The AI voice cloning tool also supports a variety of dialects and teaches you to quickly get started with the CosyVoice free voice cloning tool!

This tutorial goes through 2 examples: cloning the voices of Guo Degang and Lin Chiling to master the most real-life-likeAI voice cloning tool(Text-to-speech tool) - cosyvoice from Ali.

The entire tutorial contains:①Installation package, ②Exercise materials, ③Video tutorials, ④Text tutorials.

The tutorials are divided into "Getting Started" and "Getting Started".

Getting Started: I chose cosyvoice, which is packaged with "AI editing assistant", and it is characterized by its simplicity, which makes it easy for newcomers to get started. Disadvantages: cut out the dialect and other features.

Advanced: The cosyvoice packaged with "Walk with AI", which is characterized by: full functionality. The disadvantage is that it is slightly complicated.

so it is recommended thatBeginners read the introductory chapter first, know the basic operations. Then read the advanced chapter.

If you're a veteran, go straight to the advanced chapter.

Related resources to download:

(The resources are all the same, put more than a few network disk is afraid of being blocked, choose one of the download can be)

① Installation package:

introduction (a subject)Tools installation package (size: 11.4G):

Quark.com: https://pan.quark.cn/s/e49a0d238ba2

Baidu cloud disk: https://pan.baidu.com/s/17UtflIaaU-ZUC0EJu_Y6rg?pwd=2045

advanced chapterTool installation package (size: 8.8G):

Link to Quark.com: https://pan.quark.cn/s/a7ed622f5ae9

Baidu.com link: https://pan.baidu.com/s/1yX17U6lGDVoCI3doFt-POw?pwd=2045 Extract code: 2045

② Practice material:

It's the same for both introductory and advanced practice material.

Link to Quark.com: https://pan.quark.cn/s/19bac8977f95

(iii) Video tutorials:

(Includes introductory and advanced tutorials)

Link to Quark.com: https://pan.quark.cn/s/bb629f97ca8b

introduction (a subject)

1. What is cosyvoice?

Free voice cloning tool: provides 3 seconds of the original voice, both to copy the timbre.

Launched by Ali's Tongyi Labs.

2. 3 main features of cosyvoice?

1. With 3 seconds of original voice, you can clone (copy) anyone's speech.
2. Supports Chinese, English, Japanese, Korean, and Chinese dialects (Cantonese, Sichuan, Shanghainese, Tianjin, Wuhan, Changsha, Zhengzhou, etc.)
3. It can contain emotions: happiness, sadness, laughter, etc.

3. how does cosyvoice work? --Clone Guo's voice

This tutorial will demonstrate how to use cosyvoice by cloning Guo's voice.

Tool Description: The version of cosyvoice created by "AI Clip Assistant" is used.

Pros of this version:Especially suitable for newcomers, keeping only the most basic features and removing redundancies.

Step 1: Download the installation package

Download address for one-click installer (11.4G):

Quark Netflix:  https://pan.quark.cn/s/e49a0d238ba2

Baidu cloud drive: https://pan.baidu.com/s/17UtflIaaU-ZUC0EJu_Y6rg?pwd=2045

There are 2 files in the installation package:

Tool file: cosyvoice-2.

Exercise Material: Contains 10 seconds of audio and text from Guo Degang and Lin Chiling for practice.

Step 2: Copy the software to the root directory of C drive

Copy the software to the root directory of your C drive.

Special attention: Do not include Chinese in the installation path. Otherwise, it will cause an error.

Step 3: Run the software

① Open the cosyvoice-2 folder, find and double click the "go-web" file.

② The command line interface will appear. This interface, do not close it while the software is running.

After about 20~40 seconds, the following interface will appear: It means the software has been installed successfully.

Step 4: Add the voice of "Guo Degang".

Add Guo Degang's voice. The sound files are in the "Practice Materials" folder of the installer.

Click "Sound Model Management".

② Fill in the character name: Guo Deguang

③ Upload Guo Degang's 9-second audio

④ Fill in the reference audio text

Add new model

Check if the sound was added successfully

① Open the "Text to Audio" interface.

Click "Refresh Sound Model List".

③ In the "Sound Model List", you can see the sound you just added.

Step 5: Text-to-Speech

Enter text:

Hi everyone, this is my new friend Saiwen Yeh, he is a blogger who shares dry AI tips. Hurry up and follow him, [laughter] he's a handsome guy [laughter].

advanced chapter

How do I install it?

Download the installation package:

Link to Quark.com: https://pan.quark.cn/s/a7ed622f5ae9

Baidu.com link: https://pan.baidu.com/s/1yX17U6lGDVoCI3doFt-POw?pwd=2045 Extract code: 2045

Download practice material:

Link to Quark.com: https://pan.quark.cn/s/0070f8caeb08

Baidu.com link: https://pan.baidu.com/s/1OxVYYeAWxXKVxrNUt70Msg?pwd=2045 Extract code: 2045

Exercise material: Contains 2 voice files: the voices of Guo Deguang and Lin Zhiling, and the voice text.

