Audio & Speech

Today's Stories

1. 1

   ElevenLabs releases AI music generator that blends opera with heavy metal in single songs

   ElevenLabs launched Music v2 on May 28, an AI music generation system that can create songs transitioning between different genres like opera, heavy metal, and rap while maintaining musical coherence. The new system includes an "inpainting" feature that lets users regenerate specific sections of a song without affecting the rest of the track.

   Musicians and content creators can now generate complex, genre-blending compositions without traditional music production skills or expensive studio time.

2. 2

   Spotify launches AI audiobook creation tool powered by ElevenLabs voices

   Spotify introduced an AI-powered audiobook generation tool on May 21 that uses ElevenLabs' voice synthesis technology to convert written text into spoken audiobooks. Authors can create audiobooks without hiring voice actors and are not bound to exclusive contracts, meaning they can publish the generated audiobooks on any platform.

   Authors can now turn their books into audiobooks quickly and affordably, potentially making audiobooks more widely available to readers.

3. 3

   MOSS-TTS v1.5 improves voice cloning with better multilingual speech generation

   OpenMOSS released MOSS-TTS v1.5 on May 26, an upgraded text-to-speech system that can clone voices from short audio samples and generate speech in multiple languages. The new version offers more stable voice cloning with improved speaker similarity and supports better language switching within the same conversation.

   Content creators and developers can generate realistic speech in different languages using the same voice, useful for multilingual presentations or language learning apps.

4. 4

   Developers create speech recognition pipeline that outperforms Whisper AI

   A developer shared a custom speech-to-text pipeline on May 25 that combines multiple AI models and audio preprocessing techniques to achieve better accuracy than OpenAI's Whisper, especially for noisy or multilingual audio. The system includes audio cleanup, speaker identification, and AI-powered transcript correction.

   Businesses using speech recognition for transcription or voice commands may get more accurate results, especially in challenging audio environments.

5. 5

   LongCat releases video avatar system that syncs lip movements to audio

   Meituan's LongCat team released Video Avatar 1.5 on May 23, an AI system that generates realistic talking head videos from audio input. The system uses Whisper AI for better lip synchronization and can create long-form videos while maintaining consistent facial features and natural mouth movements.

   Video creators can generate realistic talking avatars for presentations or content without appearing on camera themselves.