Instagram has added a text-to-speech feature to Reels, allowing creators to convert video captions into audio.

This is seemingly in an effort to keep up with TikTok, which launched text-to-speech in December last year.

While popular (there are currently 1.5 billion views on the #texttospeech hashtag alone), TikTok's AI voices have received a lot of negative attention. And Instagram could go down the same route.

For starters, it's only offering two voices, both of which have an American accent: Voice 1 (a female voice) and Voice 2 (a male voice). It also experiences similar issues with inaccurate pronunciations and robotic-sounding speech.

These limitations are understandable in the context of these platforms. But audiences are increasingly exposed to advanced synthetic speech, so their expectations are high.

BeyondWords, for example, offers a library of over 720 voices across 64 languages, and the ability to create a custom voice. We also use natural language processing (NLP) and speech synthesis markup language (SSML) to ensure more accurate text-to-speech. And we've powered over one billion listens for over 120 global publishers.

So, I expect Instagram's text-to-speech feature to get its fair share of criticism. But it's an impressive feature that will no doubt assist creators and their audiences.

How to use the text-to-speech feature on Instagram

1. Open Instagram and go to the Reels camera

2. Create your video then select "Preview"

3. Tap "Aa" to add a text caption

4. Tap the text bubble, then "...", then 'Text-to-Speech"

5. Select a voice then tap "Done"

6. Make any other edits then share your Reel

With BeyondWords, you and your team can convert any text into quality audio. We also offer all the distribution, analytics, and monetization tools you need.

Please note that this post was published before SpeechKit rebranded to BeyondWords.

We've released v3 of the SpeechKit text-to-speech plugin for WordPress, and it's packed with improvements and new features that make it even easier to engage your audiences with audio. Keep reading or listening to learn what's new.

Existing user? If you have auto-updates switched on, you'll already have access to these new features. Otherwise, you'll need to update the plugin via your WordPress admin.

1. Create audio versions in bulk

You can now create and publish audio for multiple posts at once. This means you can easily make old posts listenable, or add audio to posts you'd previously excluded.

1. Open the relevant posts overview section in your WordPress admin ("Posts", "Pages", or a custom post type).

2. Use the checkboxes on the left to select which posts you'd like to request audio for (you may need to change your "View mode" in "Screen Options" first).

3. Select "Generate audio" from the "Bulk actions" dropdown, then "Apply".

The selected posts will now be processed into audio - this typically takes less than a couple of minutes per post. You can check the status of audios via the SpeechKit sidebar or your SpeechKit dashboard.

Once processed, your audios will be automatically embedded into the corresponding posts via the Small Player (like the one on this post).

2. Check audio responses at a glance

We've replaced the "SpeechKit Status" column in post overview pages with a "SpeechKit" column. This shows you the response from our API - in other words, whether:

audio has not been requested for the post (-),

audio has been requested for the post (),

the Player has been disabled on the post ( - Disabled), or

there has been an error (you will see a ! warning symbol with an error code and description).

You can check the status of a particular audio via the SpeechKit sidebar in your post, or via the "Content" section of your SpeechKit dashboard.

Tip: You can also use the SpeechKit sidebar to determine whether or not audio is requested and whether or not the Player is displayed.

3. Easily share debugging data

If you're having technical difficulties with a particular post, you can now send us the debugging data.

1. Open the SpeechKit sidebar on the relevant post.

2. Go to "Inspect" then select "Copy".

3. Paste the copied details into your message to our support team.

Tip: You can report voice issues via your SpeechKit dashboard. For any other issues, please email us or use the chat function in your dashboard.

Plus much more

We've also made a host of other updates to improve the speed, stability, and security of our WordPress plugin. For full details, take a look at our WordPress changelog.

Lots of text-to-speech service providers use AI voices from Amazon Web Services, Yandex, Microsoft Azure, and Google Cloud Platform.

But these voices sound best when you use them in BeyondWords.*

Why?

It's thanks to our natural language processing algorithms.

Setting voices up for success

AI voices can interpret text in two formats: plain text, or speech synthesis markup language (SSML).

SSML tags provide extra information to the AI voice, clarifying pronunciations and improving speech flow. Using SSML therefore ensures a higher-quality voice output.

Some text-to-speech services only allow you to input plain text, meaning you can't achieve higher-quality outputs through SSML.

Others, like Amazon Polly, give you the option to manually insert SSML tags. But this is complex and time-consuming. Let's say the voice is mispronouncing "Joe Biden". Fixing this requires an understanding not only of SSML, but of the international phonetic alphabet (IPA) - symbols that linguists use to represent speech sounds.

Joe Biden

This is not feasible for the majority of publishers.

BeyondWords, on the other hand, adds the SSML tags for you.

