nonebot plugin fishspeech tts Download - nonebot plugin fishspeech tts Source code download

nonebot plugin fishspeech tts

AI Source Code

v2024.12.09-07.43.52

Download

nonebot-plugin-fishspeech-tts

Nonebot2-based call to online fish-audio or offline fish-speech api text generation speech tts plugin

Chinese Simplified | English

introduce

Only a 5-second voice material can be used ~~to perfectly~~ clone the original sound! Just prepare the character voice you want to clone and label its voice filename (see below) to quickly generate voice.

Or use the official online api -> fish-audio to enjoy fast cloud voice generation.

Disclaimer

Caution

This plug-in is for learning and research purposes only, and users must bear the risk of using the plug-in at their own risk. The author is not responsible for any losses or problems caused by the use of the plugin. Please use the plug-in reasonably and abide by relevant laws and regulations. By using this plug-in, you have read and agree to abide by the above disclaimer . If you do not agree or fail to comply with the above statement, please do not use this plug-in.

? Install

Install

pipx installation

pipx install nonebot-plugin-fishspeech-tts -U

[!note] Add this plugin in pyproject.toml of nonebot

nb-cli installation

nb plugin install nonebot-plugin-fishspeech-tts -U

git clone installation (not recommended)

Run under the command window cmd

git clone https://github.com/Cvandia/nonebot-plugin-fishspeech-tts

Copy the folder nonebot-plugin-fishspeech-tts at the window run to src/plugins in the root directory of the bot (or other name xxx/plugins when creating the bot)

Notice

Recommended mirror site download

Tsinghua Source https://pypi.tuna.tsinghua.edu.cn/simple

Alibaba Source https://mirrors.aliyun.com/pypi/simple/

Configuration

Add the following configuration in .env

Basic configuration	type	Required	default value	illustrate
tts_is_online	bool	yes	True	Whether to use cloud API
tts_chunk_length	literal	no	"normal"	The audio shard length is normal when requested, optional: short, normal, long
tts_max_new_tokens	int	no	800	Maximum audio length, default is 800, setting to 0 is unlimited
tts_audio_path	str	no	"./data/reference audio"	Voice material path, default is "./data/reference audio"
tts_prefix	str	no	None	Trigger prefix, default to None

Note: The file name format of the reference audio is: [Role Name] The text label corresponding to the audio. [Audio suffix name]

Click here to select to download the recommended reference audio reference_audio.zip , decompress and put the audio into data/参考音频in the bot file directory.

! Support different voices of the same character!

For details on audio suffix support, please refer to AUDIO_FILE_SUFFIX in files.py

If you want to use the official API, please set the configuration item tts_is_online to True and configure the following

Configuration Items	type	Required	default value	illustrate
online_api_url	str	no	"https://api.fish-audio.cn"	The official website API address can be selected from https://api.fish.audio (walled) or default value
online_authorization	str	yes	"xxxxx"	Official API authentication key, please see the link for details
online_model_first	bool	no	True	If you want to call the official model, customize the character tone through your own reference audio, set this to `False` . Of course, if you do not prepare reference audio, you will also call the tone already available on the official website. For details, please refer to the link
online_api_proxy	str	no	None	Proxy address, such as: http://127.0.0.1:7890

If you want to use the API of your own or other fish-speech project, please set the configuration item tts_is_online to Fasle and configure the following

Configuration Items	type	Required	default value	illustrate
offline_api_url offline_api_url	str	yes	"http://127.0.0.1:8080"	Your `fish-speech` api address

use

Note

Please note your COMMAND_START and the above configuration items.

instruction:

instruction	need@	scope	illustrate	Permissions
xxx says xxxx	According to configuration	all	tts voice generation	all
Voice list	yes	all	Get a list of all roles	all
Voice balance	yes	all	Query API balance	all

? Todo

Add more configuration items
No plans yet

If you like it, remember to click a star

? Special thanks

nonebot2: The basics of this project, a very easy-to-use chat robot framework.
fish-speech: Zero sample & small sample TTS: Enter a sound sample of 10 to 30 seconds to generate high-quality TTS output

Additional Help

Brief tutorial on building windows offline

Before use

CUDA toolkit ensures that the correct version of CUDA is installed before use

Prepare fish-speech

1. Download the fish-speech repository release code ( Source code (zip) ) to the local
2. Unzip to local
3. Open the fish-speech root directory
4. Run install_env.bat to install the virtual environment and the required dependencies
5. Run start.bat for the first time
6. Modify API_FLAGS.txt and start it again

Start API service

1. Modify API_FLAGS.txt to roughly be the following content, that is, cancel # number in front of api

 # --infer
--api
--listen 0.0.0.0:8080  #监听接口
...

Additional parameters that can be added in API_FLAGS.txt

1. --complie -> Whether to start the compiled model (faster generation of tts, but slower startup)
2. --workers 数字>Start数字multi-coroutines (please be sure to set it, because the default one is easy to block)
3. --device cpu -> Use cpu to generate tts (if using gpu, please ignore)
4. --half -> Use half precision to generate tts
5. --max-text-length -> Enter the maximum length of text