetensors

Sounds good! Please add the custom url to Orpheus too

Add ability to use a custom URL to load Kokoro safetensors

Thanks @adrgrondin! Sorry for the delay we have been busy on some work behind the scenes for mlx-audio.

Any way to Speed up short TTS?

Could you share more details and perhaps a reproducible example? What version you are using?