EchoAI is building an AI-powered conversation intelligence platform that turns customer conversations into powerful insights.

Learn more about Echo AI's story and why they chose to partner with AssemblyAI here:

Google NEXT recap:

Earlier this month, many members of the AssemblyAI team traveled to Vegas for Google NEXT. We had the opportunity to connect with customers and other innovative companies who are pushing the boundaries of what’s possible.

Some highlights include:
- Our CEO

Universal-1 model is our next-gen multilingual ASR model, beating all other tested models in WER.

Here's a collection of resources to get started easily:

Research:
🟠Blog post: assemblyai.com/research/unive…
🟠Paper: arxiv.org/abs/2404.09841

Coding guides with our SDKs:
🔵Python:

Self-serve invoices and dual-channel improvement:

🔵Users of our API can now view and download their self-serve invoices in their dashboards
🔵We’ve made readability improvements to the formatting of utterances for dual-channel transcription

assemblyai.com/changelog

Anatomy of Industrial Scale Multilingual ASR

The technical report about our Universal-1 multilingual ASR model is out!

Universal-1 is our latest ASR model in production, designed for high-quality, high-throughput, and large-scale operations.

arxiv.org/abs/2404.09841

AssemblyAI is building the best API platform for developers to transform and understand voice data with AI so that you can build amazing new products and services for the world to use!

Get started today 👇
youtu.be/5Uw-r36XQYk

We're excited to share that AssemblyAI made the Forbes AI 50 list, which recognizes the most promising AI startup companies🎉

It’s an honor to be listed alongside other innovative companies that are pushing the boundaries of what's possible with AI!

forbes.com/lists/ai50/

Our very own CEO Dylan Fox was a podcast guest on the Product Market Fit Show.

Learn about the early days of AssemblyAI and how Dylan came up with the idea, went through YC, grew it to 5,000 customers, and raised $115M in venture capital:

buzzsprout.com/1889238/148220…

If you're at #GoogleCloudNext , join our VP of Technology talking about how we leveraged JAX and TPUs to create our latest Universal-1 model.

Together with cohere and NVIDIA we share what we're doing within the JAX ecosystem.

g.co/cloudnext cloud.withgoogle.com/next/session-l…

Nice to see our Universal-1 model as 'the clear leader in the Word Error Rate benchmark'

Introducing Universal-1, our most powerful speech recognition model to date.

Trained on over 12.5 million hours of multilingual audio data, Universal-1 achieves best-in-class speech-to-text accuracy across English, Spanish, French, and German.

assemblyai.com/research/unive…

LLMs work wonders on text data but if you want to use audio or video files instead, things get a bit trickier.

In this video, we’ll learn how to build a RAG application in 10 minutes that can take multiple speakers into account when answering a question.

Consider a panel discussion, where multiple people are having a conversation.

❓ If we ask, 'What is each speaker's opinion?' the RAG application that only uses a plain transcription can only provide a general opinion of the group, while the RAG application that takes speaker

Some of these features would have been unimaginable even a couple of years ago, yet with AI, you can now:

✍️ write a script for YouTube videos,
👤 remove the background of your videos,
🔈 cancel the background noise,
🌠 create images for your thumbnails,
🤖 ideate for

🤔 LLMs work wonders on text data but what about audio data? Transcribing audio recordings could be a solution but you might lose valuable information on multi-speaker files.

🚀 You can overcome this problem by using speaker diarization with the AssemblyAI & Haystack

🗣️ Unlock the full potential of your recordings with LLMs and Speaker Diarization!

🤝 Learn how to use AssemblyAI and Haystack together to enhance your RAG application by detecting multiple speakers in audio recordings, and providing a transcript that attributes each