tts

A free, fully local voice cloning and TTS tool powered by Qwen3-TTS and Kokoro runs on Apple Silicon via MLX, enabling studio-quality audiobook generation from PDFs without cloud services.

VibeVoice 1.5B, a long-form multi-speaker TTS model, is now supported in audio.cpp, a native C++/ggml runtime, achieving 4.08x real-time speed on RTX 5090, 2.86x faster than Python baseline without quantization.

The developer improved qwen3-tts.cpp to run 5x realtime on RTX 5080 and created a cross-platform desktop GUI with Kotlin Compose Multiplatform, featuring voice cloning, streaming, and speaker embedding management.

Inflect-Micro-v2 is a compact text-to-speech model with under 10M parameters, supporting CPU/CUDA inference and long-text handling, released on Hugging Face.

A fully free and open-source HTTP API service based on macOS native capabilities, offering image OCR, multi-language translation, web content retrieval, face recognition, QR/barcode recognition, and text-to-speech.

MOSS-TTS-Local Transformer v1.5 is an open-source 48 kHz stereo TTS model with zero-shot voice cloning, native streaming, and support for 31 languages, built on a Qwen3-4B backbone and served via SGLang-Omni.

MosiAI has released MOSS-TTS Local Transformer v1.5, a text-to-speech model that supports voice cloning, over 30 languages, and high-quality 48 kHz output.

Inflect-Nano, an ultra-extreme tiny 4.63 million parameter text-to-speech model, has been released.

A developer debunks the common belief that LLM latency is the primary cause of slow voice agents, explaining that delays often stem from earlier stages like audio capture, VAD, and STT. They recommend logging specific latency metrics and testing various STT/TTS providers and orchestration frameworks to diagnose issues.

A curated, open-source learning path for building voice agents, covering from STT to production, with 190+ resources and a 5-week plan.

Kokoro-82M is a highly natural text-to-speech model with 82 million parameters and over 11 million downloads, representing a significant advancement in AI voice generation.

Compares two locally running mobile TTS models, Kokoro and Supertonic, questioning their production quality beyond initial demos.

Demonstrates Hermes Agent using its Manim Video skill and TTS tool to create a video explaining itself.

Introducing the open-source project Pixelle-Video: a fully automated AI short video engine. Input a topic and it automatically generates a video with script, images, voiceover, and background music. Supports local and cloud models, modular design allows flexible replacement of each component model.

Zyphra released ZONOS2, an open-source MoE text-to-speech model trained on over 6 million hours of multilingual speech, supporting voice cloning and high-quality synthesis across many languages.

Xiaomi has released updates to its MiMo model series, including mimo-v2.5-asr (supporting multiple dialects and lyric transcription), mimo-v2.5-pro (trillion parameters, 1M context), mimo-v2.5 (full-modal perception), and a TTS series, significantly improving agent performance and recognition capability in complex acoustic scenarios.

A revamped TTS benchmark introduces objective standards and live blind voting to create an ELO rating for 46+ models, with participation open to the community.

Recommends 10 practical open-source GitHub projects, covering AI investment analysis, multi-model chat client, video generation engine, financial terminal, automated short video generation, AI email client, voice cloning, domain information collection, Claude skill sets, and API integration, etc.

RedNote releases dots.tts, a 2B parameter open-source text-to-speech model with zero-shot voice cloning and 48 kHz synthesis.

Dots.tts is a new TTS model from Xiaohongshu (RedNote) with 2B parameters, Apache 2.0 license, fully continuous architecture without codec tokens, 48kHz synthesis, and zero-shot voice cloning.