simple-voice-chat 3.0.0

A simple voice chat interface using configurable LLM, STT, and TTS providers.

Simple Voice Chat

This project provides a flexible voice chat interface that connects to various Speech-to-Text (STT), Large Language Model (LLM), and Text-to-Speech (TTS) services.

Acknowledgement: This project heavily relies on the fantastic fastrtc library, which simplifies real-time audio streaming over WebRTC, making this application possible.

Motivation

The primary motivation for creating this project was the high cost associated with OpenAI's real-time voice API. This application allows you to leverage potentially more cost-effective or self-hosted alternatives for STT, LLM, and TTS, while still providing a near real-time voice interaction experience.

Features

• Modular: Connect to different STT, LLM (including local models via proxies like LiteLLM), and TTS providers.

  • STT: Defaults to using Speaches (which allows self-hosting Faster Whisper), but can also connect directly to OpenAI's Whisper API.
  • TTS: Defaults to OpenAI TTS, but also supports alternatives like Kokoro-FastAPI.
  • LLM: Supports virtually any LLM provider (OpenAI, Anthropic, Google, Mistral, Cohere, Azure, local models, etc.) thanks to its integration with LiteLLM. You can connect to any OpenAI-compatible API endpoint, including local models served via proxies like LiteLLM itself, vLLM, or Ollama.

• Configurable: Fine-tune various parameters for STT confidence, TTS voice/speed, LLM model selection, and more via command-line arguments or environment variables.

• Web Interface: Provides a simple web-based UI for interaction.

• Cost Tracking: Includes basic cost estimation for supported LLM and TTS providers (like OpenAI).

Installation

1. Clone the repository:

   git clone https://github.com/thiswillbeyourgithub/simple_voice_chat
   
   cd simple_voice_chat
   

2. Install the Python packages:

   uv pip install -e .
   

3. (Optional) Configure services using environment variables. You can create a .env file based on the available options (see --help or utils/env.py).

Usage

Run the main script using Python:

simple-voice-chat --help

The application will start a web server and attempt to open the interface in a dedicated window (or browser tab if --browser is specified).

For a detailed list of all configuration options (STT/LLM/TTS hosts, ports, models, API keys, etc.), please use the --help flag:

simple-voice-chat --help

This will provide the most up-to-date information on available arguments and their corresponding environment variables.

Command-Line Options (--help)

