AI agent podcast studio
Agents transcribe, edit, master, and package it. You approve. Nothing ships without you.
Powering Signal & Noise — the premier show by Nnamdi Okorafor.
Audio or video, any length. The studio takes it from raw file to publish-ready package.
Choose a file →Seven agents, one pipeline. Every decision is logged.
The pipeline stopped here on purpose. Everything is produced and QC-passed, but it will not publish until you say so. A hard QC failure blocks release outright — there is no override, and the agents cannot mark their own work as passing.
Loudness
−16.2
target −16 ±1
True peak
−1.7
ceiling −1.5
Format
44.1k
stereo
Runtime
40:10
2410.3 s
Approved by you, then distributed. Every artifact below came out of one drop.
During the interview
Your phone is the microphone. The server runs the model. Cues arrive while you're still talking.
Moderation
Guest is circling — steer back with something concrete. He offered to introduce someone; don't let that drop, lock the follow-up before you move on.
⚑ Unverified claim
NEEDS_SOURCE — “backed by the US Constitution.” No match in the fact pack. Ask him to define it rather than letting it stand.
Ask next
“What's the most expensive misconception people bring to you?”
Why it's different
Agents run the entire chain, then stop dead at review. QC hard-fail means no publish — no override flag, and no agent can pass its own work. Your name stays on the episode honestly.
Fact-checks resolve to verified or needs-source, never to an invented citation. On a show about money or health, an assistant that fabricates sources is worse than no assistant.
No timeline, no track stack, no settings to learn. One drop in, finished episode out. Loudness targets, chapter marks, captions and artwork are decisions the studio already made correctly.
Every stage is a tool over MCP and HTTP. The studio isn't an app you sit inside — it's a surface any agent can operate, which is why the phone can record while the server thinks.
The page above runs on demo data. Point it at a running PodMaster3000 API to produce real episodes.