Google introduces text to speech models that generate custom voices from written descriptions

Google has launched new artificial intelligence tools that can build custom synthetic voices based on plain text descriptions. According to a post on the company's official blog, these new Gemini text to speech models can generate realistic speech in over 100 languages. Creators can prompt the system by specifying tone, style, and character traits, enabling the tool to produce multi speaker conversations directly from written text.

This development simplifies the audio creation process for podcast producers, game developers, and video creators who previously needed professional actors or technical software to record distinct roles. By making custom voice generation as easy as drafting a prompt, digital media can become significantly more dynamic and accessible across regions. To address misuse concerns, Google noted that the models embed digital watermarks into the output to verify whether an audio clip was synthetically produced.

However, the technology raises ongoing questions regarding voice actor rights and creator compensation. While digital watermarks help identify generated audio, it remains unclear how reliably these safeguards will hold up against deliberate tampering. As synthetic voice tools become routine features in media software, legal standards defining who owns a digital vocal identity will need to evolve.