Tag: #tts36 results
Kyutai has released Pocket TTS, a 100M parameter model capable of zero-shot voice cloning on CPUs using only 5 seconds of audio. It outperforms competitors in flexibility by allowing user-supplied voices without fine-tuning.
A new native C++ inference framework built on ggml that consolidates 12 audio models into a single runtime. It significantly outperforms Python-based inference, achieving up to 5x speedups on CUDA.
NetEase Youdao has upgraded its Ziyue LLM to version 4.0, introducing full-modal capabilities. The company has officially open-sourced its core multimodal and text-to-speech (TTS) models.
Curated list ranks top open-source models across audio generation (TTS, cloning, music), image generation, and image-to-video categories. Standouts include Qwen3-TTS for TTS, FLUX.1 for images, and LTX-2.3 for 4K video with audio. Provides links to Hugging Face repos for quick access.
OpenClaw 2026.4.15 updates default Anthropic models to Claude Opus 4.7 with image understanding and adds Gemini TTS support in the Google plugin. New UI features include a Model Auth status card, while memory enhancements bring cloud storage to LanceDB and GitHub Copilot embeddings. Fixes improve tool security and agent reliability.
DeepMind unveils Gemini 3.1 Flash TTS, the next generation of expressive AI speech synthesis. The model introduces granular audio tags for precise control over AI-generated audio. This enables highly customizable and natural-sounding speech output.
Fine-tuned Chatterbox-Multilingual TTS with LoRA adapters to support 8 Indian languages using only 1.4% parameters, no phoneme engineering. Extended tokenizer and used Brahmic warm-start for better init. Model and samples on Hugging Face.
Deepgram's production speech-to-text (STT) and text-to-speech (TTS) models are now available natively on Together AI. This enables access via Dedicated Model Inference for building real-time voice agents.
Mistral's Voxtral TTS model enables 'Day 0' integration with 90ms latency on Apple M4, capturing emotional nuance without cloud cold starts. Used for art-bot with personality preservation. GitHub repo for local API implementation.