Whether it is imported from your website in HTML format or manually imported as plain text, your content is automatically converted into SSML before being processed by the AI voice.

This is made possible by a layer of natural language processing (NLP) algorithms, which can programmatically read, analyze, and interpret written language.

How our NLP works

Our NLP uses a combination of rule-based and neural network-based techniques. Our deep learning models are trained on large data sets, which allow them to "learn" how humans convert particular text elements into speech and how this differs depending on context. This is particularly useful for resolving ambiguities in text.

For example, in the sentence "I read the book", our NLP identifies, through contextual features, that the homograph "read" is most likely being used in the past tense, and so should be pronounced like "red" (/rɛd/) as opposed to "reed" (/riːd/). It applies the SSML tag accordingly, ensuring the AI voice gives the correct output.

I read the book yesterday.

Also consider non-standard elements, such as dates. Our system can determine whether to read a number as cardinal (e.g. twenty nine) or ordinal (e.g. twenty ninth) numeral based on the usage context and other features.

Without this NLP and SSML layer, the AI voice may not predict which pronunciation of an element is correct and simply output a naive "best guess". This is why many text-to-speech systems struggle with ambiguities.

A customizable and evolving NLP

Our team of computational linguists make iterative and domain-specific improvements to the NLP. This means that voice outputs evolve even when the AI voices behind them stay the same, and can adapt to the needs of each BeyondWords user.

Take the abbreviation "NLP", for instance. In the context of this article, it refers to "natural language processing". But it is also the airport code for Nelspruit, South Africa. In medicine, it can be shorthand for 'no light perception'. With custom-built text normalization rules, we can ensure that our system delivers the most relevant result for a particular publisher.

Text-to-speech providers without a post-optimization NLP layer have no way to efficiently extend or make domain-specific customizations to an existing voice. If they wish to correct conversion errors, they must retrain the voice itself - something that comes at great cost and cannot guarantee accuracy, especially when it comes to unusual and idiosyncratic text-to-speech conversions.

Here are some more examples of what our NLP can do:

Apply

SSML tags to ensure correct pronunciation of novel or complex words

Apply

SSML tags to ensure foreign words are pronounced accurately

Apply

SSML tags to ensure s...

Learn how to elevate your text-to-speech audio. Whether you're creating audio from plain text or editing automated audio, you can use these tricks to make your content more engaging.

1. Mix and match your favorite voices

You can change the voice associated with any title or paragraph in just two clicks. With a variety of virtual speakers at your fingertips, you can craft richer and more engaging audio using multiple voices.

This feature is ideal for quotations. Not only can you give speakers more befitting virtual voices, but you can provide clear audio cues as to where speech begins and ends. You could also try using a different voice for your titles and subheadings.

Plus, you can change your default voices in seconds. This means you can easily switch things up before creating your next audio. Our library includes 550+ AI voices across 130+ language locales. We also offer voice cloning, so you can create custom voices that perfectly suit your needs.

2. Insert your own audio clips

You can insert audio files into your text-to-speech audio without using an audio editor. Simply enter your text like usual then upload audio wherever you'd like it to play, and we'll stitch it seamlessly into the final audio file.

Combining synthetic spoken-word audio with other listenable content gives you the best of both worlds. Make your reports more trustworthy by inserting real interview snippets, provide clips of referenced audio to improve comprehension, or use sound effects and music to create atmosphere.

When creating longer audios, it's a good idea to insert short transitional sounds, known as bumpers and stingers, to improve flow.

3. Optimize your text

When converting written content into audio for narration purposes, you may want to ensure an exact match between the written and spoken words. Otherwise, you should optimize your text to provide the best possible listening experience:

Sense-check your content. For example, where your text version says "keep reading", your audio script should say "keep listening"

Use punctuation and paragraphs to improve flow. Punctuation and paragraph spaces determine where pauses should go in synthetic speech and how long they should be, so you may wish to tweak these for optimal flow

Add audio descriptions. Where your text version has visual assets, such as illustrations and charts, you may wish to provide audio descriptions instead

Add audio links. Where your text version has a hyperlink, you may wish to read out the URL or directions to the page instead

4. Create and collaborate at your own pace

You can save drafts in the Text-to-Speech Editor and continue working on them anytime. This is particularly handy when writing from scratch or working on edits to long-form text. You can even invite teammates to collaborate on your project, so you can work on the content together.

5. Review and update your audio

If you want to make your audio the best it can be, you should review it after processing. This is known as proof-listening. If you identify any errors or opportunities for improvement, you can address them before distributing your audio.

You can also edit your audio further down the line - without disrupting your existing distributions. This means that you can keep your content up to date and make incremental improvements.

