# Cartesia Assembly AI Projects at AI Tinkerers

> Canonical HTML: https://aitinkerers.org/technologies/cartesia-assembly-ai
> Markdown URL: https://aitinkerers.org/technologies/cartesia-assembly-ai.md
> Technology record last updated: 2026-06-29T02:12:46Z
> Generated: 2026-09-23T09:39:11Z

A high-performance integration stack pairing AssemblyAI’s speech-to-text with Cartesia’s ultra-low-latency text-to-speech to build responsive, real-time voice agents.

Building production-grade voice agents requires shaving off every millisecond of latency, and pairing AssemblyAI with Cartesia is the industry-standard way to do it. This stack routes live audio through AssemblyAI’s Universal-3.5 Pro Streaming model for speech-to-text (clocking a tight 307ms P50 latency with built-in neural turn detection) and answers back using Cartesia’s Sonic 3.5 text-to-speech engine (which fires its first byte of audio in just 90ms). Developers tie these engines together using orchestration frameworks like LiveKit Agents or Pipecat. The result is a highly natural, bidirectional conversational pipeline that handles everything from automated customer support to ambient medical scribing without the typical lag of legacy voice systems.

- Official technology site: https://www.assemblyai.com
- Public AI Tinkerers demos and talks: 1
- Result page: 1 of 1

## Recent Public Talks and Demos

### [Coding on dog walks over the years](https://boston.aitinkerers.org/talks/rsvp_JsuBvLBOFkM)

I've been tinkering with coding by voice since ChatGPT came out. I'll demo my latest iteration Openbase, which to my knowledge is the most capable voice coding solution currently available (even though it is currently not the fastest). Openbase is designed for work on real production apps, and optimized for splitting work between a team of agents, each with their own names, voices, and tasks. Start Codex and Claude Code threads on your machine like normal, then pick them up from mobile with voice-to-voice. Steer, queue, abort, confirm dangerous actions, and plan mode all supported over voice. Other features of Openbase are under active development (please reach out if you'd like to team up!) - Syncs your filesystem and AI coding threads across devices (solves the problem of: "should I start this thread on my laptop or mac mini/remote devbox") - Optionally bundles Tailscale for easy single-account installation - (Codex only ATM): Start a long-running turn, and steer it from voice, even if it hasn't finished running yet. Claude Code you can resume but only after a turn has completed. Full docs at https://docs.openbase.cloud/

- Event context: Back from Summer: AI GTM Builders — 2026-09-03 — Boston
- Public talk page: https://boston.aitinkerers.org/talks/rsvp_JsuBvLBOFkM

## Related Technologies

- [Apple Callkit](https://aitinkerers.org/technologies/apple-callkit) ([Markdown](https://aitinkerers.org/technologies/apple-callkit.md)) — 1 public demo
- [Claude Code](https://aitinkerers.org/technologies/claude-code) ([Markdown](https://aitinkerers.org/technologies/claude-code.md)) — 216 public demos
- [LiveKit](https://aitinkerers.org/technologies/livekit) ([Markdown](https://aitinkerers.org/technologies/livekit.md)) — 12 public demos
- [Tailscale](https://aitinkerers.org/technologies/tailscale) ([Markdown](https://aitinkerers.org/technologies/tailscale.md)) — 4 public demos