Topic
Voice
Voice in and voice out: the spoken half of an agent, end to end.
18 posts
- An in-app agent is an accessibility feature, narrowlyAn agent that completes a task is an alternative input path, and one WCAG 2.2 criterion turns the biometric confirm into an accessibility win rather than a tax.Voice10 min
- The first 300 milliseconds of an agent turnYou cannot make the model faster from the client. Almost all of the felt wait is decided before the model has said anything, and that part is yours.Render spec10 min
- Running ten dialects of one language through one pipelineMost speech stacks assume one user speaks one language. Production breaks that in week one, and the fixes are the same whichever language you start from.Language10 min
- What barge-in really costs, and how we pay for itServer-side turn detection decides when an agent should stop talking. The client decides whether it actually does, and that is where it breaks.Voice11 min
- Agent runtimes for mobile apps: an honest comparison (2026)OpenAI Realtime, LiveKit, ElevenLabs, Vapi and Retell compared on what a mobile team actually chooses between: layer coverage, client SDKs and the write path.Business8 min
- Designing an agent conversation people can actually useThe seven decisions that make or break a conversational feature: discoverability, endpointing, barge-in, errors, confirmation, latency and the visual answer.Voice11 min
- How to evaluate dialect coverage in a speech stackA vendor language list is a claim about a corpus, not about your users. The method for measuring what a speech model does on the varieties they actually speak.Language9 min
- How fast does a voice agent have to be? A latency playbookWhere a spoken turn's milliseconds go, what the turn-taking research actually measured, and which tactics shorten which segment of the budget.Voice9 min
- Adding an AI agent to a React Native appThe real integration: five calls across the bridge, what your JavaScript owns, what the native side draws, and the week-two work nobody puts in a quickstart.Integration9 min
- Agents in a logistics app: the driver's hands are the constraintHands-free is a legal requirement in a cab, not a convenience. What a logistics agent should do, how proof of delivery works spoken, and the offline case.Business8 min
- Text-to-speech for agents when the language is hardSynthesis quality is decided upstream of the model that makes the sound. The two stages that break on a hard language, with Arabic as the worked example.Language9 min
- Why some varieties of a language get recognised and others do notEgyptian Arabic is the best-served spoken variety of Arabic, and the reasons are historical rather than linguistic. What that predicts for every other language.Language8 min
- Voice is an alternative input path, not a speed featureThe case for a spoken path into your app, argued from the curb-cut effect, WCAG 2.2 and what screen reader users actually report.Voice7 min
- Voice or chat: picking the mode for the taskSpeaking and typing are two inputs to the same agent. The situation the user is in decides which one wins, and four of those situations are predictable.Agent basics9 min
- Shipping speech for a variety with less training dataHow to measure recognition coverage for a language variety the big corpora barely contain, using Gulf Arabic as the case where the gap is documented.Language10 min
- When your user changes language mid-sentenceCode-switching is the normal way bilingual people talk, and it breaks pipelines that pick one language per utterance. What fails, and how to test for it.Language9 min
- When voice actually works in an app, and when it doesn'tAn honest framework for deciding which tasks belong to a spoken path, which belong to text, and which should stay on the screen.Agent basics8 min
- Conversational interfaces and older users: the real tensionWhy asking for a task costs older users less than navigating to it, and why the same people are least likely to have tried an AI product.Voice8 min