Uploaded August 2026 | Updated September 2026, 2 weeks ago

---

In this video, I test a Live Vision-Language Model (VLM) running locally on the NVIDIA Jetson Orin Nano (8GB) using the live-vlm-webui from Jetson AI Lab.

I test real-time vision capabilities using Gemma 3 4B—the sweet spot for memory efficiency, speed, and accuracy on this hardware. I evaluate Scene Description, Object Detection, Activity Recognition, Safety Monitoring, and OCR text extraction using both an iPhone and a low-cost ESP32-S3 camera module via RTSP.

*Hardware & Software Specs
SBC: NVIDIA Jetson Orin Nano (8GB)
Model: Gemma 3:4B (Quantized)
Software: live-vlm-webui (Jetson AI Lab)
Input Devices: iPhone (Local Wi-Fi) / ESP32-S3 + OV03660 (RTSP)

[Jetson AI Lab - Live VLM WebUI]
jetson-ai-lab.com/tutorials/live-vlm-webui

[ESP32-S3 + OV3660]
waveshare.com/esp32-s3-cam-ov5640.htm?&aff_id=116255