AI-powered speech recognition models are making strides, with new implementations emerging for both native and browser-based applications. Simultaneously, a London-based startup is leveraging AI to revolutionize energy transactions, securing significant funding for expansion. These developments highlight the rapid advancements and diverse applications of artificial intelligence across various sectors.
A pure Rust implementation of Mistral's Voxtral Mini 4B Realtime model, called "voxtral-mini-realtime-rs," was released on GitHub, allowing for streaming speech recognition natively and in the browser. This implementation, using the Burn ML framework, enables the Q4 GGUF quantized path (2.5 GB) to run entirely client-side in a browser tab via WASM WebGPU. Users can try it live, according to Hacker News (Source 1). The project offers a quick start guide for native CLI use, including downloading model weights and transcribing audio files.
Another development, a pure C implementation of the Mistral AI's Voxtral Realtime 4B model, was also made available on GitHub (Source 2). This implementation has zero external dependencies beyond the C standard library and includes MPS inference. Audio processing utilizes a chunked encoder with overlapping windows, managing memory usage regardless of input length. The C implementation also allows for audio input from stdin or live microphone capture, making it easy to transcode and transcribe various formats. A streaming C API (voxstreamt) is included, allowing for incremental audio input and token string output.
In the energy sector, London-based startup Tem secured a $75 million Series B funding round, valuing the company at over $300 million, according to TechCrunch (Source 5). Tem uses AI to optimize energy transactions and currently serves over 2,600 UK businesses, offering potential energy bill savings. The company plans to expand to the US and Australia, starting with Texas, with the ultimate goal of going public.
These developments come amid a broader landscape of AI advancements. Other news includes Discord's global age verification rollout, the release of entertainment trailers, and the use of 3D-printed whistles (Source 4). These examples demonstrate the wide-ranging impact of AI and related technologies.
Discussion
AI Experts & Community
Be the first to comment