first time using f5, it wouldn't generate the audio at all

[helprequiredthanks]

Checks

• [X] This template is only for usage issues encountered.
• [X] I have thoroughly reviewed the project documentation but couldn't find information to solve my problem.
• [X] I have searched for existing issues, including closed ones, and couldn't find a solution.
• [X] I confirm that I am using English to submit this report in order to facilitate communication.

Environment Details

windows 10, 64 bit, most recent Python, downloaded miniconda as instructed ALSO importantly the app wouldn't work at all until I further downloaded ffmpeg-7.1-full_build, from gyan, and then added the bin folder to PATH as ChatGpt advised me to when I consulted it on the error message showing on prompt Now I have already downloaded the ffmpeg, the "actual behavior" happened well after this

Steps to Reproduce

Sorry I have no experience in tech, I'm just trying to create audiobooks.

Installation I did as specified by

Create a python 3.10 conda env (you could also use virtualenv)

conda create -n f5-tts python=3.10 conda activate f5-tts

Install pytorch with your CUDA version, e.g.

pip install torch==2.3.0+cu118 torchaudio==2.3.0+cu118 --extra-index-url https://download.pytorch.org/whl/cu118

Then within a folder in my C file I opened command and did as specified by

git clone https://github.com/SWivid/F5-TTS.git cd F5-TTS pip install -e .

Then I used the line specified for launching the app f5-tts_infer-gradio

Then I opened http://127.0.0.1:7860/

Then I uploaded a brief audio clip and put in a brief text passage. See actual behavior pictures.

✔️ Expected Behavior

It was a brief audio reference, and a brief text passage, couldn't be simpler.

❌ Actual Behavior

It just won't generate the audio result. Also, when "generating", the computer case started getting very loud - only getting quiet when I terminated the program.

[SWivid]

there are two possible reasons:

1. check if you have used gpu to do inference
2. make sure not to click on the cmd inferface, otherwise there could be unintentionally pause command passed in

commonly, after getting to the phase you have got to, there is nothing blocking the inference process also, if you succeed in generating samples but receive blank output, consider upgrading torch to 2.4.0+cu121 or to order the dtype in infer_utils.py as fp32 (make it fixed)

thanks for providing detailed info, hope would help

(The repo owner's edit 2024/12/1: WARNING!!! The URL in the response of @helprequiredthanks below is an unofficial FAKE website, has no relationship with us.)

[helprequiredthanks]

Thanks. Could my internet connection be at fault somehow? (Does the app rely on a connection?) I just tried a similar thing in literally the online try section https://f5tts.org/playground

And 200 seconds later, nothing

That was bizarre

I haven't tried what you suggested yet, because this one really threw me off

[SWivid]

this is an unofficial fake website, has no relationship with us!

you can just follow the readme we have provide, where there are online demos!

[helprequiredthanks]

this is an unofficial fake website, has no relationship with us!

you can just follow the readme we have provide, where there are online demos!

Thank you, you are right, the demos you referred to are the correct ones. I'll maybe try Pinokio later, seems much easier platform to use apps on. Thanks for replying to my queries.

[SWivid]

glad could help

will close this issue, feel free to open if further questions~