04:39 I was mistaken, it actually works by generating audio that matches the duration you specify. For example, if you set it to 10 seconds, it will produce 10 seconds of audio, adjusting the speed as needed based on the length of your input text.
I have a question, I love your videos man! so whenever I write like a pharagraph (500 words) it keeps crashing and giving an error, do you have a fix for it?
I just tested them both, F5-TTS is wayyy faster. I was able to generate a 10mins audio in about 1 min and 30 seconds. MaskGCT could only generate 15 seconds of audio in 25 seconds.
I have a question, I love your videos man! so whenever I write like a pharagraph (500 words) it keeps crashing and giving an error, do you have a fix for it?
I fixed the bug. Check the GitHub repository and click on the Google Colab link. I found an error related to NLTK in Google Colab. Make sure to click on 'Long Text = True' in the Gradio app. Still, I suggest using smaller text and later merging all the cloned voices with any editing software, as there is a chance of encountering a CUDA out-of-memory error with large text. However, we can try it. I wrote some code to handle the CUDA out-of-memory error, but I have never tried it with large text.
For now it only supports English and Chinese, they promise to upload training code github.com/open-mmlab/Amphion/issues/289 I think we can train the model for different language in future. For now you need to use XTTS v2 for Portuguese
04:39 I was mistaken, it actually works by generating audio that matches the duration you specify. For example, if you set it to 10 seconds, it will produce 10 seconds of audio, adjusting the speed as needed based on the length of your input text.
Plz make video for how can locally setup xtts in laptop but I can not knowledge about python plz make video I need pl
@@Rightpath-dr7ys Just Download Portable version.
github.com/daswer123/xtts-webui
I have a question, I love your videos man! so whenever I write like a pharagraph (500 words) it keeps crashing and giving an error, do you have a fix for it?
5:48 is creepy 😁😅
Hummm
I just tested them both, F5-TTS is wayyy faster. I was able to generate a 10mins audio in about 1 min and 30 seconds. MaskGCT could only generate 15 seconds of audio in 25 seconds.
Yes F5-TTS is faster compared to the MaskGCT. But it also depends on which GPU you are using.
This better than f5tts
Sometime it works better 😀
I have a question, I love your videos man! so whenever I write like a pharagraph (500 words) it keeps crashing and giving an error, do you have a fix for it?
Enable debug mode on Gradio to identify the exact error.
I will try it myself later and let you know.
I fixed the bug. Check the GitHub repository and click on the Google Colab link.
I found an error related to NLTK in Google Colab. Make sure to click on 'Long Text = True' in the Gradio app. Still, I suggest using smaller text and later merging all the cloned voices with any editing software, as there is a chance of encountering a CUDA out-of-memory error with large text. However, we can try it. I wrote some code to handle the CUDA out-of-memory error, but I have never tried it with large text.
Do you think it's possible to fine tune it for Portuguese? I wanted another TTS model besides XTTS v2 for my research
For now it only supports English and Chinese, they promise to upload training code
github.com/open-mmlab/Amphion/issues/289
I think we can train the model for different language in future. For now you need to use XTTS v2 for Portuguese
Watch this video : th-cam.com/video/UO4usaOojys/w-d-xo.html
@@neuralfalconthank you!! I’ll take a look
@@diogopereira0323 did you take a look whats the best one right now. I use xtts for now through pinokio
does this support to clone Indonesian voices? please answer🙏
NO, Only English and Chinese
You can try paid elevenlabs
You could try using Coqui TTS. Someone has trained a model on the Indonesian language, available here: github.com/ZahrizhalAli/indonesian-tts-vits
@@neuralfalconNext time, please make a video that supports all languages, one of which is Indonesian🙏😥
Watch this video : th-cam.com/video/UO4usaOojys/w-d-xo.html