TextToSpeech¶

class agentlego.tools.TextToSpeech(model='tts_models/multilingual/multi-dataset/xtts_v2', speaker_embeddings=SPEAKER_EMBEDDING, device='cuda', toolmeta=None)[source]

A tool to convert input text to speech audio.

Parameters:

model (str) – The model name used to inference. Which can be found in https://github.com/coqui-ai/TTSHuggingFace . Defaults to tts_models/multilingual/multi-dataset/xtts_v2.
speaker_embeddings (str | dict) – The speaker embedding of the TTS model. Defaults to a default embedding.
device (str) – The device to load the model. Defaults to ‘cuda’.
toolmeta (None | dict | ToolMeta) – The additional info of the tool. Defaults to None.

Default Tool Meta¶

name: TextToSpeech
description: The tool can speak the input text into audio. The language code should be one of ‘zh-cn’ (Chinese), ‘en’ (English), ‘es’ (Spanish), ‘fr’ (French), ‘de’ (German), ‘it’ (Italian), ‘tr’ (Turkish), ‘ru’ (Russian), ‘ar’ (Arabic), ‘ja’ (Japanese), ‘ko’ (Korean).
inputs:
- text (str)
- lang (str): The language code of text.
outputs:
- AudioIO

Examples¶

Use the tool directly (without agent)

from agentlego.apis import load_tool

# load tool
tool = load_tool('TextToSpeech', device='cuda')

# apply tool
audio = tool('Hello, this is a text to audio demo.')
print(audio)

With Lagent

from lagent import ReAct, GPTAPI, ActionExecutor
from agentlego.apis import load_tool

# load tools and build agent
# please set `OPENAI_API_KEY` in your environment variable.
tool = load_tool('TextToSpeech', device='cuda').to_lagent()
agent = ReAct(GPTAPI(temperature=0.), action_executor=ActionExecutor([tool]))

# agent running with the tool.
ret = agent.chat(f'Please introduce the highest mountain and speak out.')
for step in ret.inner_steps[1:]:
    print('------')
    print(step['content'])

Set up¶

Before using the tool, please confirm you have installed the related dependencies by the below commands.

pip install TTS, langid

Reference¶

This tool uses a XTTS-v2 model in default settings. See the repo details.