Meta Launches NotebookLlama – An Open-source AI-powered Podcast Generator
- Meta just announced the release of NotebookLlama, an AI podcast generator that can create audio conversations out of PDFs.
- A lot of people are comparing Meta's NotebookLlama with Google's NotebookLM, saying that the latter is much better at creating AI conversations.
- The good thing is Meta has addressed the issues with its tool and is working on making things better.
Meta releases NotebookLlama, an open-source AI-powered podcast generator that can create audio conversations out of written documents. For now, you have two voices that can be used to make it sound like a genuine conversation between two people.
The tool was developed with the help of Meta's Llama model. Files related to the product have already been shared on GitHub - check it out here. As for the tool itself, it's already available for public download.
How Does It Work?The process is very simple - you just have to feed a text document, in a PDF format.
- Then the tool uses the Llama 3.2 1B model to convert it into a transcript.
- Then the Llama 3.1 70B model makes a podcast script out of it while the smaller Llama 3.1 8B model makes it more conversational.
- Lastly, Meta's Parler TTS tool converts the text to audio, adding all the dramatization, voice effects, and pauses to make it sound natural.
As soon as Meta announced NotebookLlama, a lot of people started comparing it to Google's NotebookLM, and rightfully so. It's not just the name that's similar. Both platforms have a podcast generation feature that works in a similar way. However, NotebookLlama is lagging behind a bit.
- For starters, the AI voices clearly sound robotic.
- Secondly, the flow of the conversation doesn't sound very human-like.
- There are odd pauses and sometimes the voices interrupt each other at the wrong time. A user on X shared a sample demonstrating the tool.
But the good thing is that Meta has acknowledged these shortcomings and assured that it's working on making things better.
Right now, it uses a single model to create the voices, which might make the back-and-forth sound unnatural. So it's planning to introduce different language models for different AI characters in the hope that it will make the flow much more natural.This isn't the first time Meta has tried to create a rival for NotebookLM. But most previous attempts have failed. NotebookLlama also has a long way to go.
The post Meta Launches NotebookLlama - An Open-source AI-powered Podcast Generator appeared first on Techreport.