Introduction
On this page, we discuss various strategies you can implement to safeguard yourself in the evolving landscape of AI. We intend to continually update and expand this page with more insights and tips in the coming months.
On this page, we discuss various strategies you can implement to safeguard yourself in the evolving landscape of AI. We intend to continually update and expand this page with more insights and tips in the coming months.
We believe brand image will become increasingly important when competing with AI voices. AI voices are faceless, so you need to stand out against them. We recommend reviewing your brand image to enhance its human element. Now, we advise replacing avatar style imagery with real photos of yourself. If your appearance doesn't match your brand voice, then maybe showcase your studio instead.
The quality of AI voice recordings typically ranges from awful to okay. So we recommend you review your own studio sound and see if any improvements can be made.
Enhance the directed session experience by having a webcam you can use for sessions which support it.
There is a real benefit of a client doing a directed session rather than an undirected session.
These reasons include:
- A human connection is made between you and the client and/or producer
- A lot of voices at the cheap-end of the market are unable to take on directed sessions
- You give the client/producer complete control, something they will get used to having
- The recording session is more of an 'event'
- You can try out different styles of read and get instant feedback
If it is an undirected recording, then we recommend being flexible and open to free pickups for style.
If you hear an awful AI text-to-speech voice on a Youtube video, then consider making a comment (if comments are open). Especially if it's a big brand. You don't need to link it to you as a voice if you don't want to. Ad agencies seeing negative feedback on their clients videos will make them question if using a text-to-speech voice was a good idea.
The more angry or emotional response we have may only make clients wonder what all the fuss is about and draw attention to the tech.
We believe any such discussions concerning AI should be kept to fellow voice artists and close partners for the time being.
We need to retain the appearance of a cool confidence. Confident because we are aware of the strengths and limitations of AI, and that although it may be acceptable for some projects, most projects will benefit from a real human.
Tempting as it might be, we recommend avoiding getting a professional clone of yourself created, at least until better protection and laws are in place.
By professional clone we mean providing an AI company with more than 5 minutes of your audio data.
Also review the tools you use. If you use Adobe Cloud then opt out of feeding data into their future models (Google it).
Avoid dry studio samples on your demo audio where possible - add music/reverb/effects.
Whether it be an human or an automated scraping bot, AI systems work best when trained with clear, dry, quality audio - so don't make it easy for them to obtain this!
The client may not know they want it until they hear it. Consider giving extra takes, where you really sound more personable and emotional - even if the client didn't ask for it.
Consider:
- Closer mic reads may give more personable takes
- Avoid repeating the same tonal rhythms in your voice
- Be a bit more 'you', maybe a little more cheeky or be more expressive with your delivery
- Varying your pace where convenient
If you have any tips you want to share then let us know.
If you don't already, consider varying your fees depending on whether the script is expressive, or just an easy straight read.
Basically, factor-in whether it would be easy to do with AI and consider this when costing up.
If in the future legislation or lobbying allows you to do a 'take down' request e.g. On YouTube, when your voice likeness is used without your permission, then you may need to prove you sound like the voice on the video/content. One likely way they'd do this by taking a recording of you and comparing how 'far' your voice is in an AI model context, compared to the voice in question. Often changes in studio/recording sound can alter the voice so it would say it's not a close match. Being able to recreate the sound from when you were cloned would be important. Thus we recommend were possible keeping your old gear.