• Dotika
  • Posts
  • AI-generated voices achieve human-level realism

AI-generated voices achieve human-level realism

ALSO : xAI offers Grok to American government cheaply

News of the day

1. New research shows AI-generated voices, including clones, are now indistinguishable from real human voices Read more

2. Elon Musk's xAI is making its Grok chatbot available to the U.S. federal government for an incredibly low price of 42 cents per year  Read more

3. Samsung launches TRUEBench to accurately measure enterprise AI model productivity in real-world scenarios  Read more

4. Meta FAIR introduces Code World Model (CWM), a 32B parameter open-weights LLM for code generation  Read more

Our take

Hi Dotikers!

AI voices are now undetectable, and it changes everything. A study from Queen Mary University of London published in PLOS One has just confirmed this technological turning point: AI-generated voices are now as realistic as human voices, making it nearly impossible to distinguish between them.

The team tested 80 English voices with ElevenLabs. The result? Cloned voices (generated with just 4 minutes of audio) fooled listeners 58% of the time. You might as well flip a coin. More disturbing: these synthetic voices are perceived as more dominant and sometimes more trustworthy than real voices.

This revolution opens up fascinating possibilities. Google's NotebookLM transforms any document into a podcast with two hosts who naturally discuss the topic. Who hasn't dreamed of listening to their reading materials in audio format during their commute? Creators generate multilingual content without a studio, companies create personalized training materials...

However, be aware: this impressive performance mainly concerns English. In other languages, the quality remains significantly inferior, with artificial accents and less natural intonations. French is improving, but we're still far from the realism of English-speaking voices.

The challenges are real: sophisticated phone scams, audio deepfakes, political manipulation (like that fake Biden call to American voters). But rather than panic, we need to understand that we're entering a new era. The tools are here, they work, and the question is no longer whether this technology will transform our daily lives, but how to use it intelligently.

A.

Meme of the day

Reply

or to participate.