Twisted50 vol III Audiobook… I have run some tests…
So what do we think of this text to speech version of Sasha’s story from Twisted50? Listen below…
As you know Sasha has volunteered his story for tests, the first one being the Feedback test from last week. You can read about that HERE.
The tool I used to create this audio is from ElevenLabs which, to my ear, is by far the best one I have heard. I used the voice An-Rita with default settings FYI. The music came from Envato Elements.
A couple of phrases were not landing right, so I rewrote bits to get it to pronounce or ‘perform’ correctly or more appropriately. It’ struggled with ‘cos‘ for instance, and eventually I got it to work by changing it to ‘coz’.
Performance wise, it delivered some lines too abruptly, so I would add a phrase like, ‘she the said softly…’ and I would edit out my prompt later, inserting the new delivery to the edit.
I did an final edit using Adobe premiere, to add music too. I suspect the whole thing took less than an hour end to end. And this was my first one, working out the process and wrinkles. Moving forward, I think I could get around 20 to 30 mins of completed audiobook per hour of my time.
So what do you think? Should we do this? Comments below please…
If you want to try ElevenLabs you can get a free account for testing HERE. It’s an affiliate link so if you do sign up, Twisted will get some additional help toward the costs of running the project.
Chris Jones
www.Twisted50.com
Filmmaker, author, screenwriter and founder of Twisted50. Currently directing splinter unit on Mission: Impossible 8.
18 thoughts on “Twisted50 vol III Audiobook… I have run some tests…”
What I wonder in my ignorance is if there is a way we can write our own stories then feed them into an AI which will read them back to us and whether we could do the same with any entries put onto the site to hear how they play out for evaluation and enjoyment purposes? No idea about the implications about that.
About this reading? It’s pretty good, has some emotional voice to it though of course one can’t help but wonder if say it were possible to hear it read in say in Michael Sheens wonderful dramatic voice, but I imagine large sums of money would be involved with that!
It’s a great way to envision the story and audiobooks are so popular these days and of course useful for those who struggle to read and a change from reading at which many of us have a short attention span.
Yes, it would be great if we could have amazing actors do this, but there is zero budget for it.
And to answer your question, yes you can add your own story and hanve it read it back to you. I do this all the time now, even though its not always perfect, hearing your prose is so valuable to spot things the eyes dont pick up on.
It’s a lovely idea but cutting voice actors out of the loop, Chris.
Actors turn up to help for Table reads etc at LSF and I like that interaction as it makes work come alive. So, one hand, sounds amazing on the other real people who become friends and mentors are better.
I class many people I’ve met via LSF as genuine friends and have met up for coffee etc. I don’t want to have a drink and giggle with my laptop!
I understand. So the simple question is this. We have no budget and no actor can commit to this with no money. And shepherding 50 actors to do as freebies is a thing. So it’s this or nothing. Unless each author is prepared to pay for actors and mastering. And this is just an idea ATM. I had no plans to do this. I likley won’t as there is likley to be a backlash. So it’s, as the blog says, a question. What do you think?
“to hear how they play out for evaluation and enjoyment purposes?” Great idea, Lee.
It certainly sounded very professional, crisp and clear, the music addition really helped, and the voice sounded very realistic. For me it’s hard to know where to draw the line with AI – I’d taken your experiment with the feedback on and fed some more of my own work into that to see how it would work, and have been happy with the results. My dilemma would be is this taking work away from a human who would be paid for providing feedback? I likely wouldn’t pay someone to provide feedback on a short story, but certainly would if I was working on a novel or something longer, and I think this applies to the voice actors who would be contributing to this and any other work as well. But then everything costs money, and paying for a voice actor/actors to read the stories for an audiobook might be too expensive, and could lead to projects such as this not happening. I think it has to come down to affordability and suitability – if it’s at all possible for a human to provide the service (but not at the expense of the project itself), then that should be the way forward. But it certainly is an interesting conversation!
As you suggest, it’s the boundary and where to place it? There was and is no plan to have an Audiobook. This is just an exploration. I am mindful of two things. One, the vitriolic attacks that some people will mount for even considering AI. Second, the very likley reality that in five years, the deabte over AI will be over and it will just be part of our reality. I am not right about this, it’s just my view on where the tech is going, the speed it’s going, and how humans have embraced or rejected tech in the past. The genie is out of the bottle.
