Emotive voice AI startup Hume launches new EVI 3 model with rapid custom voice creation

[ad_1]

Join our daily and weekly newsletters for the latest updates and exclusive content in the industry’s leading AI coverage. Learn more


New York-based AI Beginner Hume has introduced the latest empathic sound interface (house) spoken AI modelThe house 3 (like Pokémon’s character “is the” Evene “Evene” Evee “Evee” Evee “Evene” Evee “Everything to” like Pokemon character “, as a” three, pokemon character).

The house allows 3 users to create their voices by conversation with the model (noise talking) and a new standard to adjust their responses in terms of expression and “empathy and” empathy and word choice.

The house is expanding the previous sound models of the Hume by offering a house 3, more complex personalization, faster answers and developed emotional understanding, designed for developers and creators.

Individual users can interact with this today Hume’s live demo on the website and iOS application, but the acquisition of the developer through the Property Application Programming Interface (API) is said to be given “in the future weeks” as a The company’s blog post States.

At this point, Developers will be able to get 3 of the house 3 to their customer service systems, creative projects or virtual assistants – price (see below).

Demo’s use is a new, special synthetic voice in seconds for the qualities I explained to me – a warm and confident mixed and men’s tone. Speaking that other AI models, more naturalist and easy to feel, and of course, the stock market sounds from Legacy Tech leaders with Siri and Amazon.

WhoDevelopers and enterprises need to know about 3

Hume’s house 3, for a number of use – The interaction of customer services and applications has been built up to the creation of audiobooks and content in the game.

It allows users to specify accurate personality features, vocal qualities, emotional tones and conversations.

This means that “a french, whispering a French scored from the cheese from cheese from cheese, a french, a naughty mouse, is a hot, empathic guide to a mischievous.

The main power of the house is the ability to integrate the emotional intelligence directly in voice-based practices.

Unlike traditional conversations or voice assistants in scratched or text-based interactions, how to speak 3 people naturally, the pitch, Prosodia and Vocal explodes to create conversations as more attractive.

However, a great feature Hume’s models are currently not currently offered by an open source and owner by opponents, such as a user or other voice, such as a user or other voice.

Again, as noted as “coming soon”, the Hume website will add such an ability to the speech model and previously reported that the company really reported the company’s report will be repeated for five seconds of votes.

Hume reported that the security and ethical considerations were prioritized before presenting this feature. Currently, in this cloning ability home, instead of the flexible sound customization is not available in Hume.

Internal assessments show users of the show users prefer to prefer 3.

Hume’s 1,720 user Hume’s own tests, home 3 was preferred Openai’s GPT-4O Rated in every category: The concept of nature, expressiveness, empathy, empathy, uninterrupted processing, response, sound quality, sound quality, sound quality, sound quality, voice emotion / style modulation and demand “Instructions” features are shown below).

Also, Generally Google’s Gemini Model Family and New Open Source is the best AI Model Company Sesame Past Oculus Compassionate Brendan irib.

Also, the lower delay (~ 300 millisities), healthy multilingual support (English and Spanish, more languages) and with effective unlimited sounds) are delayed lower). Hume writes on the website (see SWICEXOOT immediately below):

The main opportunities include:

  • The generation of Prosodia and not talk about expressive text with modulation.
  • StagnancyProvides dynamic speech flow.
  • Voice customization in the conversationTherefore, users can adjust the style of speech in real time.
  • API-Ready Architecture (Soon), so the developers can integrate 3 direct applications and services.

Access to the evaluation and development

Hume, house, Octave TTS and expression offer flexible, use-based price along the Measurement Apple.

House 3’s special API prices have not yet been announced (are marked as TBA), sample, it will be used with enterprise discounts for large placements, is based on use.

For reference, the house is a relatively 30% price in 2 minutes to $ 0.072 – 30%, home 1 (0.102 / minute).

For the developers and developers working with text and speech projects, the Hume’s Octave TTS is free from one step (~ 10,000 characters of speech, ~ 10,000 audio). Here’s an accident:

  • Free: 10,000 characters, unlimited special sounds, $ 0 / month
  • Beginner: 30,000 characters (~ 30 minutes), 20 projects, $ 3 / month
  • Creator: 100,000 characters (~ 100 minutes), 1000 projects, use-based other ($ 0.20/10 characters), $ 10 per month
  • Projector: 500,000 characters (~ 500 minutes), 3000 projects, $ 0.15 / 1,000 Extra, $ 50 / $
  • Scale: 2.000.000 characters (~ 2000 minutes), 10,000 projects, $ 0.13 / 1000 Extra, $ 150 / month
  • Work: 10,000,000 characters (~ 10,000 minutes), 20,000 projects, $ 0.10 / 1000, $ 900 per month
  • Venture: Special price and unlimited use

For real-time voice interactions or developers working on emotional analysis, Hume also offers a payment for the lack of $ 20 in free loans and the lack of pre-commitment. High-volume enterprise can choose a special enterprise plan, which are customers, database, on-service solutions, special integration and advanced support.

HUME History of emotional EU sound models

In 2021, Google DePrmind was established by a former researcher Alan Cowen, the Hume is aimed at eliminating the gap between human emotional nuances and AI interactions.

The company has developed models in a wide range of databases compiled from hundreds of thousands of participants who are not only speech and text in the world.

“Emotional intelligence, intentions and advantages and advantages of behavior, this is a very nucleus that the AI ​​interfaces are trying to achieve,” Cowen Venturebeat said. Hume’s mission is to make AI interfaces more sensitive, as more sensitive, as a person – a customer to go to an application or explaining a story with a proper mixture of drama and humor.

Earlier 2024, the company’s house launched 40% low delay and a 30% reduced price compared to 1, dynamic sound customization and conversational style decreased by 30% compared to new features.

February 2025, Octave’s debut, for text requests for text requests, saw a text engine from a text that regulates sensations in the sentence.

Huma, which is only handed by hand-handing and only the full API, allows you to restore what is possible with a voice AI for the corner of the full API.


[ad_2]
Source link

Leave a Reply

Your email address will not be published. Required fields are marked *