The Izwi AI platform has released its latest update, v0.1.0-alpha-12, boasting significant advancements in automatic speech recognition (ASR) and text-to-speech (TTS) technologies.

This update introduces long-form ASR with automatic chunking and overlap stitching, resulting in faster and more efficient speech recognition. Additionally, it features faster ASR streaming and reduced unnecessary transcoding on uploads, streamlining the overall process.

Notable enhancements include support for MLX Parakeet, new 4-bit model variants for Parakeet, LFM2.5, Qwen3 chat, and forced aligner. TTS improvements include model-aware output limits and adaptive timeouts, further refining the platform’s capabilities.

The update also brings a revamped model-management UI, including the My Models and Route Model modal, designed to enhance user experience. For detailed information on the update and its features, users can visit the Izwi AI documentation.

As the Izwi platform continues to evolve, users are invited to test the updated features and provide feedback on their speed and quality, contributing to the platform’s ongoing development.

Photo by Will Chen on Pexels
Photos provided by Pexels