Voices.com has launched its beta program for a new AI, and we had the opportunity to test the platform. Voices.com has already answered important questions about the AI Studio and how it deals with talent and customers in its own FAQ. We would like to not only shed more light on this information, but also provide a hands-on critical impression of how the platform performs in everyday use.
What is the AI Studio?
Voices.com’s AI Studio positions itself as a platform that enables customers to pre-listen to, purchase and download the “cloned” voices of professional speakers via text-to-speech (TTS). For the talents, i.e. the speakers themselves, this offer is supposed to provide a new source of income by making their voices available for the cloning process. But what about the quality of the generated audio files? Can AI Studio really reach the level of professionally recorded texts? We will examine these questions in detail in the further course of our test.
Our experience with the AI Studio beta
Once logged into the beta platform, the process begins with a simple question: What kind of project do you want to create? A limited selection of voices is then presented, currently around 10 different options. Each of these voices is presented with its own demos, so you can click through the examples and get a feel for the particular “personality” of the voice. Once you have decided on a voice, you are taken to the next step, where you can enter the voiceover text. In addition, the interface offers the possibility to add pauses and adjust accentuation individually.
Source: https://www.voices.com/studio?utm_source=voicesai&utm_medium=email&utm_campaign=voicesai_early_access
Afterwards, it is possible to listen to a preview of the entered text. However, at the time this blog article was written, it was not yet possible to download the generated audio file. Therefore, it remains unclear how the quality of the “preview” differs from that of the final, purchasable audio. What is clear, however, is that the output of Voices.com’s AI delivers comparable results to other AI tools currently on the market. So it doesn’t seem to offer any significant quality advantage, at least in the beta phase. Apart from the fact that Voices.com allows you to identify the artists behind the Ai Voices as individual artists (people).
The quality of Voices.com Beta AI output
Since it is currently not yet possible to pay for and download the generated voices at Voices.com, we cannot (yet) present any audio samples here for legal reasons. Instead, we would like to share our experiences with the previews generated there.
– The basic emphasis and naturalness of the AI output are comparable to other text-to-speech voice generators.
– As expected, the AI-generated voices lack the ability to take context into account that a professional speaker would bring. As a result, emphasis is often not placed as it ideally should be in a professional context.
– There are sometimes small artifacts, especially at the beginnings or ends of words.
– Editing the pauses proves to be tedious, but works quite well on the whole.
– Audio quality varies from voice to voice, which could possibly be due to different training material or equipment of the respective speakers.
– For short sentences and simple applications, the AI seems quite usable. However, it is far from making subtext audible or specifically arousing emotions.
– For branding, this AI does not seem suitable at the moment and it remains questionable whether AI will ever have what it takes to fulfill the requirements for this.
– Downloading the output appears to be comparatively expensive compared to other AI tools. As an example, $0.50 per word for a voice named Rachael West resulted in a cost of $72.80 for a 140-word script in our case.
Overall, we can conclude that Voices.com’s AI shows potential in its current beta version, but still has some weaknesses that need further improvement. Furthermore, Voices.com emphasizes in their own blog article:
“For longer listening experiences, there is still no suitable substitute for the real human voice. For things like corporate training, educational content, or public service announcements where the goal is to inform, or anything meant to inspire or entertain, a real, dynamic voice should deliver the content.”
https://www.voices. https://www.voices.com/blog/your-ai-voice-questions-answered/
The ethics behind the Voices.com approach
Voices.com emphasizes acting ethically. Thus: consent, recognition, and reward – are central principles in the collaboration with talents.
The transparency with which Voices.com communicates its ethical principles is commendable, and the idea of creating an easy source of income for talent sounds good, too. Voiceover artists can even use restriction filters to control which words their cloned voice can and cannot say. After completing a project, talent even gets to see the script used.
Conclusion: AI voices vs. human voices
Voices.com, with the introduction of AI-generated voices, at least shows a willingness to offer voice actors a new source of revenue. This is commendable in this day and age when AI is increasingly being used in various fields. It’s also positive that Voices.com has established ethical guidelines for working with talent, including the intention to compensate them for the use of their voices.
However, it must be clearly stated: In its current beta phase, Voices.com’s AI technology is simply not mature enough for professional applications. From inconsistent audio quality to undesirable artifacts and a lack of ability to effectively convey emotion or subtext, the voices generated in the preview are far from a substitute for a real human speaker. In addition, the cost of downloading the final generated audio files seems disproportionately high compared to other providers.
In short, while the idea and ethical considerations behind the project are commendable, the execution still has a long way to go to be truly attractive for professional applications.
After all that AI chirping, longing for the sound of real human voices? You can hear some good real human voices on the artists overview page of Audiobird.com.