How to use

Unzip the file and click on the "Start.exe" file.

Open it up and it looks like this:

1. Speaking in tongues

How to use cosyvoice's dialect

Select "Natural Language Control".

② Select the voice you want to clone. (You can choose Lin's voice from the installation package. (The sound cannot be less than 3 seconds and not more than 30 seconds)

③ Input the text of the cloned sound

④ Input the text to be converted to speech

⑤ Enter the dialect to be used. For example: in Sichuan

Currently supports Cantonese, Sichuan, Shanghai, Tianjin, Changsha and Zhengzhou.

2. Accession: laughter, breathing, emphasizing

How to use

How to use cosyvoice's dialect

Select "Natural Language Control".

② Select the voice you want to clone. (You can choose Lin's voice from the installation package. (The sound cannot be less than 3 seconds and not more than 30 seconds)

③ Input the text of the cloned sound

④ Enter the text to be text-to-speech. Note: Add laughter and emphasized grammar.

⑤ Generate audio

⑥ Download Audio

respiratory

Function: at the place of joining, there will be a breath change sound

Example: I'm not the poor boy I was then [breath], now I'm the poor boy of this year [breath].

Effect:

laughter

Function: To make a laughing sound

Example usage: I wrote the names of his whole family on the window in response to the fog, and when the fog cleared, his whole family was gone [laughter] [laughter].

effect

Function: To make a laughing sound

Usage examples: It's okay not to have a breath, it's great to have a breath.

Effect:

emphasize (a statement)

Role: Emphasizing content

Usage examples: why must people go to higher places, people can go to the<strong>round</strong>.

Effect:

3. Add emotions

Use the same method as in the dialect, write in "Enter instruct text" what emotions are needed.

Say it in a happy tone.

Example: I don't fight those who can beat me.

Effect:

Select "Natural Language Control".

② Select the voice you want to clone. (You can choose Lin's voice from the installation package. (The sound cannot be less than 3 seconds and not more than 30 seconds)

③ Input the text of the cloned sound

④ Input the text to be text-to-speech.

⑤ Input emotion words: e.g., say in a happy tone of voice

⑥ Generate audio

⑦ Download Audio

Say it in an angry tone.

Example: I felt very angry when I encountered a reckless driver cutting in line during rush hour traffic. This kind of uncivilized behavior always makes people helpless.

Effect:

4. Role-playing

In an innocent, romantic child's voice.

Select "Natural Language Control".

② Select the voice you want to clone. (You can choose Lin's voice from the installation package. (The sound cannot be less than 3 seconds and not more than 30 seconds)

③ Input the text of the cloned sound

④ Input the text to be text-to-speech.

⑤ Input Role

⑥ Generate audio

Download Audio

Instruct text (instruct text): an innocent child, always full of fantasy and endless curiosity.

TEXT: It's okay to be unimpressive, it's great to have breath.

5. General vocabulary of task descriptors (structural texts)

Organized from the official documentation.

Link to official documentation: https://funaudiollm.github.io/cosyvoice2

categorization	element
speak in tongues	Sichuan dialect
	Cantonese language
	Shanghai dialect
	Zhengzhou dialect
	Changsha dialect
	Tianjin dialect
Join: laughter, breathing, emphasis	[laughter]
	♪ laughter ♪
	<strong></strong>
	[breath]
Add emotion	Say it in a happy tone.
	Say it in a sad tone.
	Say it in a surprised tone.
	Say it in an angry tone.
	Expressed with the emotion of fear
	With disgusting expressions of emotion
	sobriety
	grave
	fast
	Very fast.
	slow speed
	Very slow.
role-playing (as a game of chess)	mystical
	fierce
	curiosity
	classy
	alone
	Mimic Robot Style
	I'd like to hear you mimic the tone of Peppa Pig's voice.
	A spunky, adventurous pixie, always with a smile and a whimsy.
	An authoritative, majestic ancient general with a voice like a bell and full of determination.
	A melancholy poet whose words always reveal a hint of sadness and romance.
	An innocent child, always full of fantasies and endless curiosity.
	An intelligent detective with a knack for picking up on nuance and thoughtfulness.

3. Frequently Asked Questions

3.1 Unsuccessful installation: No module named 'ttsfrd'

Cause of the problem:  Installation path in Chinese

Solution:  Copy the installation package to the root directory of your C drive

3.2. Installation path with Chinese characters

Error: FstIOError:read failded

Problem Cause: In the file installation path, there is Chinese.

Solution: Modify the installation path without Chinese.

3.3 Localhost is not accessible

Error: ValueError: When localhost is not accessible, a shareable link must be created. Please set share=True or check your proxy settings to allow access to Please set share=True or check your proxy settings to allow access to localhost.

Reason for the problem: Opened the international network

Solution: Close the international network

3.4 Other issues arise

There can be other problems of all sorts, mostly with the configuration of the computer.

Solution:

1. Rent an AliCloud computer (first month free).
2. Rent a GPU computer.
3. Tools for using the Magic Tower Online: https://www.modelscope.cn/models/iic/CosyVoice2-0.5B