AB
AB
AiBoss

All Categories

project

Speech 2.6 - A speech generation model introduced by MiniMax

Speech 2.6 is a brand-new speech generation model from MiniMax, designed specifically for the next generation of voice-activated agents. It features ultra-low latency (less than 250 milliseconds) to ensure smooth real-time conversations; it supports URLs, emails, phone numbers, etc., in multiple languages.

What is Speech 2.6?

Speech 2.6 is a brand-new speech generation model from MiniMax, designed specifically for next-generation voice-activated agents. It boasts ultra-low latency (below 250 milliseconds) to ensure smooth real-time conversations. It supports direct conversion of non-standard text formats such as URLs, emails, and phone numbers in multiple languages, eliminating the need for cumbersome preprocessing. Utilizing Fluent LoRA technology, the model further enhances the naturalness of phonology and the fluency of timbre reproduction, generating high-quality speech even from original materials with accents or non-fluent delivery. The model is suitable for various scenarios, including intelligent customer service and smart hardware, and supports over 40 languages, providing users with an efficient and natural voice interaction experience. Users can access the model through the MiniMax Open Platform and the MiniMax Audio website.

Main features of Speech 2.6

  • Ultra-low latencyEnd-to-end latency is less than 250 milliseconds, ensuring fast and smooth audio generation in scenarios such as real-time dialogue.
  • Professional format without barriersIt supports direct conversion of non-standard text formats such as URLs, emails, phone numbers, dates, and amounts in multiple languages, without the need for cumbersome text preprocessing.
  • Greater naturalness and Fluent LoRAEnhances the naturalness of the phonology, supports timbre replication, and preserves the accents, speech patterns, and other characteristics of the original timbre. Fluent LoRA technology makes speech more fluent and natural, and can generate high-quality speech even if the original material has an accent or is not fluent.
  • Multilingual supportSupports 40+ languages and is applicable to voice interaction scenarios worldwide.
  • High-efficiency voice interactionIt is applicable to various scenarios such as intelligent customer service and smart hardware, and provides a smooth and natural voice interaction experience.

How to use Speech 2.6

  • Register/LoginVisit the MiniMax Audio website, register an account and log in.
  • Select speech synthesisIn the left navigation bar, click the "Speech Synthesis" option to enter the speech synthesis page.
  • Input textEnter the text you want to convert to speech in the text input box.
  • Select tone and modelBelow the input box, select your preferred voice (such as "composed executive") and speech synthesis model (such as "speech-2.6-hd").
  • Select application scenarioChoose the application scenario for speech synthesis as needed, such as "news broadcasting", "storytelling", "film and television dubbing", etc.
  • Generate audioClick the "Generate Audio" button, and the platform will generate audio based on the input text and selected parameters.
  • Download or play audioThe generated audio can be played online or downloaded and saved locally.

Application scenarios of Speech 2.6

  • Customer ServiceProvide natural and fluent voice interaction in call centers or online customer service systems to enhance customer experience.
  • audiobooksGenerate high-quality audio readings for ebooks, online articles, or educational materials.
  • voice assistantIt can act as a voice assistant to provide voice interaction services in smart home devices, mobile phones, or in-vehicle systems.
  • Radio and podcastGenerate professional-quality voice for radio programs, news broadcasts, or podcast content.
  • Language learningIn language learning applications, it provides accurate pronunciation demonstrations and language practice.
Back