January 2023 Summaries

ElevenLabs, an AI voice technology startup, has launched a Beta platform designed to enable creators and publishers to narrate long-form content, following a successful $2 million pre-seed funding round led by Credo Ventures. The platform utilizes a deep learning model for speech synthesis to produce human-like intonation and inflections, offering features such as voice cloning and synthetic voice design. ElevenLabs aims to revolutionize audio storytelling and expand multilingual audio support through their AI dubbing tool, which will allow users to re-voice audio or video content in different languages while maintaining the original speaker's voice. The company, founded by ex-Google and Palantir employees, prioritizes research and innovation to overcome challenges in artificial audio, positioning itself as a leader in AI speech technology with potential applications across education, gaming, and media. Supported by venture capital and individual investors, ElevenLabs aims to make high-quality spoken content universally accessible, promising to significantly enhance the reach and engagement of audio content in various industries.

ElevenLabs has released an updated version of its speech synthesis technology, which now offers improved performance on longer text fragments due to changes in training methodologies. This update includes support for cased input, enhancing the model's ability to accurately read names and create pauses, as well as components necessary for infilling and extending the model's language support. Users are encouraged to test the enhancements via the usual panel, with the expectation of minor changes to cloned or default voices. Additionally, ElevenLabs has introduced Agent Workflows, a visual editor for designing complex conversation flows, and shared a customer success story about Avidio's use of AI voices for personalized video outreach.

ElevenLabs is introducing Voice Settings as part of their Speech Synthesis product to give users more control over cloned voices, focusing on two main settings: Stability and Similarity. Stability allows users to decide whether they want the voice to be more consistent or expressive, with higher stability leading to cleaner speech and more variability potentially causing slight artifacts. Similarity adjusts how closely the output resembles the uploaded voice samples, with cleaner samples producing better results. The platform also highlights its capabilities in inbound sales through an AI Sales Development Representative (SDR) that can qualify 78% of leads in over 30 languages, as well as its proficiency in voice cloning across 12 Indian languages, showcased live at IIT Delhi.

Eleven Labs is pioneering the use of generative voice AI, emphasizing its potential in creating synthetic voices for various applications like audiobooks, games, and advertising. Their model allows users to design unique voices by setting parameters such as gender, age, accent, pitch, and speaking style, offering a solution for users who need tailored voices for specific content. This innovation is seen as a way to expand the accessibility of audio content, providing opportunities for creators and publishers to engage audiences with unique, brand-specific voices. Despite concerns about AI replacing professional voice actors, Eleven Labs envisions a future where voice actors can license their voices for AI training, enhancing their reach and preserving their vocal legacy. The company is committed to ethical AI use, implementing safeguards against misuse and supporting intellectual property rights. Looking forward, Eleven Labs plans to enable users to enhance and manipulate their own voices using their technology, broadening the possibilities for personalized audio content creation.