AssemblyAI

In the last few years we've seen an explosion of audio data available online. This coupled with advances in AI technology have allowed organizations to unlock the value of voice data in ways that were previously impossible. As a result, we've seen organizations build new products, services, and capabilities that serve millions of people around the world. Today, we’re announcing Universal-1, our most powerful and accurate model to date, trained on 12.5M hours of multilingual audio data to help power the next generation of Speech AI products and features. Some key stats on Universal-1: • 72% preferred to our most recent model Conformer-2 in human evals • 71% better speaker count estimation and 14% better word timestamp estimation compared to our prior models • Up to 30% fewer hallucinations than seq2seq models like Whisper • Just 38 seconds to process 1 hour of audio Learn more about Universal-1 on our blog: https://lnkd.in/e5inQ-x9

💡 New Tutorial on our blog 💡 Learn how to quickly get started using Claude 3.5 Sonnet with audio data.

Microsoft's Florence-2 is a foundational image model that can perform almost every common task in computer vision and represents a significant step towards a unified vision model. In this guide, learn: What Florence-2 can do How Florence-2 works How to use Florence-2 What’s next for Florence-2 and Large Vision Models

Our Work-Bench NYC Enterprise Operators Retreat is one of our most anticipated events of the year. Why? We bring together 150+ operators to connect on building from 0 to 1. Our first speaker of the day, Dylan Fox, Founder and CEO of speech AI model AssemblyAI, walked us through the tactical takeaways he’s learned now running a company with $115M in fundraising, 100+ employees, and 5000+ customers. We recapped his top 10 pieces of advice for founders (or operators marinating on founding a company in the future). See a few below and more in our blog post linked in comments. 🤑 On fundraising: Articulate what you’re building clearly. Dylan sees a lot of Seed fundraising decks that clearly explain the market, but generally lack one critical element: a description of what the actual product is and does. YC's application is designed specifically to get founders to articulate this messaging clearly. So, he recommends even if you're not going to apply, to look at their application questions as a forcing function to describe what you're building, fluff free. Create urgency with a compelling “why now” event. Dylan doesn’t believe it's effective narrative of “I need cash and here is information about my company.” Investors want to be dazzled with creativity, imagination, potential, and opportunity. To meet that demand, try to be specific on why they should invest, but more importantly, why they should invest now. Why is this specific moment in time important? Did you just sign up a slew of customers? Did you just reach product-market fit? Are you seeing a big increase in top of funnel? 🚀 On scaling: Take time to define your company values/operating principles early on. Having this info at the ready will be a competitive advantage to talent scouting out different companies. Leaders you hire are super important. While it's tempting to recruit executive leaders from some of the world’s biggest tech companies (think Google, Meta, etc.), they often don’t have the hustle and startup experience needed to lead at an earlier stage. Make sure each hire has experience around your current stage - these are the people most well-equipped to handle the challenges that will be thrown at you as you scale. Invest in GTM before you need it. Most early hires will be builders (developers, engineers, product) to get the company product off the ground. Oftentimes marketing is put on the backburner in the earliest days, however, it's nearly impossible to scale to $25M+ in ARR with no marketing and on word-of-mouth alone. So make sure to get your marketing function set up before you need it, so they can hit the ground running when it’s time.

If you’re an application developer, taking the time to understand and address security concerns helps you protect your users' data, and become a trusted partner. In this article, we explore the top foundational security questions to consider for your next project using speech, including: - Have I accounted for defense in depth while accounting for risk? - Does the API provider adhere to industry standard frameworks? - How much transparency is provided in code-level controls? And more. Read the full article here: https://buff.ly/3Wv0vpt

💡 New YouTube Video 💡 Since GPT-4 was announced, we’ve seen many AI tools released to support both our everyday tasks and also our professional lives. In this video, we discuss the best AI tools now available specifically for software engineers, for tasks such as coding, testing, and UI building. Watch here: https://buff.ly/3Ww8Ub1

We've added new language support for PII Text Redaction and expanded our Entity Detection model, giving you more power and control in protecting sensitive information. What's New: - PII Text Redaction is now available in 47 additional languages - 16 new entity types were added to Entity Detection, for a total of 44 types available Learn more about these updates here:

💡 New Tutorial on our blog 💡 Accurately translated speech into different languages can be vital for accessibility and successful communication. Follow along in this detailed tutorial to learn how to translate speech in real-time in JavaScript using AssemblyAI and DeepL.

✨ AssemblyAI achieves the lowest Word Error Rate among automatic speech recognition (ASR) providers, according to 3Play Media's latest State of ASR report, which explores how ASR engines perform and underlines the importance of looking at diverse evaluation metrics when choosing speech-to-text technology. View the complete report here: https://lnkd.in/eMNVyn6M

💡 Video Tutorial 💡 In this step-by-step guide, learn how to create subtitles that dynamically change color based on the speaker using AssemblyAI's Speaker Diarization model and Python. Watch here: https://buff.ly/3zBGiVK