Summarization (beta)

Learn how to generate and distribute AI-generated audio summaries with BeyondWords

Introduction

In today's fast-paced world, it can be challenging to stay on top of the latest news and information. AI summarization lets you convert your articles into audio and video summaries, making it easier than ever for your readers and subscribers to keep pace with the news.

Enable summaries

To get started, go to the Summarization tab in your project settings, activate the Summarization content enabled toggle, and then select a default voice and voice speaking rate for your summaries.

Once enabled, you will be able to submit your own summaries through our API or enable AI summarization so we can summarize the articles for you based on a default prompt (see next step).

Enable AI summarization

Activate the AI summarization enabled toggle to let BeyondWords summarize your articles using the default prompt.

For the beta release, we're using gpt-4 from OpenAI to generate summaries. In the future, we may explore letting customers select their preferred LLM and model.

Edit the default summarization prompt

For the beta release, the default prompt is in English. We strongly recommend re-writing the prompt in the same language as your content. Failing to do so may result in your summaries being generated in English, even if your original content is in a different language.

We recommend customizing the default prompt to your specific summarization needs. Tailoring the prompt can significantly improve the quality and relevance of the generated summaries.

Feel free to experiment with the wording, length, and instructions to optimize the output for your use case, content, and desired summary characteristics.

Once satisfied, press Save changes (you can always change it later and regenerate summaries).

Generate summaries

With Summarization and AI summarization enabled, each time you generate an audio and video (if enabled) for an article, an audio and video summary will be generated too.

Go to the Content section in your project to listen to the summaries:

Edit summaries

You can also edit summaries on a per article basis (great for prompt experimentation). Go to an article and click Edit summary:

You will then be able to edit the summarized text through the TTS editor or Edit AI summarization to regenerate the text with a different prompt. Once you're happy with the summary, click Regenerate to regenerate the audio and video (if enabled).

Distribute summaries

Like audio and video articles, summaries can be distributed using the BeyondWords Player. You just need to add the loadContentAs parameter to the player script and set it to ['summary']. See example:

<script async defer src="https://proxy.beyondwords.io/npm/@beyondwords/player@latest/dist/umd.js" 
  onload="new BeyondWords.Player({ 
    target: this, 
    projectId: <ID>,
    contentId: '<ID>',
    playerStyle: 'standard',
    loadContentAs: ['summary']
  })">
</script>

To make both the full article audio and summary available in the BeyondWords Player SDK, you can do the following:

<script async defer src="https://proxy.beyondwords.io/npm/@beyondwords/player@latest/dist/umd.js" 
  onload="new BeyondWords.Player({ 
    target: this, 
    projectId: <ID>,
    contentId: '<ID>',
    playerStyle: 'standard',
    loadContentAs: ['summary', 'article']
  })">
</script>

For the beta release, the default player UI does not support both summaries and articles. Therefore, if you want to make both playable through a player, we recommend setting showUserInterface: false and building your own player UI with our player SDK.

Fetching summaries through the API

You can fetch the summary text through the Content or Player resources in the API.