ASR From Scratch
- Part 0 Prologue: The Questions 2026-07-19
TTS From Scratch
- Part 0 Prologue: The second set of questions 2026-08-09
- Part 1 Reading is not writing: the asymmetry at the heart of TTS 2026-08-10
- Part 2 What should a voice be made of? Two jobs, two kinds of token 2026-08-11
- Part 3 The wall: why you can't just predict the sound 2026-08-12
- Part 4 Say it, then voice it: splitting one hard job into two 2026-08-13
- Part 5 Coherent residuals: from independent heads to a depth transformer 2026-08-14
- Part 6 Adapt the consumer to the producer, again 2026-08-15
- Part 7 Where it fits: a map, and the voice it was aimed at 2026-08-16
Other writing
- 2026-08-09 Three things that fooled me building a codebook TTS
- 2026-08-08 Hand off the vector, not the id
- 2026-08-07 One codebook, both directions: a frozen alphabet as a contract
- 2026-08-06 Watching attention write speech: the cluster diagonal, in reverse
- 2026-07-26 Why Attention Isn't the Problem: Understanding Classification Errors in Attention-Based Encoder-Decoder (AED) Models
- 2026-07-19 Watching attention read speech: cluster-attention maps from a units-to-text transformer
- 2026-07-16 Beyond WER and CER: Why Indian-Language ASR Needs Akshara-Level, Phonetically-Weighted Metrics
- 2026-07-08 Speech recognition as independent parts: notes from a frozen-codebook ASR experiment
- 2026-07-08 Hello World
- 2026-07-08 What I'm trying to learn from my speech recognition experiments