I have a Raspberry Pi 5 running BirdNET and a mic thrown out into my garden. It’s been incredible finding out who my noisy neighbours are.

The species IDs are used to make prompts for the MiniT2I model, which generates the portraits on my DGX Spark.

Big shoutout to Cornell’s K. Lisa Yang Center for Conservation Bioacoustics for making this technology available, and to Xianbang Wang and co. for their great work on text-to-image models.