RealtimeTTS

RealtimeTTS is a Python text-to-speech library that turns strings, generators, and LLM streams into audio with low latency.

RealtimeTTS is a Python text-to-speech library designed for applications that require low-latency audio generation from strings, generators, and LLM token streams. It plays speech locally, streams chunks to other processes, writes WAV files, and falls back across multiple engines. The project supports local system voices, cloud APIs, free service wrappers, local neural models, and voice-cloning stacks including Qwen, Inflect, OpenAI, Azure, and ElevenLabs. It is built for developers integrating text-to-speech into Python applications. Key capabilities include streaming text inputs, callbacks, voice switching, and OpenAI-compatible HTTP server support.

RealtimeTTS at a glance

Best for: Software engineer
Runs on: Python, Windows, macOS, Linux, self-hosted, open source, API
Pricing: Free
Released: August 2023
Category: AI audio tool
Website: github.com/KoljaB/RealtimeTTS

Key features

Low-latency playback from strings, generators, and streamed model output
Multiple engines with local, cloud, free-service, and neural model options
Fallback engines for more resilient synthesis
Sync and async playback with pause, resume, stop, and state inspection
Text, audio, sentence, character, word-timing, and audio-chunk callbacks
WAV output, muted synthesis, selected output devices, and volume control
Voice switching and voice-cloning workflows

Use it for

Streaming text-to-speech from LLM token streams
Generating audio files from text strings
Running local neural text-to-speech models
Integrating cloud text-to-speech APIs

Alternatives to RealtimeTTS

RealtimeSTT — Provides fast local speech recognition with voice activity detection and wake words.
Speech-to-Speech — Fully modular, swappable components with local or hosted execution options.
Whisper Timestamped — Generates accurate word-level timestamps without extra neural networks.
WeNet — Achieves state-of-the-art results on public speech datasets.
ESPnet — Provides reproducible recipes and hundreds of pretrained models for various speech tasks.
AssemblyAI — Industry-leading speech models deliver high accuracy and low latency for voice applications.
Pipecat — Builds real-time voice and multimodal agents with frame-level control.
kokoro-onnx — Runs local TTS inference with efficient ONNX runtime performance.
MOSS-TTS — Delivers zero-shot voice cloning and fine-grained pronunciation control.
audio.cpp — Runs modern audio models locally with high-performance, native C++ execution.
Piper — Runs locally and offers fast neural text-to-speech synthesis.
TTS WebUI — It offers a free web interface with over 30 AI models for audio generation.

All RealtimeTTS alternatives