In September, Google’s NotebookLM was able to turn users’ documents into a podcast, where two hosts talk through the material in a dialog format. Now Meta is introducing its own version of this feature, TechCrunch reports.
The tool, called NotebookLlama, will create a transcription from a file, article, or blog in a podcast format. After that, “more dramatization” and pauses will be added to the transcription, and then the finished script will be passed to speech synthesis models.
For this purpose, Meta uses its own language models, including Llama 3.2, Llama 3.1, and the open-source speech synthesis model Parler.
So far, NotebookLlama’s finished podcasts are inferior to what Google’s NotebookLM has produced. The generated voices sound like robots and speak over each other. However, Meta researchers say that the result could be improved by using more powerful models.
You can join the waiting list for the new tool on the official NotebookLlama website.
Loading comments …