AI Hacker News Jul 14, 2026 1 min read
A benchmark that Apple did not publish itself drew HN attention: SpeechAnalyzer posted lower word error rates than both the old SFSpeechRecognizer and Whisper Small on the same Apple hardware.
A benchmark that Apple did not publish itself drew HN attention: SpeechAnalyzer posted lower word error rates than both the old SFSpeechRecognizer and Whisper Small on the same Apple hardware.
A LocalLLaMA post details recurring Whisper hallucinations during silence and proposes a layered mitigation stack including Silero VAD gating, prompt-history reset, and exact-string blocking.