Ecosyste.ms: Timeline
Browse the timeline of events for every public repo on GitHub. Data updated hourly from GH Archive.
ZhikangNiu created a comment on an issue on SWivid/F5-TTS
> Sorry for having so many questions! But can you explain what differences? The parameters look compatible: https://huggingface.co/nvidia/bigvgan_v2_24khz_100band_256x/blob/main/config.json "num_me...
SWivid created a comment on an issue on SWivid/F5-TTS
> Im using it with Pinokio Not sure if pinokio also compatible with training/finetuning. If you just want to do inference, use `gradio_app.py` and `inference-cli.py`
kmn1024 created a comment on an issue on SWivid/F5-TTS
Sorry for having so many questions! But can you explain what differences? The parameters look compatible: https://huggingface.co/nvidia/bigvgan_v2_24khz_100band_256x/blob/main/config.json "nu...
ZhikangNiu created a comment on an issue on SWivid/F5-TTS
> Why does this need a retrain? I thought F5 just outputs mel-spectrograms without any dependencies on the downstream vocoder used for inference? They have some difference in extract mel-spectro...
kunci115 closed an issue on SWivid/F5-TTS
preparing dataset for finetune
hi, i know you have scripts/prepare_emilia.py my question is how if i have my own dataset? what format should i provide? what is the structure needed to finetuned? the docs is not that clear
kmn1024 created a comment on an issue on SWivid/F5-TTS
Why does this need a retrain? I thought F5 just outputs mel-spectrograms without any dependencies on the downstream vocoder used for inference?
kmn1024 created a comment on an issue on SWivid/F5-TTS
Does this work? https://huggingface.co/nvidia/bigvgan_v2_24khz_100band_256x The params seem to match: sample_rate: 24000 n_fft: 1024 hop_length: 256 n_mels: 100
kunci115 created a comment on an issue on SWivid/F5-TTS
> @SWivid thank you for the amazing repo ! > > @kunci115 i create simple script prepare dataset to use with format csv local > > ``` > my_data/ > │ > ├── wavs/ > │ ├── audio1.wav > │ ...