Create your best audio yet

Our platform has everything you need to create rich and engaging audio content. Book a meeting to learn more.

TikTok's text-to-speech feature lets creators add virtual voiceovers at the touch of a button. It makes videos more accessible and eliminates the need to read.

So why do so many people hate on it? What makes TikTok voices so annoying?

Well, to start with, it's pretty common for TikTok's text-to-speech to get pronunciations wrong.

"Drop the mic" comes out "drop the mick".

It can't work out whether "read" should be pronounced like "red" or "reed".

And it definitely can't handle "quinoa".

The lack of voice options is a problem, too.

Creators currently have access to just one (region-dependent) voice. This can create a jarring effect for the listener, where the voice doesn't match up with the message.

@lucyedwardsblind NEW FEATURE ALERT! This is how you use Text-to-speech #learnontiktok #texttospeech #newfeaturetiktok #newfeature #blind Mood (feat. iann dior) - 24KGoldn

Take, for example, Scottish TikTok. As on Scottish Twitter, lots of captions are written in regional dialect. They simply wouldn't work with the British-English text-to-speech voice.

The North American text-to-speech voice has a "Valley Girl" speech pattern, which many consider to be annoying. Not only is "Valspeak" highly stigmatized in the US, but its chirpy nature simply feels inappropriate with a lot of TikTok content.

And it doesn't help that this replaced a more liked voice - one that was removed because the voice artist said she never gave permission for her voice clone to be used.

@leahova Guys, I had to be super raw here. So please be respectful. #texttospeech original sound - Leah

TikTok's synthetic voices sound robotic, too. Even if you ignore mispronunciations, their tone and inflections often sound very unnatural.

All of this is especially problematic when you consider that TikTok is a social media network. It's personal by nature. When creators use a synthetic voice rather than their own, it reduces the human connection.

Plus, the read-aloud feature offers limited benefits to TikTok viewers. Only small amounts of text appear on videos and the platform is highly visual anyway, so it isn't empowering users to escape their screen or multitask. Nor is it having a huge impact on accessibility.

Users can't really opt out of listening, either, as text-to-speech is played automatically if enabled by the creator.

Is this the best that TTS has to offer?

In a word, no.

TikTok's text-to-speech feature is no doubt impressive. And with 1.4 billion views on #texttospeech alone, it's almost certainly raised awareness of the technology's benefits. It's updated perceptions, too (for many, synthetic speech still brings Stephen Hawking's voice to mind).

But it doesn't showcase the best of what modern text-to-speech has to offer. Which is fair enough: it's not a major feature, and the current version is fit for purpose.

The most advanced text-to-speech can interpret and read text like a human. Often better. And while that might not be necessary on a social media network, it's crucial when converting articles, guides, and other written content into audio.

BeyondWords text-to-speech

At BeyondWords, we use natural language processing (NLP) to apply speech synthesis markup language (SSML) to text inputs. This acts like a virtual voice director, telling the AI voice how to pronounce elements like "mic", "quinoa", and "read" properly.

We also use advanced synthetic voices, developed using deep learning algorithms. These AI voices are trained on real human speech, ensuring more naturalistic inflection and intonation. We even offer voice cloning, allowing creators to "speak" in a voice that truly resonates with their audience.

Our text-to-speech platform is specially developed for publishers. Book a demo with our team to learn more.

Converting articles into audio is one of the most rewarding audience engagement strategies. Not only can adding narration be quick, easy, and affordable, but people are listening more than ever.

Keep reading or listening to learn about the seven key benefits of converting articles into audio:

1. Improve engagement

Making your articles available in audio allows you to better engage audiences with listening needs and preferences.

For example, people who are multitasking, prefer listening to reading (or reading and listening simultaneously), have certain vision or reading difficulties, or want to take a break from their screen are more likely to stick around when audio narration is available.

"Good marketing is about engaging with people the way they want you to." - Rachel Worsley, CEO at Neurodiversity Media

Plus, studies suggest that audio can resonate more deeply than other content formats. UCL found that "listening to audiobooks elicits a more intense physiological and emotional reaction than watching films or television", while the National Literacy Trust says that listening can aid comprehension.

So, when people do opt to listen, they may have a more positive experience with your content. This can have a number of benefits, such as making them more likely to return and recommend your articles to others.

2. Increase traffic

There are many cases where someone will want to listen to an article but not read it. If you cater to this audience and target them through your marketing efforts, you can drive more traffic to your site.

You can promote audio functionality in social media posts, meta descriptions, and many other areas to secure click-through from listeners - not just readers.

Plus, people may be more likely to share and link to your content if it has an audio version, because they understand this makes it accessible and engaging to a wider audience. This can boost your traffic even further.

