How AI Can Transform the Reader Experience
While Maintaining Writing Authenticity, Without Compromising on Originality
I have always been a skeptic of AI writing tools. I don’t even use Grammarly because its AI rephrasing engine makes me look insignificant. Its suggestions for rewording my entire sentences make me feel demotivated. My limitations as a human writer are exposed and I don’t appreciate it. My wisdom, my vulnerabilities, my flaws, my vocabulary, my slangs, my personal anecdotes — all of it put together is what adds originality to my writing.
So I have never used tools like ChatGPT to assist me with my writing.
Enter Eleven Labs.
ElevenLabs is a voice AI research & deployment company with a mission to make content universally accessible in any language & voice. ElevenLabs creates the most realistic, versatile and contextually-aware AI audio, providing the ability to generate speech in hundreds of new and existing voices in 29 languages.
Adding voiceovers is the best way to add life to our words. But hiring voiceover professionals requires a careful vetting process of selecting the perfect voice, tone, accent, pace, and price, suitable for our needs. We are charged by the word count and it takes several iterations to complete the project.
Eleven Labs has made it simpler, quicker, and easier to get started. With a freemium option limited to 10000 characters a month it now takes only a few clicks to generate voiceovers.
I tested the tool with my recent flash fiction post, The Staircase, a quick read of 234 words (1252 characters). The narrative transports the reader to a suspenseful moment on a staircase dominated by uncertainty. The voiceover needed to draw out that emotion in its narration, without which the post would be a bland read.
It took less than 5 mins to generate the file, and no payments were required.
Honestly, what I expected was a robotic voice sans emotion. Instead, the result was mind-blowing. This AI tool generated a text-to-speech audio file that’s as human as it can get, IMO.
Without further ado, here’s “The Staircase” narrated by AI Rachel.
I went with the “Rachel” premade voice that was pre-selected for me on the screen. Rachel’s voice is that of a calm American narrator. There are many premade voice variations for you to select from, as shown in the screenshot below.
The tool can also record our voice and store it as a custom premade. I have not tried this out, but others have reported a near-perfect cloned result.
In a world where the lines between fact vs. fake are becoming increasingly blurred, I am sure there are those who will misuse this tool to deepfake voices.
In February, VICE reporter Joseph Cox published findings that he had recorded five minutes of himself talking and then used ElevenLabs to create voice deepfakes that defeated a bank's voice-authentication system.
But for writers, I believe it is harmless to use, saves effort and money, and the result is an elevated reading experience that humanizes our writing with AI voiceovers…
i.e. if it actually works.
UPDATE 3/5/24: My 2nd experiment with the voiceover generation using the American, casual, conversational voice of “Chris” to translate a 5000 character post proved unusable. The voice was exactly what I need for the post but within a minute of the audio review, Chris started to fail miserably. I noticed the falling flaws:
I used #(number) to indicate numbers but Chris failed to say “number”.
I used $(amount)B to indicate a billion dollar amount but Chris couldn’t say Billion or Million when it was not spelled out.
When it hit the previous error, the voice went from casual and conversational to robotic and non-American for a few sentences.
“popped” was read as “poppered”; “wear” was “veer”; “gaga” was “gaygah”.
That’s when I stopped and sent a feedback to the creators. I am waiting to hear from them.
So for now, I recommend we use simpler texts without abbreviations or slangs. These algorithms are not perfect but will likely improve over time.
Try it out and share your feedback in the comments below, or tag me in your post if you decide to use it. I hope it helps.