FluidVoice is a new macOS dictation app with on-device speech-to-text and custom AI enhancement, positioned as a local Wispr Flow alternative.
The Alaya-EVOKE paper discusses moving from linear-scaling supervision to endless world generation. It has gained 64 upvotes on HuggingFace.
MiniMax has released its Music3 model for music generation, according to community reports.
A Show HN release presents Reverb ASR+Diarization as the best open-source ASR solution for long-form audio. The claim comes from the developer.
Lightricks has released the official Python inference and LoRA trainer package for the LTX-2 audio-video generative model, gaining 205 stars.
Index-TTS is an open-source system for controllable, efficient, zero-shot text-to-speech, gaining 105 stars today.
Macro is a newly trending unified workspace that integrates email, chat, docs, tasks, agents, and CRM with shared AI memory. It gained 1,239 stars today.
DeepSeek's OCR codebase has been updated, showing active development on a high-performing OCR tool with 23,786 stars.
A new research paper introduces AutoDesign, a method for optimizing meta-harnesses to improve long-horizon agentic design. The paper received 28 upvotes on HuggingFace.
RAGFlow, an open-source RAG engine that fuses RAG with Agent capabilities, gained 465 stars today, cementing its leading position.
NVIDIA has released Switchyard, a tool that allows LLM applications to route traffic across models and providers while preserving OpenAI and Anthropic API compatibility. It enables flexible model selection and cost/performance optimization.
GLM 5.3 has been released, with community reports highlighting its strong capacity-to-size ratio. The release includes open weights.