Prototype 01 · Podcast

Turn source material into a conversation worth listening to.

Upload material, tell the tool what you want to understand, and let it research across the sources before shaping a cited multi-voice discussion. The goal is not a generic summary. It is a useful conversation you can listen to and inspect.

What I was testing

There are plenty of tools that can summarise a document. This prototype asks a different question: can a set of sources become a conversation that is easier to absorb, still grounded in evidence, and useful enough to listen to away from a screen?

Source first

Can episodes stay tied to their material?

The source files are part of the episode identity, not hidden behind a generic prompt. You can see what an episode was built from and return to the evidence.

Conversation

Can synthesis sound less like a report?

Different speaker roles can extend, question and pressure-test the material rather than reading a sequence of polished summary paragraphs.

Listenability

Would you actually keep listening?

The quality bar is not whether audio was generated. It is whether the conversation has enough rhythm, challenge and clarity to be worth 20–45 minutes of attention.

Explore the workflow

Choose a demonstration episode below. The examples and source names are synthetic. The page uses no private data, makes no model calls and does not connect to the private Podcast database.

Demonstration mode Everything below runs in the browser. The short “play” sample uses your browser's built-in speech capability where available; it does not call a paid voice API.
Sample episode

Public demo · $0
Browser audio sample~45 sec excerpt · no external call

Conversation excerpt

The working application stays private.

The full workspace contains the persistent PDF library, real generated episodes, audio, usage data and paid generation controls. Its URL can be known, but the application itself is protected by authentication before private content or APIs are available.

Open private workspace ↗

The public demonstration is deliberately separated from the working application so a visitor cannot expose source files, trigger generation, consume API credits or reach private episode data.

Show the provenance.A useful episode should make it easy to understand not only what was asked, but what material the conversation was built from.