3. Improve accessibility

Adding audio narration to your articles improves accessibility and inclusivity for people with vision impairments and reading limitations. This includes people with dyslexia, blindness, and low literacy.

Another benefit is that you have control over how your content sounds, because users don't need to resort to screen-reading software, which can provide a poor listening experience.

One of our publishers' readers, Troy Phillips, said: "Audio articles help me get the news easier and faster. I have dyslexia, which makes it hard to read the news. Screen-reading software is unreliable, and I don't always have access to it. So I really appreciated it when my chosen news source started using BeyondWords."

4. Increase subscriptions and reduce churn

In a subscription-based revenue model, offering audio articles as a subscriber perk can improve your conversion rate:

In the US, 36% of digital news subscribers say one of the main benefits is getting content that is only available to paying customers

In Mexico, access to more audiovisual content is one of the most important factors in digital media subscription decisions

In the UK, 48% of digital news subscribers say they pay for online access because they get a convenient package of news and information

However, a bigger benefit may be reduced subscription churn.

13% of news subscription cancellations result from subscribers having too much to read in too little time. The Economist's then Head of Strategic Product Development, Denise Law, referred to this as "unread guilt factor" when talking about the publication's move into audio.

When you offer audio, subscribers can consume articles while they're on the go. And because they find it easier to make time for your content, they're more likely to feel that their subscription is justified.

"Our evidence suggests that the audio edition is a very effective retention tool; once you come to rely on it, you won't unsubscribe." - Tom Standage, Deputy Editor at The Economist

5. Increase ad revenue

Audio articles can also increa...

Please note that this post was published before SpeechKit rebranded to BeyondWords.

As consumer demand for audio increases, more publishers are turning to text-to-speech. AI audio publishing platforms like SpeechKit offer a quick and cost-effective way to make written content listenable.

But how are visitors engaging with these AI audio-enabled articles?

To find out, we've compared listener and non-listener engagement across more than 28 million sessions in which the SpeechKit Player was loaded last year.

AI audio captures the attention of new users

The average new user spends just 2 seconds on-site when they don't engage with AI audio. This increases to 225 seconds (+11,150%) when they do press play.

In an increasingly competitive online world, publishers need to do more to stand out and convince new readers they're worth their time. Offering your written content in an audio format is an effective way to capture their attention.

AI audio keeps visitors coming back for more

Listeners are 32% more likely to engage in multiple sessions than non-listeners, suggesting that audio keeps users coming back for more. With a Northwestern University study showing that frequency of consumption is the biggest predictor of subscriber retention in digital news, AI audio can play a key role in sales and customer lifetime value.

"The number of cases for how audio fits in with media companies' subscription businesses is growing." - Lucinda Southern, Media Editor at Adweek

Plus, returning users are 38% more likely to press play than new users. This means audio articles are more popular with habitual visitors - visitors who drive the most revenue in both advertising- and subscription-based revenue models.

When returning visitors do opt to listen, they stay on-site for longer (+688%) and visit more pages along the way (+11.48%).

Listeners visit more pages

The average non-listener visits 1.17 pages per session, whereas the average listener visits 1.39 pages (+19%) before leaving a site, according to our analysis.

This finding suggests that listening experiences encourage visitors to stick around and explore. By providing content in a high-quality audio format, which some may find more accessible or engaging than text, publishers give themselves a better chance of keeping visitors on-site, potentially driving more ad revenue.

Listeners spend longer on-site

We found that non-listening sessions typically last just 30 seconds, whereas sessions involving audio last for 322 seconds (+973%). This means that users stay on-site over 10x longer when they press play.

As well as boosting brand loyalty and revenue, this increased engagement could benefit your search engine optimization (SEO) performance.

According to Backlinko: "Google pays very close attention to 'dwell time': how long people spend on your page when coming from a Google search. [...] The longer time spent, the better."

Making your written content listenable could therefore improve your search engine rankings, meaning that more people discover your content organically.

Audio engages at every age

Our analysis backs the established idea that younger demographics gravitate towards audio content. 18-to-34-year-olds represented 28.02% of our sample but 37.92% of listening sessions, making this the age group most likely to press play (1.5x more likely than 35 and overs). Audio engagement also had the biggest impact on session duration here (+1,109%).

However, it was listeners aged 55+ who remained on-site for longest, clocking an average session duration of 407 seconds. This group also visited the most pages per session (1.765), which represented the biggest increase versus non-listeners (+48%). This suggests that engaging older people with audio can result in the biggest pay-off.

Audio interaction also had a positive impact on the engagement of 35-to-54s, corresponding with a 2% increase in pages per session and 847% increase in session duration.

SpeechKit is an AI audio publishing platform that ...