Tag: #tts36 results
Amazon Polly launches Bidirectional Streaming API for real-time TTS in conversational AI. It allows simultaneous text input and audio output, ideal for incremental LLM responses. This enables low-latency speech synthesis before full text is ready.
Mistral has released a new open-source model for speech generation. The lightweight model can run efficiently on resource-limited devices like smartwatches and smartphones.
KoboldCpp celebrates its 3-year anniversary with version 1.110, introducing native high-quality Qwen3 TTS (0.6/1.7B) supporting voice cloning and Ace Step 1.5 for music generation. A demo video showcases these features, including an epic music piece. Download available on GitHub.
iFlytek has published a new patent for controlling accents and speaking styles in human-computer interaction systems. The technology enables dynamic state inheritance and context-aware voice synthesis.
A comprehensive CPU-based benchmark comparing small-scale TTS models using UTMOS MOS scoring. The analysis highlights the unique streaming architecture of Kyutai's Pocket TTS and the performance trade-offs between different inference backends.
NagaTranslate is an initiative to develop translation and speech pipelines for Nagaland's low-resource languages like Nagamese, Ao, and Sema. The project currently utilizes a hybrid approach of commercial LLM APIs, fine-tuned VITS for TTS, and Whisper for ASR.
Inflect-Nano is a highly efficient text-to-speech model with only 4.63M parameters, designed for local execution on low-end hardware. It offers a unique size-to-functionality ratio, making it suitable for embedded devices and offline voice agents.
Resemble AI has released DramaBox, a new voice model built on the LTX 2.3 architecture. It is designed to deliver the most expressive voice synthesis currently available in the open-source ecosystem.
.png)
Together AI launches Voice Finder, a new tool enabling developers to search, match, filter, and audition over 600 voices across their TTS models. It supports natural-language prompts or uploaded audio samples for precise voice selection. This streamlines integrating the right voice into apps.
OpenClaw 2026.4.25-beta.4 delivers comprehensive TTS upgrades with new providers like Azure Speech and ElevenLabs v3, chat-scoped controls, and per-agent overrides. Plugin management shifts to a persisted registry for efficient updates and discovery. Expansions in OpenTelemetry, browser automation, UI features, and install hardening enhance reliability.