How the Podcast Agent works: research, script and audio
The Podcast Agent researches your topic, writes a multi-host script and voices it. Learn what it can create, how web research works and how to revise in chat.
The Podcast Agent is the chat that turns a prompt into a finished podcast episode: it researches the topic, writes a script for your hosts and generates the audio. For a step-by-step first run, follow Creating your first episode.
What you can create
The agent makes podcast episodes. A single prompt produces one episode, even when the topic covers several subtopics. To get a series, explicitly ask for more than one (for example, "make 3 episodes about X, Y, and Z"). Jellypod chains the episodes so each script builds on the one before it. See Generating multiple episodes for limits and the outline review option.
Start a conversation
- Choose Podcast Episode on the Studio start screen.
- Describe the episode in your prompt. Tips for better prompts covers what to include.
- Optionally attach sources such as URLs, files or YouTube links.
Sources are optional because the agent researches on its own. If you paste 2,000 or more characters into the start screen input, Jellypod converts the text into a source automatically (see Supported source types).
From prompt to audio
The agent researches the topic, reads any sources you attached, then writes a full multi-host script and generates the audio in one pass. There is no approval step in between. The one exception is planning a series from an outline, where you review the episode plan before generation starts.
Credits are charged for narration audio, not for research. See Understanding credits for rates.
Web research
The agent can search the web and read pages whenever fresh information would improve the episode. Web research is always available, and there is no setting to turn it on or off.
- It reads public web pages only. To use paywalled material, copy the text and attach it as a source.
- Web research does not use credits.
- To keep an episode to your own material, say so in the prompt, for example "use only the attached sources". You can combine your sources with web research, and the agent draws on both.
Revise in chat
After the episode generates, ask for changes in the chat, such as "make the intro shorter" or "add a section about X". The agent edits the existing script rather than starting over. When a change needs fresh information, the chat shows Web Search and Read URL activity, and the agent cites the sources it used with their URLs.
Was this page helpful?