Open source · Self-hosted · Podcasts. Your way.

Any topic. Your podcast.

Start with an idea, article, pasted text, or forwarded message. Herald can research, write, narrate, and deliver a finished podcast through Telegram on infrastructure you control.

Topic-to-podcast Source-based episodes Zero-AI reading Telegram remote control
Herald logo
Supported today

Start with whatever you have.

Herald is not limited to converting an article. It can begin with a subject you want to learn about or source material you already have.

Topic

An idea or question

Give Herald a topic, headline, question, or short concept and let Topic mode research and build the episode.

URL

An article or web page

Send a public URL and choose whether Herald should stay source-bounded or add outside context.

Text

Pasted content

Paste notes, an article, a newsletter, or other text directly into Telegram and turn it into audio.

Forward

A Telegram message

Forward a message to your Herald bot and use that content as the starting point for the podcast.

Simple workflow

Send it. Shape it. Listen.

1. Send a topic or source to your Telegram bot.
2. Choose the mode, target length, voice, and speed.
3. Listen when Herald returns the finished MP3.
Generation modes

Choose what kind of podcast you want.

The same input can be handled very differently depending on your goal.

Topic

Research and write

Build a research-backed episode from a subject, question, headline, or brief prompt.

Source

Stay with the source

Create a structured podcast using only information from the material you supplied.

Expanded

Add outside context

Use your source as the foundation and augment it with bounded web research.

Literal

Just read it

Clean and narrate the supplied text locally without making any LLM API calls.

Open by design

Your server. Your models. Your workflow.

Herald keeps the pieces interchangeable instead of locking the product to one AI service or hosted platform.

Open source

Herald is released under the MIT License. Inspect it, deploy it, change it, or contribute to it.

Provider-neutral AI

Use supported hosted providers or local Ollama models, with configurable provider slots and failover.

Local narration

Kokoro text-to-speech runs locally while Herald handles pacing, assembly, and final audio delivery.

Go deeper

Want to see how Herald is put together?

The About page covers current inputs, generation modes, AI providers, Telegram controls, local TTS, architecture, deployment, and the open-source design in more detail.