... days ago, we released our Qwen3-TTS-Streaming-ONNX repository. Today, it has already passed 750+ downloads on Hugging Face , and people are still downloading it every day. Honestly, it s exciting to see growing interest around the real-time voice AI technology we re building at VertoX . Especially because this is still an early version, and we have many more updates coming. Right now, a lot of work is happening behind the scenes around streaming speech translation, low-latency ...

... things I enjoy most while building VertoX is seeing the technology improve week by week. Patrick Lumban Tobing , our AI/ML Engineer, recently integrated Qwen3-TTS Streaming into our Streaming Speech Translation framework, and honestly, it s exciting to watch everything slowly come together. Lower latency. More natural voice output. Better real-time multilingual communication. There s still a lot to improve, but seeing the system evolve like this is a great feeling. It also makes me appreciate how powerful ... ... better week by week progress, not just product hype. Lower latency + more natural TTS is exactly what makes real-time translation start feeling usable instead of experimental Comment from Yaoshen Luo(@lyshen): What is your latency of your requests of Qwen3

... general tasks, 250 prompts total. All finetuned models were trained with 500 examples, 3 epochs, LoRA rank 16. I used Perplexity to create prompts and judge independently. The observations below are based on Perplexity's evaluation. Customer Support Classification (Qwen3-4B) : The improvement compared to base model was small, but definite. Edge cases where the base model confused "Account Access" with "Technical Issue", and feature requests it kept mislabelling. Finetuning fixed those. Invoice Extraction (Qwen3

... Mimic AI Desktop Assistant A privacy-first, AI-powered desktop assistant with voice synthesis, persistent memory, and local model inference. Mimic AI runs entirely on your local machine using Ollama for language models and supports Browser TTS and Qwen3-TTS for voice synthesis. Table of Contents Features System Requirements Installation Quick Start Architecture Overview Data Flow Component Interaction Configuration Usage Guide Troubleshooting Privacy & Security Features Core Capabilities Local AI Inference : Uses Ollama for running LLMs locally (Llama, Mistral ... ... Voice Synthesis : Browser TTS (no setup) and Qwen3