• ThePrompt
  • Posts
  • AI stack for audio and speech data

AI stack for audio and speech data

PLUS: StackOverflow AI, AI job at Netflix pays $900K, Shopify Sidekick and more

Hi folks!👋🏻 This is The Prompt! We're like your favorite coffee place - warm, cozy and always serving the best shots of AI!

Let's get it

FEATURED

LeMUR: LLMs for Speech and Audio

LeMUR, is an “AI Stack” for building LLM apps with spoken data.

You can take your meetings, phone calls, videos, podcasts, and more and build LLM apps with just a few lines of code.

LeMUR is built on top of their speech recognition model called Conformer-2 which is trained on more than 1.1M hours of spoken data, and in theory, it should give better results than Whisper.

Whisper is trained on 680K hours.

Pricing

Looking at their pricing page, we can see that for 60 min audio file (that’s about 15k tokens) and an output size of 1500 words (2000 tokens), the price would be $0.353.

Whisper is cheaper, but it can’t ingest a 60min audio file at once.

How to use it

They have curated a Prompt library, where you’ll find best practices and prompt examples on how to use it.

Try the demo here.

🚨 What else is going on

📕 Resources

🧰 Tools of the trade

  • Rewind: Capture everything from your phone/mac and search it with AI

  • Insumo: ADHD brain planner

  • HuddleUp: AI-powered peer recognition in Slack

  • Contember: From concept to web app in minutes

  • AI logo art: Transform your logo into AI Art

✍🏼 Prompt of the Day

TOOL

Midjourney

PROMPT

Oscar winning special effects and cinematic action aerial from above photography still from an upcoming live action cinematic sci fi blockbuster,an ultra futuristic utopian young woman in a sleek flying craft on a bright sunny future utopian city street background aerial view of futopian future city, intense special effects portrait --s 250 --ar 16:9

RESULT

Subscribe to keep reading

This content is free, but you must be subscribed to ThePrompt to continue reading.

Already a subscriber?Sign In.Not now