With genAI, it has become much easier and quicker to automatically transform written content into audio content. It’s exceptionally efficient for diversifying the way your content can be consumed, making it more accessible for your audience.

However, whilst that may be true, whether your content will actually be listened to depends on a lot more: from the resonance it has with your target audience, to how crisp it sounds… and whether you are truly delivering it in the right format.

More often than not, simply changing the format from text to audio fails to increase engagement. That’s where alternative content types, crystal-clear audio and great concepts usually win the race - and AI can support you to deliver those, too.

How? Let’s explore.

Text-to-audio: transforming written text into spoken word

As you can imagine, repurposing existing content with little effort is a very attractive idea. With text-to-audio, you can take content that has already been written (i.e. already paid for) and transform it into a new format (audio) and make it available to a broader audience, through different channels and with different commercial mechanisms. As AI audio technology matures, this is becoming increasingly easy to do.

However, in order to produce audio that your audience will actually listen to from start to finish (rather than MP3s that will just ‘sit there’), two other things have to happen: 

1. The audio is of a high-enough quality that the listener will remain engaged

2. You have a great idea for content that your listeners want to listen to 

With regards to making engaging text-to-audio, there have been 4 approaches we have seen emerge in the last few years, each with their own strengths and weaknesses.

Let’s take a look at each

1. Article readers: Great for accessibility, bad for engagement

The easiest approach to repurpose content is to take an existing content piece (e.g. an article on a website) and then use a Text-to-Speech (TTS) voice to read it from start to finish. 

Nowadays, there are fantastic, lifelike voices out there that can read text with human-like delivery, and these are exceptionally useful for making articles more accessible to people with difficulties related to sight or reading.

In fact, you can listen to this article with one of these right now:

However, based on our experience in numerous text-to-speech projects, the content that is produced in this way almost always disappoints. Why? 

In the same way that Instagram content largely doesn’t do well on YouTube, and TV shows aren’t adapted for movie theaters, an article is usually made for reading. 

Reading articles out loud is not something listeners often enjoy, and as a result, they usually switch off and eventually read it when they have the time.

In short, “article readers” are great for accessibility, but very bad at keeping listeners engaged by themselves.

2. Summaries and aggregation: great… if the format is right

If AI that simply reads back written audio content isn’t inherently engaging, then what is? 

The key takeaway is that audio content tends to perform much better if it is meaningfully adapted to create a worthwhile listener experience.

That doesn’t mean it takes a lot more work - simply a compelling way of repackaging it.

For example, text to audio can help you produce a short summary of the written content (think a “60 second summary”, or an aggregation of several pieces of content into a digestible “weekly summary” of dense information like financial news or world events).

It takes some experimentation to get it right, and listeners will appreciate a bit of added “sound design”: a jingle at the beginning, some backing music or auditory queues to make it easier to follow go a long way in this respect.

With these kind of formats, we have seen fantastic results in listener engagement by synthesizing already-existing content into these engaging, short-form formats.

Try listening to this article in that format right now:

3. Short-form podcasts: the more factual, the better

Taking the above concept even further, we have seen successful projects that managed to repurpose their written content into unique podcast formats, such as ”5 minute daily news updates” or “10 things to know about X” (be it celebrity news or the stock market).

This is one of those:

The potential of AI audio production here is the possibility to create new formats of content that were not viable to produce before, be it because they were too niche or too time-critical and short-lived to produce. 

With AI-driven text-to-audio, it’s possible to (almost) fully automate the creation of such a short-form podcast, including all production elements like music and sound design.

So far, we at AudioStack have seen that the more “factual” the content is (i.e. news, financial updates, sport odds or scores) the better the engagement. 

It’s well worth evaluating your own content output to see if you might benefit from using AI to transform your written content into a podcast format, as the engagement with this kind of content is overwhelmingly positive.

4. AI-generated conversations: incredible potential (but the jury is still out)

In just the last few months, another new option has emerged: thanks to AI models such as NotebookLM, it is now possible to transform and present written content into a realistic podcast-style conversation between two people.

This is what one of those conversations about this very article sounds like:

The results sound impressive and are extremely easy to listen to. Of course, we are very excited about these new possibilities for content creation and synthesis, but we have not seen enough testing as of yet to come to a conclusive verdict about the performance of such content.

(In this respect, testers are always wanted.)

If you are thinking about transforming your written content into compelling audio, you can do so much more than simple, functional screen readers. Audio can breathe new life into content and unlock possibilities that broaden and engage your audience in new ways.

Why not get in touch with AudioStack to find out how?