diff --git a/new_specie/README.md b/new_specie/README.md
index 681579813a274eeb35e8cdfe74f085d02b983791..70bb5169361610fae4b6ac88f51deda438ad1ad2 100755
--- a/new_specie/README.md
+++ b/new_specie/README.md
@@ -9,6 +9,7 @@ Vocalisation detection needs to be done prior to this process (stored in a .csv
 If you want to train your own auto-encoder (optional since the generic one might suffice), use `python train_AE.py detections.csv`
 
 Use trained auto-encoder weights to project your vocalisations with `python compute_embeddings.py generic_embedder.weights detections.csv`
+The architecture for the generic_embedder.weights is with a bottleneck of size 256
 
 Visualise vocalisation embeddings and resulting clusters with `python sort_cluster.py embeddings.npy detections.csv`