usage: simple_voice_chat.py [-h] [--host HOST] [--port PORT] [-v]
                               [--auto-start | --no-auto-start] [--browser]
                               [--system-message SYSTEM_MESSAGE]
                               [--llm-host LLM_HOST] [--llm-port LLM_PORT]
                               [--llm-model LLM_MODEL]
                               [--llm-api-key LLM_API_KEY]
                               [--stt-host STT_HOST] [--stt-port STT_PORT]
                               [--stt-model STT_MODEL]
                               [--stt-language STT_LANGUAGE]
                               [--stt-api-key STT_API_KEY]
                               [--stt-no-speech-prob-threshold STT_NO_SPEECH_PROB_THRESHOLD]
                               [--stt-avg-logprob-threshold STT_AVG_LOGPROB_THRESHOLD]
                               [--stt-min-words-threshold STT_MIN_WORDS_THRESHOLD]
                               [--tts-host TTS_HOST] [--tts-port TTS_PORT]
                               [--tts-model TTS_MODEL] [--tts-voice TTS_VOICE]
                               [--tts-api-key TTS_API_KEY]
                               [--tts-speed TTS_SPEED]
                               [--tts-acronym-preserve-list TTS_ACRONYM_PRESERVE_LIST]
Run a simple voice chat interface using a configurable LLM provider, STT
server, and TTS.
options:
-h, --help            show this help message and exit
--host HOST           Host address to bind the FastAPI server to. Default:
127.0.0.1
--port PORT           Preferred port to run the FastAPI server on. Default:
7860. (Env: APP_PORT)
-v, --verbose         Enable verbose logging (DEBUG level)
--auto-start, --no-auto-start
Automatically start the connection when the
application loads. Default: True
--browser             Launch the application in the default web browser
instead of a dedicated GUI window. Default: False
--system-message SYSTEM_MESSAGE
System message to prepend to the chat history.
Default: (from SYSTEM_MESSAGE env var, empty if
unset).
--llm-host LLM_HOST   Host address of the LLM proxy server (optional).
Default: None. (Env: LLM_HOST)
--llm-port LLM_PORT   Port of the LLM proxy server (optional). Default:
None. (Env: LLM_PORT)
--llm-model LLM_MODEL
Default LLM model to use (e.g., 'gpt-4o',
'litellm_proxy/claude-3-opus'). Default:
'litellm_proxy/litellm_best'. (Env: LLM_MODEL)
--llm-api-key LLM_API_KEY
API key for the LLM provider/proxy (optional, depends
on setup). Default: None. (Env: LLM_API_KEY)
--stt-host STT_HOST   Host address of the STT server (e.g., 'api.openai.com'
or 'localhost'). Default: 'localhost'. (Env: STT_HOST)
--stt-port STT_PORT   Port of the STT server (e.g., 443 for OpenAI, 8002 for
local). Default: '8002'. (Env: STT_PORT)
--stt-model STT_MODEL
STT model to use (e.g., 'whisper-1' for OpenAI,
'deepdml/faster-whisper-large-v3-turbo-ct2' for
local). Default: 'deepdml/faster-whisper-
large-v3-turbo-ct2'. (Env: STT_MODEL)
--stt-language STT_LANGUAGE
Language code for STT (e.g., 'en', 'fr'). If unset,
Whisper usually auto-detects. Default: None. (Env:
STT_LANGUAGE)
--stt-api-key STT_API_KEY
API key for the STT server (REQUIRED for OpenAI STT).
Default: None. (Env: STT_API_KEY)
--stt-no-speech-prob-threshold STT_NO_SPEECH_PROB_THRESHOLD
STT confidence threshold: Reject if no_speech_prob is
higher than this. Default: 0.6. (Env:
STT_NO_SPEECH_PROB_THRESHOLD)
--stt-avg-logprob-threshold STT_AVG_LOGPROB_THRESHOLD
STT confidence threshold: Reject if avg_logprob is
lower than this. Default: -0.7. (Env:
STT_AVG_LOGPROB_THRESHOLD)
--stt-min-words-threshold STT_MIN_WORDS_THRESHOLD
STT confidence threshold: Reject if the number of
words is less than this. Default: 5. (Env:
STT_MIN_WORDS_THRESHOLD)
--tts-host TTS_HOST   Host address of the TTS server (e.g., 'api.openai.com'
or 'localhost'). Default: 'api.openai.com'. (Env:
TTS_HOST)
--tts-port TTS_PORT   Port of the TTS server (e.g., 443 for OpenAI, 8880 for
local). Default: '443'. (Env: TTS_PORT)
--tts-model TTS_MODEL
TTS model to use (e.g., 'tts-1', 'tts-1-hd' for
OpenAI, 'kokoro' for local). Default: 'tts-1'. (Env:
TTS_MODEL)
--tts-voice TTS_VOICE
Default TTS voice to use (e.g., 'alloy', 'ash', 'echo'
for OpenAI, 'ff_siwis' for local). Default: 'nova'.
(Env: TTS_VOICE)
--tts-api-key TTS_API_KEY
API key for the TTS server (REQUIRED for OpenAI TTS).
Default: None. (Env: TTS_API_KEY)
--tts-speed TTS_SPEED
Default TTS speed multiplier. Default: 1.00. (Env:
TTS_SPEED)
--tts-acronym-preserve-list TTS_ACRONYM_PRESERVE_LIST
Comma-separated list of acronyms to preserve during
TTS (currently only used for Kokoro TTS). Default: ''.
(Env: TTS_ACRONYM_PRESERVE_LIST)

---

This README was generated with assistance from aider.chat.