Low-latency access to SAM-3 via real-time WebSocket sessions. Pay only for the time you use — $0.00155/s, no subscriptions.
SAM-3 segmentation running on live camera data from a Unitree Go2, powered by the Vizion API.
GPU-accelerated inference with optimized model serving for real-time applications.
Get started in minutes with our Python SDK. A few lines of code is all you need for real-time inference.
$0.09/min of GPU time. Load any amount you need — no subscriptions or hidden fees.
Install the SDK and start segmenting in under a minute.
pip install git+https://github.com/CoreVisionX/vizion-sdk.gitfrom vizion import VizionClient
with VizionClient("vz_live_...") as client:
client.connect()
result = client.segment(jpeg_bytes, prompts=["person", "car"])
for det in result.results:
print(f"{det.prompt}: {len(det.instances)} found")
for inst in det.instances:
mask = inst.decode_mask() # numpy (H, W) bool arraySegment Anything Model 3 — state-of-the-art image segmentation
$0.00155/s · $0.09/min · $5.58/hr
Monocular depth estimation for any image
$0.00155/s · $0.09/min · $5.58/hr
Real-time object detection and tracking