It’s good but I’d give it 7/10 compared to a human.
It’ll probably be light years better in a couple of years.
The question is how long till we’re all sitting up and begging for treats from our AI masters. Hoho.
No one told me we’d be getting treats! Roll on, the singularity!
Pretty much in accord with the other commenters I think. Better than I expected, in a lot of ways.
I guess personally I’d be happy to record a reading of my own story myself, but I admit there might be quality/mastering implications there which would be hard to overcome.
With my challenges relating to reading, I welcome the fact that an audio representation of a story can be facilitated via either a live human or an AI-generated voice.
Chris, you did an excellent job creating this audio presentation of Sasha’s story.
Your choice of background music, using music only for transitional breaks, and the blending in of mood music for the final story climax added the cues to trigger my emotional senses.
I can visualize a story when I write, but it is not easy for me when I read. Reading slows the process, audio increases it.
Listening to this audio, allowed me to get, a different perspective on Sasha’s story. I no longer had to stumble and re-read a paragraph to get the visual representation of what the author would be trying to convey.
Yes, there will always be a concern/fear about using AI, which could, under certain conditions and developments reduce or eliminate trained employment in a multitude of disciplines.
AI can use your voice to train, so you can instantly convert your story from text to speech, saving you countless hours, to create an audio-book of your story/s. Imagine, you have perfected your story, you use your own AI-trained voice generator, and your story’s audio is now created within seconds. You have now maintained your creative integrity.
I have mixed feelings regarding AI technologies, but I have learned, over the years, that I will always need some form of editing, guidance, support, and creative assistance, which for me, has been difficult to get access from humans.
But what is interesting, is that AI is typically trained using humans knowledge and creativity.
Respectful acknowledgment of each person’s opinion is crucial.
Human contact is always my first choice.
Pretty good. The faux breathing became irritating after a while; I know it’s there to make it seem real, but a professional voice actor, and a proper mic, would cut out all the extraneous breaths. Or maybe, as a trained actor, I just find it threatening.
It was alright, in the sense that it was clear, but I wonder how well it would do with a script containing several characters: most of this was reported speech, so no variation was necessary, and the couple of direct speeches were marginally different. Yet, if you have a multitude of characters, like Harry Potter, how difficult would it be to identify the different characters talking. Is there a prompt for you to highlight a quotation marked part of the text, say in different colours, and match each colour to a different pitch? Time consuming.
Having tried, and still trying, to use ai image generators for my YT vids, and battling with six fingers and 3 legs on so many creations, and the AI not paying attention to all of the instructions, I feel deeply sorry for whoever has to control the program doing a longer text.
Technically, I was ‘gobsmacked.’ I thought it was amazing.
Aesthetically, I was not so convinced. The voice was far too young for the story. Almost childlike. It also lacked any light and shade. Emotion is vital for a narrator. This AI version lacked ‘the human’ touch. Not surprising really as it isn’t human. Probably great for reading technical manuals or works memos. But for stories that touch the soul, I would wait until they can get more human emotion into the reading.
Overall, I thought it worked well. The voice was clear and pleasant to listen to (though a bit flat, maybe, here and there), and for me suited the main character POV, that of a young teenage girl. The music complemented the reading . However, one particular glitch did show up the limitations of these systems, when Helen in the recording says “I can’t do it for *you*” [asterisks to show voice emphasis], where a voice actor, with understanding of what was going on in the scene, would have said “I can’t do it *for* you.”
I thought the A.I. reading was good, better than I expected and a useful tool to hear the story as it often sounds, evokes a different tone, when read out aloud.
With budget being an issue, if real actors were to be employed, could drama students help? It would gain them some exposure, some practical work experience… just my thoughts.
The problem is working with 50 inexperienced performers. Along with their inexperience at voice recording. This is six weeks work for a small team, with no pay, let alone the actors.
I wondered about drama students, myself. But aside from the above, another issue might be they are likely to be all around the same age, while some stories might be better voiced by an older or younger person. The AI has a whole menu of voices to choose from. If it’s acceptable to consider an AI solution for the audiobook, this particular one looks pretty good, to me.