Noise at the end of generated voice #478

Closed
opened 2024-03-04 08:26:29 +00:00 by tortoise · 6 comments

Dear friends, I have to ask that after the generation of voice from the mode, there is a wired noise at the end which is always the part of exported wav. Is there something which is missing in my workflow or any idea how to get rid of it.

Dear friends, I have to ask that after the generation of voice from the mode, there is a wired noise at the end which is always the part of exported wav. Is there something which is missing in my workflow or any idea how to get rid of it.

Could be temperature or often its just something missed in the training samples.

Could be temperature or often its just something missed in the training samples.

Hey! Have you checked whether you have a blank line in your prompt? I mean something like this:
"The first line of your prompt
[an empty space here]"

I've noted that you need to make sure there is no other line behind the last line of your prompt, else it will generate a noise like the speaker is having a stroke.

Not sure whether I'm being clear, do tell me if I can explain better / if it helps!

Hey! Have you checked whether you have a blank line in your prompt? I mean something like this: "The first line of your prompt [an empty space here]" I've noted that you need to make sure there is no other line behind the last line of your prompt, else it will generate a noise like the speaker is having a stroke. Not sure whether I'm being clear, do tell me if I can explain better / if it helps!
Author

@DoctorPopi So far, there is no empty line I have at the end of the line. It is the case, where it finishes sentence and combines with other sentence, this is also the case. Sometimes, for 1-2 seconds in the start of sentence too.

@DoctorPopi So far, there is no empty line I have at the end of the line. It is the case, where it finishes sentence and combines with other sentence, this is also the case. Sometimes, for 1-2 seconds in the start of sentence too.
Author

Could be temperature or often its just something missed in the training samples.

where and how I could check the samples or the temperature. and is there any temperature value while training I have seen only while voice generation. I have kept that temperature 0.8? if you could guide.

> Could be temperature or often its just something missed in the training samples. where and how I could check the samples or the temperature. and is there any temperature value while training I have seen only while voice generation. I have kept that temperature 0.8? if you could guide.
Author

@gforce @DoctorPopi if there is any more advise to make it better and get rid of such noise at the end of sentence? This occurs specially when text is more than 2 sentences.

@gforce @DoctorPopi if there is any more advise to make it better and get rid of such noise at the end of sentence? This occurs specially when text is more than 2 sentences.
Author

Hey! Have you checked whether you have a blank line in your prompt? I mean something like this:
"The first line of your prompt
[an empty space here]"

I've noted that you need to make sure there is no other line behind the last line of your prompt, else it will generate a noise like the speaker is having a stroke.

Not sure whether I'm being clear, do tell me if I can explain better / if it helps!

Thank you so much again. I have almost fixed with your advise.

> Hey! Have you checked whether you have a blank line in your prompt? I mean something like this: > "The first line of your prompt > [an empty space here]" > > I've noted that you need to make sure there is no other line behind the last line of your prompt, else it will generate a noise like the speaker is having a stroke. > > Not sure whether I'm being clear, do tell me if I can explain better / if it helps! Thank you so much again. I have almost fixed with your advise.
Sign in to join this conversation.
No Milestone
No project
No Assignees
3 Participants
Notifications
Due Date
The due date is invalid or out of range. Please use the format 'yyyy-mm-dd'.

No due date set.

Dependencies

No dependencies set.

Reference: mrq/ai-voice-cloning#478
No description provided.