How to use Claude to build a formant speech synthesizer
An experienced developer used Claude Code to translate phonetics research into a Klatt-style synthesis runtime, retaining ownership of specifications, architecture and corrections to domain errors.
news.ycombinator.com
Prompt to try
Editorial adaptation, not the source author’s exact prompt. Replace bracketed inputs and review tool permissions first.
Help implement [formant speech synthesizer] from [phonetics papers] and [behavioral compatibility specifications]. Use declarative YAML and CEL expressions for phased rules, cite each parameter to its research source and present coefficient interpretations and formant transitions for my domain review.
Reach people exploring AI agents
See placements →What happened
The author reports a working runtime developed over months with Claude Code; the stated commercial-synthesizer compatibility target remains an ongoing effort.
Reported by ctoth on Hacker NewsHow the agent did it
- Read phonetics papers and wrote the synthesis specifications.
- Chose an architecture using YAML and CEL for phased rule execution.
- Used Claude Code to translate research into a working runtime.
Where a person stepped in
The author chose the architecture and specifications and personally catches errors using phonetics knowledge and expected speech output.
A person approved each step before the agent continued.
The author is building a Klatt-style formant speech synthesizer aimed at behavioral compatibility with a specific commercial synthesizer. This is a specialist project rather than an off-the-shelf assembly: they describe the architecture as driven by hundreds of academic papers and every parameter as backed by citations to phonetics research.
The author wrote the specifications, chose the architecture and read the papers. The runtime uses declarative YAML with CEL expressions for phased rule execution. Claude Code helped translate the large body of acoustic phonetics research into the working runtime. The author, a professional developer for more than 15 years, says they could not have built this project without AI coding tools.
They contrast the months spent with Claude Code against the years they believe solo implementation would have required. That comparison is their assessment, rather than a measured speedup. Domain knowledge remained central: the author says they still catch Claude misinterpreting a Klatt coefficient or mishandling a formant transition rule, because they know what the output should sound like.
The source links the Qlatt repository but describes development as ongoing. It does not claim completed compatibility or report a formal evaluation result.
Who it's for: Speech-synthesis researchers or experienced developers able to interpret phonetics papers and review the generated runtime against domain requirements.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · ctoth
Frequently asked questions
What did the human author contribute?
They wrote specifications, chose the architecture, read the research papers and caught misinterpreted Klatt coefficients or incorrect formant transition rules.
Does the source claim compatibility is complete?
No. The author says they are building toward behavioral compatibility with a particular commercial synthesizer. The post does not report completed compatibility or a formal evaluation result.
More details
Remaining steps
- Caught coefficient interpretation and formant transition errors.
About this account
Claude Code helped translate a large research corpus into runtime code over months of ongoing development.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=47434697. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: ctoth on Hacker News
Source published 19 Mar 2026
Source checked 14 Oct 2026
Tools and topics
Claude Code · YAML · CEL
Speech synthesis · Phonetics · Research-backed code
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.