Documentation

[HibsMaxGit]

Hello.

I saw your TTS on a recent YouTube video, and it looks like an exciting development. I installed it locally, but I am having problems running even a simple example. The documentation on the local web UI is a little light on information. Is that something you are going to expand, or maybe even provide some YouTube videos of your own? For example, what is the difference between Prompt Text and Target Text? Is it:
Prompt Text - the speech you want to generate
Target Text - the transcript of the audio recording that is uploaded for cloning

Thanks, Max

[HibsMaxGit]

Here's an example from the Web UI:
"If the task requires modifying text content (such as clone, para-linguistic), fill in the text to be synthesized in the "clone text" field. For all other tasks, keep the uploaded audio text content unchanged;"

There is no "clone text" field.

I think I am mostly there, but some fine-tuning of the documents will definitely help (at least me).

[yanchaomars]

"clone text" -> "target text"
Sorry for the typo. 🤦 we'll fix it.

[HibsMaxGit]

It doesn't seem this Discussion area is actively monitored. I will continue using VibeVoice for now.

[yanchaomars]

Hello.

I saw your TTS on a recent YouTube video, and it looks like an exciting development. I installed it locally, but I am having problems running even a simple example. The documentation on the local web UI is a little light on information. Is that something you are going to expand, or maybe even provide some YouTube videos of your own? For example, what is the difference between Prompt Text and Target Text? Is it: Prompt Text - the speech you want to generate Target Text - the transcript of the audio recording that is uploaded for cloning

Thanks, Max

Yes, we plan to release a detailed video.
“Prompt Text” is the transcript of the audio recording that is uploaded for cloning or editing
"Target Text" is the speech you want to generate.

[OccasionallyCrepuscular]

Yes, we plan to release a detailed video. “Prompt Text” is the transcript of the audio recording that is uploaded for cloning or editing "Target Text" is the speech you want to generate.

1. Please make a detailed accessible tutorial video where you read out exactly what you are doing with details, people who need TTS who cannot read due to fatigue (whether Long Covid or other illnesses) find it incredibly difficult to learn new things such as how to install this on Linux/Windows from GitHub and deal with any installation issues on top of having enough energy to function on a daily basis.

2. An option to have an emotionless voice (that old TTS has) would be helpful too for those with complex trauma who simply may want to read an angry email they have to read ,without being intensely impacted (a calm voice takes the edge off).

3. Lovely to see an open source accessibility tool that is so advanced but sad that it's not accessible except to the tech savvy for now. Please do consider an option for low and very low sibilance voices too.

4. I can't tell if you have fine grain controls of speed as that's very important especially in the slow category for people recovering from stroke and other medical conditions which most TTS do not allow below 0.5 (also speed is not standardised across companies that created the TTS)

Thanks for the hard work you've put into this and I hope some day I will get to use it to read book PDFs in voices I actually really enjoy because for now I'm stuck with the Google speech synthesis voices.Best of luck with the project!