diff --git a/nvidia/spark-reachy-photo-booth/README.md b/nvidia/spark-reachy-photo-booth/README.md
index d104fca..94d1e9a 100644
--- a/nvidia/spark-reachy-photo-booth/README.md
+++ b/nvidia/spark-reachy-photo-booth/README.md
@@ -20,7 +20,7 @@
 
 ## Basic idea
 
-![Teaser](images/teaser.jpg)
+![Teaser](assets/teaser.jpg)
 
 Spark & Reachy Photo Booth is an interactive and event-driven photo booth demo that combines the **DGX Spark™** with the **Reachy Mini** robot to create an engaging multimodal AI experience. The system showcases:
 
@@ -33,7 +33,7 @@ Spark & Reachy Photo Booth is an interactive and event-driven photo booth demo t
 
 The demo is based on a several services that communicate through a message bus.
 
-![Architecture diagram](images/architecture-diagram.png)
+![Architecture diagram](assets/architecture-diagram.png)
 
 > [!NOTE]
 > This playbook applies to both the Reachy Mini and Reachy Mini Lite robots. For simplicity, we’ll refer to the robot as Reachy throughout this playbook.
@@ -149,7 +149,7 @@ Run the following command to make sure the Reachy speaker can reach the maximum
 ./robot-controller-service/scripts/speaker_setup.sh
 ```
 
-![Setup](images/setup.jpg)
+![Setup](assets/setup.jpg)
 
 ## Step 4. Start the stack
 
diff --git a/nvidia/spark-reachy-photo-booth/assets/architecture-diagram.png b/nvidia/spark-reachy-photo-booth/assets/architecture-diagram.png
new file mode 100644
index 0000000..b1594cb
Binary files /dev/null and b/nvidia/spark-reachy-photo-booth/assets/architecture-diagram.png differ
diff --git a/nvidia/spark-reachy-photo-booth/assets/setup.jpg b/nvidia/spark-reachy-photo-booth/assets/setup.jpg
new file mode 100644
index 0000000..fac3108
Binary files /dev/null and b/nvidia/spark-reachy-photo-booth/assets/setup.jpg differ
diff --git a/nvidia/spark-reachy-photo-booth/assets/teaser.jpg b/nvidia/spark-reachy-photo-booth/assets/teaser.jpg
new file mode 100644
index 0000000..efd4414
Binary files /dev/null and b/nvidia/spark-reachy-photo-booth/assets/teaser.jpg differ
diff --git a/nvidia/unsloth/assets/test_unsloth.py b/nvidia/unsloth/assets/test_unsloth.py
index a11367d..74bfc6e 100644
--- a/nvidia/unsloth/assets/test_unsloth.py
+++ b/nvidia/unsloth/assets/test_unsloth.py
@@ -46,7 +46,7 @@ fourbit_models = [
 ] # More models at https://huggingface.co/unsloth
 
 model, tokenizer = FastModel.from_pretrained(
-    model_name = "unsloth/gemma-3-4B-it",
+    model_name = "unsloth/Phi-3.5-mini-instruct",
     max_seq_length = 2048, # Choose any for long context!
     load_in_4bit = True,  # 4 bit quantization to reduce memory
     load_in_8bit = False, # [NEW!] A bit more accurate, uses 2x memory