TLDR: AI startup Hume AI has launched EVI 3, the latest iteration of its Empathic Voice Interface, featuring a groundbreaking ‘hyperrealistic voice cloning’ tool. This allows users to create an AI replica of their voice by uploading a short audio recording, enabling verbal interaction with their AI clone. EVI 3 is designed to capture and reproduce the subtle nuances of human speech, including emotional context and speaking style, pushing the boundaries of empathic AI communication.
Hume AI, an innovative artificial intelligence startup, has introduced its third-generation Empathic Voice Interface (EVI), dubbed EVI 3, which includes a pioneering ‘hyperrealistic voice cloning’ feature. This new tool allows individuals to generate an AI replica of their own voice, offering a novel way to interact with technology. The announcement, made on July 19, 2025, marks a significant leap in AI’s ability to mimic human expression.
The core of EVI 3’s voice cloning capability lies in its ability to analyze and reproduce the intricate nuances of a person’s speaking style. Users can create their AI voice clone by uploading a brief audio recording, ideally between 30 and 90 seconds. The model then processes this input to replicate not just pitch and timbre, but also cadence, intonation, pauses, and even subtle emotional inflections that make a voice uniquely human. Hume AI states that EVI 3 aims to create a ‘true vocal fingerprint,’ capable of generating new speech that is virtually indistinguishable from the original speaker.
EVI 3 has been trained on extensive amounts of text and speech data, enabling it to interpret emotional context and respond to specific emotional prompts. This goes beyond traditional text-to-speech, moving into the realm of empathic communication. The model can also capture ‘aspects of the speaker’s personality,’ according to Hume’s blog post, although some early testers note that while the voice imitation is realistic in many ways, capturing personality remains a nuanced challenge.
Founded in 2021 by Alan Cowen, a former Google scientist, Hume AI’s mission is to build AI that understands human feelings. Cowen’s vision is to address the gap where current AI often misses the emotional expression in voices, aiming for AI that better serves human well-being through empathic interactions. EVI 3 is a testament to this goal, designed to make AI voices feel more natural in conversation, including factors like turn-taking cues and appropriate emotional undertones.
The implications of this technology are profound and multifaceted. On the positive side, it opens doors for ultra-personalized digital assistants that genuinely understand user emotion, new forms of content creation, and realistic voice restoration for individuals who have lost their ability to speak. However, the launch of such a powerful tool also raises critical ethical questions. The potential for misuse, such as creating sophisticated deepfakes for misinformation campaigns, fraud, or harassment, cannot be ignored. Hume AI acknowledges these concerns, stating that as this new era in voice AI unfolds, robust safeguards and ethical guidelines will be crucial to ensure the technology benefits humanity.
Also Read:
- Gamma Unveils Agentic AI to Revolutionize Presentation Creation
- OpenAI’s New ChatGPT Agent Poses Significant Bioweapon and Financial Misuse Risks
For those interested in trying the feature, Hume AI offers a free tool. The company’s website indicates that user data from interactions with the EVI API are collected and anonymized by default for model training, though users have the option to disable this through a ‘Zero data retention’ feature in their profile. For non-API products like the demo, data ‘may’ be collected for model improvement, with a similar toggle option available upon creating a personal profile.
