AssemblyAI

Announcing Universal-2, our next-generation speech-to-text model! Building on Universal-1's industry-leading performance, we've made significant improvements in just 6 months, focusing on areas that matter most for real-world conversations: - 24% better at handling proper nouns - 21% improvement in alphanumeric accuracy - 15% enhanced text formatting - Maintains 30% reduction in hallucinations versus other speech-to-text models like Whisper - 73% of users prefer Universal-2 outputs compared to Universal-1 What makes Universal-2 special isn't just better accuracy - it's solving the hardest challenges in conversational AI. From handling messy real-world speech to accurately capturing proper nouns, phone numbers, and formatting, Universal-2 delivers transcripts that are both accurate and clean. Universal-2 is available today through our API. Start building on Universal-2 today: https://lnkd.in/ex2QU_SP

Check out these results from Jack McDermott, who tested how accurately Speech AI models like AssemblyAI perform on speech with varying degrees of stutter ??

We ?? customer videos! VEED.IO's cofounders were determined to reduce the average user’s barriers to producing high-quality videos, so they set out to create a next-generation video editing platform -- learn how speech to text and summarization functionality from AssemblyAI helped them make it happen ?? https://lnkd.in/eRdPbWVq

We have a fresh AssemblyAI Required video out today! ?? Synthesia co-founder and CEO Victor Riparbelli met with Dylan Fox to discuss his journey in building a visionary AI-first company built on customer value and trust. The two discuss: 1?? How sometimes you have to adjust your vision to fit market needs 2?? How finding success as an AI-first company involves a lot of trial and error 3?? How you need to anchor yourself to your customer to create a successful roadmap And much more. Watch their full conversation here: https://lnkd.in/eG-Z7Qbf

?? Push the boundaries of speech AI with AssemblyAI in the latest DEV Community challenge! Build something incredible and you could win $1,000, a dev.to membership, and more prizes! Join the challenge here: https://lnkd.in/evU3kzj3

Announcing Multichannel: We've expanded our capabilities to include multichannel audio processing. Now, process up to 32 separate audio channels, enhancing speaker identification, readability, and reliability. What makes Multichannel important? It's solving real challenges across critical industries: - Call tracking - Customer support - Telehealth - VoIP Services The results? - Speech is accurately attributed to specific channels - Speech-to-text accuracy is improved through isolated channel processing - Contextual insights are enriched through multi-participant conversations Get $50 hours of free usage, and start using Multichannel through our API: https://lnkd.in/gTVKdVqM

Announcing Multichannel: We've expanded our capabilities to include multichannel audio processing. Now, process up to 32 separate audio channels, enhancing speaker identification, readability, and reliability. What makes Multichannel important? It's solving real challenges across critical industries: - Call tracking - Customer support - Telehealth - VoIP Services The results? - Speech is accurately attributed to specific channels - Speech-to-text accuracy is improved through isolated channel processing - Contextual insights are enriched through multi-participant conversations Get $50 hours of free usage, and start using Multichannel through our API: https://lnkd.in/gTVKdVqM

AI has quickly transformed into a core pillar of product strategies across nearly every industry and the mounting pressure on businesses to adopt and integrate AI has never been greater—in fact, more than 90% of organizations already have. Learn more about this race to integrate AI, including what product leaders and founders think about AI implementation and their top use cases. https://lnkd.in/eJPHhyPH

Interested to learn how our latest model, Universal-2 compares to Universal-1, and two Whisper variants? ?? Check out this article to learn about model performance regarding the finer details that are crucial for readable transcripts and downstream tasks: ?? Proper nouns (e.g. person names, places, brand names) ?? Alphanumerics (e.g. digits, years, phone numbers) ?? Text formatting (e.g. upper/lower case, punctuation) ?? Hallucinations Read the full analysis here: https://lnkd.in/em9RDTJv

?? Calling all NYC area developers: We are hosting a Hackathon on Friday, December 6 in NYC. Swag, food, prizes, usage of our models + follow-up credits for all attendees! Get more information and sign up here: https://lu.ma/newafxrf See you there!