A single API for audio-driven LLM apps
LeMUR (Leveraging Large Language Models to Understand Recognized Speech) is a framework for applying Large Language Models to spoken data. In a few lines of code, you can do things like generate summaries or ask questions about your meetings, phone calls, videos, or podcasts.
Rapidly ship high-quality Generative AI features with voice data
- Unifies your AI stack for audioLeMUR is a single API that connects all of the spoken data in your application to an LLM to build generative features in your product. No need to chain multiple technologies together to go from audio file to LLM output.
- Optimized for specific tasksThe LeMUR API is optimized for high accuracy on specific tasks like generating answers to questions and writing custom summaries. The API leverages advanced retrieval and compression techniques to offer high-quality LLM responses.
- Faster time-to-marketThe LeMUR API enables you to find product-market fit faster. Rapidly ship new AI features, iterate on use cases, and enhance your development team’s productivity.
- Powered by AssemblyAI’s speech recognition modelsHigh-quality LLM outputs on audio data starts with quality transcriptions. LeMUR operates over AssemblyAI’s state-of-the-art speech recognition models, which leads to high-quality LLM outputs from your audio data.
- Continuously updated with the latest in researchWe are constantly experimenting with the latest research in LLMs and updating LeMUR with new techniques in retrieval, compression, prompt engineering, LLM performance, and more.
- Effortlessly scaleProcess over 100 hours of audio in a single API call. The LeMUR API handles over 1M tokens as input and is priced to scale as your audio data grows.
“LeMUR works incredibly well out-of-the-box. It allowed us to focus on product instead of infrastructure. As a result, we were able to bring a transformative new product to market in half the time.”
Optimized for high accuracy on specific tasks
Get answers to questions about your spoken data with LeMUR, ranging from asking about a customer’s history in a call center to asking for an explanation of a concept mentioned in a podcast. LeMUR can provide answers to questions that include citation and reasoning.
INPUT
What were the three most common issues customers called in with this week?
OUTPUT
Quality Control issue, Order Status issue, Billing issue
Automatically summarize virtual meetings, phone calls, and more in a flexible way with LeMUR’s Summarization endpoint. You can add additional context to provide information that is not explicitly referenced in the audio data being analyzed, like specific topics LeMUR should pay particular attention to when summarizing.
INPUT
Summarize key points from this product launch announcement in 2 bullets.
OUTPUT
• Apple announced Apple Vision Pro, augmented reality smart glasses.
• Vision Pro allows users to see digital content overlaid on the real world and interact using eye movements, taps, and voice.
Automatically generate a list of action items from virtual meetings with LeMUR. You can provide a specific format to follow and add context on the speakers to assign action items to specific meeting attendees.
INPUT
Recap action items from this meeting and assign them to attendees.
OUTPUT
**Metrics**
Christopher - Review and improve metrics around open bugs to accurately reflect issues within SLO, rather than just closed bugs. Work with Data team to implement.
**Database Replication**
Christopher - Review and improve metrics around open bugs to accurately reflect issues within SLO, rather than just closed bugs. Work with Data team to implement.
**Database Replication**
Steve - Provide update on postgres replication issue and mitigation plans in next Infra key review. Work with Craig and Data team to determine root cause and solutions.
**Customer Satisfaction**
Product Team - Continue focus on improving customer satisfaction (CSAT) and net promoter score (NPS). Monitor to ensure recent improvements are an actual trend.
LeMUR is built flexibly to allow you to define your own tasks and prompts with a customizable endpoint.
INPUT
You are a helpful coach. Provide an analysis of the transcript and offer areas to improve with exact quotes. Start with an overall summary then get into the examples with feedback. Under each example, place the corresponding feedback in a bulleted list.
OUTPUT
Overall, this seems like a productive meeting where the team discusses key metrics, areas for improvement, and next steps. The conversation flows well and multiple team members provide input. However, there are a few areas that could be improved:
**Clarifying Targets**
*"So if we were to look back historically here, the goal here was actually lower, and Christopher was ambitious and we kept raising and we kept meeting that. So it should stairstep here and we could go back and reconstruct that if we really wanted to."*
- The team seems to frequently change their targets for metrics like the MR rate. They should clarify their targets and the rationale behind changes to avoid confusion.</li>
- They propose "reconstructing" the target changes historically, which seems unnecessary. Going forward, any target changes should be clearly documented and shared to avoid having to reconstruct the history.
**Prioritizing Security Work**
*"We need to get more security work prioritized or hearing that from the team. But neither that problem nor that activity is sort of currently reflected in our security metrics."*
- The team recognizes that more security work needs to be prioritized, but their current metrics do not reflect this. They should consider adding metrics to track security work and ensure enough of it is being prioritized.
Try LeMUR today
See how LeMUR works through our Playground, where you can upload a file or enter a YouTube link to experiment with LeMUR endpoints in just a few clicks
Weekly product and accuracy improvements
LATEST UPDATES
- Pricing decreases
- Significant Summarization model speedups
- Introducing LeMUR, the easiest way to build LLM apps on spoken data