Emotive Voice Ai Startup Hume launches a new model for EVI 3 with a quick dedicated voice

Its headquarters AI Startup Hume has unveiled the latest AI model for the sympathetic voice interface (EVI)Evi 3 (Evee “Three, such as Pokémon), and targeting everything from the operation of customer support systems and health training to the novel of overwhelming stories and virtual companionship.

EVI 3 allows users to create their own voices by speaking to the model (it is a voice to a net/speech to a speech), and aims to set a new standard for nature, expression and “sympathy” according to what it is, that is, how users see understanding the model for their emotions and its ability to contradict or control its own responses, in terms of tone and selection of the word.

EVI 3, designer of companies, developers and creators, is designed on the previous audio models of the home by providing a more advanced allocation, faster responses, and enhanced emotional understanding.

Individual users can interact with it today Hume’s Live on its website And the iOS application, but it is said that the arrival of the developer through the HUME (API) application interface is available in the “coming weeks”, as a Blog post from the company Countries.

At this point, developers will be able to include EVI 3 in their customer service systems, their creative projects or apparent assistants – for the price (see below).

I allowed me to use to clarify the creation of a new artificial sound dedicated in seconds based on the qualities that I described to – a mixture of warmth, confidence, and male tone. Talking to her looked more natural and ease than other artificial intelligence models, and certainly stock sounds from old technology leaders such as Apple with Siri and Amazon with Alexa.

DoYou should know developers and companies about EVI 3

Hume’s Evi 3 is designed for a set of uses-from customer service and within the application to creating content in audio books and games.

Users are allowed to determine the characteristics of accurate personality, vocal characteristics, emotional tone and conversation topics.

This means that it can produce anything from a warm sympathetic evidence to a foam and harmful narrator – to requests such as “the mosquitoes that are urgently whispering in a French accent about its plan to steal the cheese from the kitchen.”

Evi 3 is the basic strength of its ability to integrate emotional intelligence directly into sound -based experiments.

Unlike traditional chatbots or vocal assistants who depend heavily on text or text-based reactions, EVI 3 adapts to how people speak normally-pick up on the stadium, patients, stopping, and vocal bangs to create more attractive conversations, and humans.

However, the HUME models that currently lack – which are presented by open source and ownership competitors, such as ElevenLabs – is sound cloning, or rapidly identical copies of the user’s voice or others, such as the CEO of the company.

However, Hume has indicated that he will add such the ability to a text or octav model to words, as it is referred to to “coming soon” on Hume on the web, and the previous reports have found to the company that it will allow users to repeat sounds from less than five seconds of sound.

Hume stated that she determines the priorities of guarantees and moral considerations before this feature is widely available. Currently, this cloning can not be available in EVI itself, as Hume emphasizes the allocation of a flexible sound instead.

The internal standards of users appear prefer EVI 3 over Openai’s GPT-4O audio form

According to Hume’s special tests with 1720 users, EVI 3 was preferred Openai’s GPT-4O In each category evaluated: nature, expression, sympathy, interruption processing, response speed, sound quality, sound/style modification upon request, understanding of emotion upon request (features are covered at “request” in “the following instructions” below).

Also, it is usually assumed to surpass the Google Gemini Model family and the new open source AI Sesame Model Company From the former author Oculus participating Brendan IRIBE.

It also includes less transmission time (about 300 milliliters), strong multi -language support (English and Spanish, with more upcoming languages), and unlimited unlimited sounds. Hume also writes on his website (see screenshot below):

The main abilities include:

• Times generation The expressive text to speak with the amendment.
• BoycottEnably the flow of dynamic conversation.
• Voice allocation in the conversationSo that users can adjust the actual speaking pattern.
• Architecture ready API (Soon), so that developers can integrate EVI 3 directly into applications and services.

Pricing and reaching the developer

Hume offers flexible price -based prices via EVI, Octave TTS and APIS to measure expression.

While the EVI 3 application programming interface has not been announced yet (it was marked with the name TBA), the pattern indicates that it will be used, with discounts for institutions for large bulletins.

To return to it, the price of EVI 2 is $ 0.072 per minute – 30 % lower than its predecessor, EVI 1 ($ 0.102/min.

For creators and developers who work with text projects to words, Hume’s Octave TTS plans range from free layer (10,000 letters of speech, about 10 minutes of sound) to plans at the institution level. This is the collapse:

• free: 10,000 letters, unlimited custom sounds, $ 0 monthly
• beginning: 30,000 letters (about 30 minutes), 20 projects, $ 3 per month
• creator: 100,000 letters (about 100 minutes), 1000 projects, use based on use (0.20/1000 dollars), $ 10 per month
• Professional: 500,000 letters (about 500 minutes), 3000 projects, additional 0.15/1000 dollars, $ 50/month
• size: 2,000,000 letters (about 2000 minutes), 10,000 projects, additional 0.13/1000 dollars, 150 dollars per month
• a job: 10,000,000 letters (about 10,000 minutes), 20,000 projects, additional 0.10/1,000 dollars, 900 dollars per month
• project: Custom pricing and unlimited use

For developers who work on actual time or emotional analysis, Hume also provides salaries as it plans with $ 20 in free credits and there is no obligation provided. Large -sized Enterprise customers can choose a dedicated institution plan characterized by licenses for data collections, local solutions, allocated integration, and advanced support.

Hume’s history in emotional artificial intelligence models

Founded in 2021 by Alan Quinn, a former Google DeepMind, Hume aims to bridge the gap between human emotional differences and self -interaction.

The company trained its models on an expanded data set derived from hundreds of thousands of participants around the world – not only talking and text, but also sound structures and expressions of the face.

“Emotional intelligence includes the ability to deduce intentions and preferences from behavior. This is the essence of what AI’s facades try to achieve,” Queen told Venturebeat. Hume’s task is to make artificial intelligence facades more responsive, human, and in the end more useful – whether it helps the customer to move in the application or narration of a story with a correct mixture of drama and humor.

In early 2024, the company launched EVI 2, which provided 40 % less than cumin and 30 % of pricing compared to EVI 1, along with new features such as dynamic audio customization and Holocaust.

February 2025 witnessed the emergence of Octave for the first time, a text engine to the words of the content of the content who are able to control emotions at the sentence level with the texts of the text.

With EVI 3 now for practical exploration and full access of API, just around the corner, Hume hopes to allow developers and creators to re -imagine what is possible with VOICE AI.

Daily visions about business use cases with VB daily

If you want to persuade your boss at work, you have covered VB Daily. We give you the internal journalistic precedence over what companies do with obstetric artificial intelligence, from organizational transformations to practical publishing operations, so that you can share visions of the maximum return on investment.

Read with us privacy policy

Thanks for subscribing. Check more VB news bulletins here.

An error occurred.