How a Vintage Phone Became a Cloud Voice Agent (Full Transcript)

A DIY build using an ESP32 and AssemblyAI’s Voice Agent API to stream audio over WebSocket, no on-device AI required, plus a live demo and open-source code.
Download Transcript (DOCX)
Speakers
add Add new speaker

[00:00:00] Speaker 1: This is how I built a voice agent into this vintage phone powered by Assembly AI's Voice Agent API. So the first thing I did was I opened up the vintage phone itself. I got it off of Facebook Marketplace and didn't know what to expect. I also picked up this ESP32 kit. It's like a tiny $10 computer. And here's the trick, it doesn't need to run any AI at all. Assembly AI's Voice Agent API is just one web socket. So audio goes in and the agent's voice comes back. So the next order of business was to figure out which of these wires carry audio. So it turns out red and green are the mic and yellow and black are the earpiece. So that's what will wire into the ESP32. So I'm labeling each wire as I free it because once I start soldering I don't want to mix these up. This was my very first time soldering by the way and honestly it was way less scary than I thought. So wires out and the labels are on and we have everything connected to the ESP32. We plug in USB-C and we should be good to go. Cloud Code Fable had my back all the way. All of the code was written by agents and I've open-sourced the code for you in the GitHub link in the description.

[00:01:10] Speaker 2: All right so here it is. Let's give it a try. Yeah, what's the weather normally like in New York in July?

[00:01:27] Speaker 1: It is usually a bit of a mixed bag. You might get sunny skies or a sun drizzle, especially if you are visiting in the spring. Are you planning a trip to the Big Apple?

[00:01:38] Speaker 2: Sure am, see ya, bye.

[00:01:41] Speaker 1: Yay. I had way too much fun with this and the voice agent part was genuinely the easiest step. So over to you. What do you think this phone should do next? Let me know in the comments. Bye.

ai AI Insights
Arow Summary
A maker explains how they converted a vintage phone into a cloud-powered voice agent using AssemblyAI’s Voice Agent API and an ESP32. They opened the phone, identified and labeled the mic and earpiece wires (red/green for mic, yellow/black for earpiece), soldered connections to the ESP32, and streamed audio via a single WebSocket so the device runs no on-device AI. They demo the phone answering a question and note the project code is open-sourced on GitHub, asking for ideas on what the phone should do next.
Arow Title
Turning a Vintage Phone into a Cloud Voice Agent
Arow Keywords
vintage phone Remove
voice agent Remove
AssemblyAI Remove
Voice Agent API Remove
ESP32 Remove
WebSocket Remove
audio wiring Remove
soldering Remove
open source Remove
GitHub Remove
hardware hack Remove
maker project Remove
Arow Key Takeaways
  • You can build a voice-enabled device without running AI on-device by streaming audio to a cloud voice agent over WebSocket.
  • An ESP32 is a low-cost microcontroller suitable for prototyping voice hardware interfaces.
  • Carefully identifying and labeling mic/earpiece wires is critical before soldering.
  • First-time soldering can be approachable with careful preparation.
  • Open-sourcing the code lowers the barrier for others to replicate and extend the project.
Arow Sentiments
Positive: Upbeat, playful tone with excitement about the build (“way too much fun”) and reassurance about first-time soldering being less scary than expected.
Arow Enter your query
{{ secondsToHumanTime(time) }}
Back
Forward
{{ Math.round(speed * 100) / 100 }}x
{{ secondsToHumanTime(duration) }}
close
New speaker
Add speaker
close
Edit speaker
Save changes
close
Share Transcript