Tag
The author built a local realtime voice stack using Parakeet STT, Qwen 2.5 7B, and Qwen3-TTS, integrated with Ollama.
A developer shares a postmortem of building a Hindi-English voice agent for fintech, highlighting challenges with number readback, code-mixed TTS, latency under load, and compliance. Key fix was choosing TTS with first-class support for Indian code-mixing and testing at real concurrency.
Voicebox is a locally run open-source AI voice studio that supports 7 TTS engines, 23 languages, and voice cloning. All processing is done locally to protect privacy. The project has received 33.8k stars on GitHub.
Audio8's open-source TTS model reached #1 on Hugging Face's TTS trending list and #10 overall, with the team expressing gratitude and plans to keep improving.
Gradium has released a new TTS model in public beta that accurately reads phone numbers, emails, IBANs, and time expressions natively, offering API access with 1M credits for testing.
A new toolkit enables fine-tuning the tiny Inflect-Nano/Micro TTS models on custom voice and language, supporting warm-start, resumption, and export to PyTorch/ONNX.
Release 0.4 of audio.cpp adds C++/GGML inference for Higgs Audio v3 TTS 4B (10x real-time) and Fish Audio S2 Pro, with full GGUF loading and Q8 speed/VRAM gains.
NeuTTS-2E is an open-source on-device TTS model that supports seven controllable emotions.
LangChain launches LangSmith tracing for voice frameworks (Pipecat, LiveKit, OpenAI Realtime, Gemini Live), enabling full audio monitoring, STT/TTS latency tracking, interruption detection, and VAD analysis with minimal code.
A curated list of text-to-speech resources for voice agent builders, organized around the decision between real-time streaming synthesis and offline high-fidelity synthesis, with emphasis on streaming latency and mid-stream cancellation.
A curated reference of open-source text-to-speech models organized by license type, highlighting which models can be used commercially.
NVIDIA released Nemotron, an audio-native model capable of transcription, translation, sound recognition, audio Q&A, TTS, and full speech-to-speech, with open weights in 2B and 30B sizes.
Scylla's Band is a new TTS model and inference framework, including an Android sample for deployment.
Release of C++/GGML based implementations of Supertonic 3, MOSS-TTS, IndexTTS2, and Irodori-TTS in audio.cpp, capable of generating 10 hours of audio in 3 minutes on an RTX 5090.
The Cicero project enables self-hosted, bidirectional voice conversation with AI agents, supporting local TTS/STT and integration with various agent protocols and tools like Claude Code and Hermes Agent.
Matt Pocock shares an idea for a workflow using LLMs to generate podcast summaries of code diffs, aiming to prevent comprehension debt in fast-moving repos; he has implemented it for his personal wiki with good results.
Whispera 是一个基于 VoxCPM 的 Windows 本地实时语音助手，集成了 SenseVoice ASR、llama-server 本地 LLM 推理、VoxCPM 流式 TTS 和 Mem0 长期记忆，完全离线运行。
Maya-2-Native has achieved the #2 position on Voice Arena's Hindi TTS leaderboard, surpassed only by Gemini 3.1 Flash.
A CPU TTS benchmark compares Kokoro, Supertonic, Inflect-Nano, and Kyutai's Pocket TTS using UTMOS MOS scores, revealing interesting findings about RTF scaling, UTMOS limitations with small vocoders, and undocumented output caps. Pocket TTS offers unique zero-shot voice cloning on CPU.
A guide demystifying voice agents using speech-to-text and text-to-speech, with four browser-based demo agents and instructions to build your own using RAG and tools.