Whisper

Evidence (4)

• 7Hacker News·7/19/2026open_source
  transcribe.cpp launches as a ggml-based speech-to-text library

  transcribe.cpp is a new v0.1.0 transcription library built on ggml, aiming to support a wide range of modern speech-to-text models with GPU acceleration. The author says each model is numerically validated and WER-tested against reference implementations, and the project includes bindings for Python, JavaScript/TypeScript, Rust, and Swift/ObjC.

• 8Hacker News·7/13/2026research
  Apple SpeechAnalyzer beats Whisper in on-device speech recognition benchmark

  A benchmark comparing Apple’s new SpeechAnalyzer API with Whisper models found SpeechAnalyzer to be the most accurate on-device speech engine tested for English transcription. It also outperformed the legacy SFSpeechRecognizer by a wide margin, while running faster than Whisper Small on Apple hardware. The report argues that developers using SFSpeechRecognizer should migrate, and that SpeechAnalyzer is now the strongest Apple-platform on-device option for English speech recognition.

• 5Hacker News·7/7/2026product_launch
  Local CPU-Only TTS with Kokoro

  The article describes running Kokoro, a small text-to-speech model, entirely on a local machine using the CPU while keeping the GPU free for other workloads. It highlights a containerized setup with Kokoro-FastAPI, OpenAI Speech API compatibility, and reported synthesis times across several CPUs.

• 7Hugging Face Blog·5/6/2026framework_update
  Open ASR Leaderboard Adds Private Anti-Gaming Datasets

  Hugging Face updated the Open ASR Leaderboard with new high-quality English speech recognition datasets from Appen Inc. and DataoceanAI, covering scripted and conversational speech across multiple accents. The datasets will remain private to reduce benchmark gaming and test-set contamination, while the public leaderboard average still uses only public datasets by default, with an optional toggle to include the private sets.