Field notes on speech AI, on-device ML, and evaluation.https://rodwoods.dev/https://rodwoods.dev/posts/round-trip-cer/https://rodwoods.dev/posts/round-trip-cer/The quality gate I use for text-to-speech when no human can proofread the audio: synthesize, transcribe, score, gate.Tue, 14 Jul 2026 00:00:00 GMThttps://rodwoods.dev/posts/the-judge-was-stale/https://rodwoods.dev/posts/the-judge-was-stale/Every quality score in my TTS pipeline comes from a model. Models age. Mine aged silently and nearly flipped a real decision.Tue, 14 Jul 2026 00:00:00 GMT