XR demands AI that understands spatial context, responds to natural language, and processes visual input with near-zero latency. Reka's efficient models are built for exactly this — on the headset, not in the cloud.
[USE CASES]
Real-Time Scene Understanding for AR Overlays
Recognise objects, surfaces, and context in milliseconds so overlays anchor where they should — and update when the world moves.
Multimodal Assistants Inside the Headset
Voice in, vision in, spatial output. Users ask about what they are looking at and get answers grounded in the actual scene — without leaving the device.
Industrial Inspection & Maintenance AR
Technicians point at equipment and get guided procedures, fault diagnosis, and part recognition overlaid in real time. The most commercially mature XR use case — and the one Reka is ready for now.
Training Simulation with AI-Driven Feedback
Trainees perform tasks in XR; the model watches the same scene they do and gives feedback grounded in what actually happened — not pre-scripted branches.
[HOW IT WORKS]
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 20 20" xmlns="http://www.w3.org/2000/svg"><path d="M 7.807 0.01 L 7.811 0.011 C 11.653 0.217 14.76 3.118 14.761 6.875 L 14.761 7.476 L 16.532 10.043 L 16.54 10.054 L 16.539 10.054 L 16.542 10.058 L 16.543 10.06 L 16.605 10.173 L 16.61 10.182 C 16.758 10.496 16.751 10.867 16.59 11.181 C 16.405 11.54 16.043 11.767 15.645 11.767 L 14.761 11.767 L 14.761 13.359 C 14.76 14.231 14.073 14.95 13.218 14.95 L 10.524 14.95 L 10.524 16.715 L 9.321 16.715 L 9.321 13.716 L 13.218 13.716 C 13.399 13.716 13.557 13.561 13.558 13.359 L 13.558 10.533 L 15.397 10.533 L 13.742 8.136 L 13.735 8.126 L 13.733 8.122 L 13.733 8.121 C 13.62 7.942 13.558 7.734 13.558 7.519 L 13.558 6.877 C 13.557 4.163 10.994 1.436 7.743 1.244 L 7.743 1.243 L 7.425 1.234 C 5.871 1.227 4.403 1.801 3.256 2.87 L 3.028 3.094 L 3.027 3.093 C 1.851 4.304 1.203 5.918 1.203 7.64 C 1.203 9.506 1.967 10.965 3.336 12.154 L 3.624 12.395 L 3.625 12.395 L 3.729 12.484 L 3.925 12.679 C 4.349 13.152 4.593 13.78 4.593 14.428 L 4.593 16.716 L 3.39 16.716 L 3.39 14.428 C 3.39 14.062 3.249 13.723 3 13.472 L 2.884 13.367 L 2.552 13.09 C 0.934 11.679 0 9.89 0 7.639 C 0 5.59 0.774 3.664 2.18 2.219 L 2.182 2.217 L 2.445 1.958 L 2.448 1.955 C 3.819 0.677 5.578 -0.01 7.437 0 L 7.439 0 Z M 5.918 9.228 L 6.498 9.828 L 6.599 9.933 L 6.498 10.036 L 5.276 11.301 L 5.169 11.412 L 5.061 11.301 L 4.481 10.7 L 4.38 10.596 L 4.481 10.492 L 5.702 9.228 L 5.811 9.116 Z M 4.61 8.122 L 2.581 8.122 L 2.581 6.974 L 4.61 6.974 Z M 5.276 3.798 L 6.498 5.063 L 6.599 5.166 L 6.498 5.271 L 5.918 5.871 L 5.811 5.983 L 5.702 5.871 L 4.481 4.607 L 4.38 4.503 L 4.481 4.399 L 5.061 3.798 L 5.169 3.687 Z M 11.165 3.796 L 11.745 4.397 L 11.846 4.501 L 11.745 4.605 L 10.523 5.869 L 10.416 5.981 L 10.308 5.869 L 9.728 5.269 L 9.627 5.164 L 9.728 5.061 L 10.949 3.796 L 11.058 3.685 Z M 8.67 4.752 L 7.559 4.752 L 7.559 2.657 L 8.67 2.657 Z" fill="rgb(46, 47, 47)" height="16.715915838627694px" id="hAWnflC58" transform="translate(1.641 1.641)" width="16.716183878636212px"/></svg>)
Deploy to the Headset
Reka runs on XR-class compute — the silicon already inside modern headsets, smart glasses, and ruggedised industrial wearables.
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 20 20" xmlns="http://www.w3.org/2000/svg"><g d="M 0 15 L 0 0 L 20 0 L 20 15 Z M 3.192 14.562 L 1.877 14.562 C 1.149 14.562 0.56 13.979 0.56 13.259 L 0.56 1.787 C 0.56 1.067 1.149 0.484 1.877 0.484 L 18.211 0.484 C 18.939 0.484 19.528 1.067 19.528 1.787 L 19.528 12.816 C 19.528 13.782 18.737 14.562 17.763 14.562 L 5.302 14.562 M 2.556 11.956 L 17.575 11.956 M 16.368 10.876 L 16.368 13.036 M 8.669 8.958 L 11.376 8.958 C 11.95 8.958 12.417 8.496 12.417 7.927 L 12.417 7.567 L 13.538 8.121 C 13.883 8.293 14.292 8.044 14.292 7.66 L 14.292 4.691 C 14.292 4.309 13.885 4.06 13.538 4.23 L 12.417 4.783 L 12.417 4.423 C 12.417 3.855 11.95 3.393 11.376 3.393 L 7.837 3.393 C 7.262 3.393 6.796 3.855 6.796 4.423 L 6.796 9.127" fill="transparent" height="15px" id="s5gsFdzeB" transform="translate(0 2.5)" width="20px"><path d="M 0 15 L 0 0 L 20 0 L 20 15 Z" fill="transparent" height="15px" id="XOcpGfEC4" width="20px"/><path d="M 2.632 14.079 L 1.317 14.079 C 0.59 14.079 0 13.495 0 12.775 L 0 1.303 C 0 0.583 0.59 0 1.317 0 L 17.651 0 C 18.379 0 18.968 0.583 18.968 1.303 L 18.968 12.332 C 18.968 13.298 18.177 14.079 17.204 14.079 L 4.742 14.079" fill="transparent" height="14.0787465px" id="H1gCgEfX5" stroke-dasharray="" stroke-linecap="butt" stroke-linejoin="miter" stroke-miterlimit="10" stroke-width="1" stroke="rgb(46, 47, 47)" transform="translate(0.56 0.484)" width="18.968360000000004px"/><path d="M 0 0 L 15.019 0" fill="transparent" height="1px" id="XiA4Evs9v" stroke-dasharray="" stroke-linecap="butt" stroke-linejoin="miter" stroke-miterlimit="10" stroke-width="1" stroke="rgb(46, 47, 47)" transform="translate(2.556 11.956)" width="15.019115px"/><path d="M 0 0 L 0 2.16" fill="transparent" height="2.16px" id="JGZv1rPIA" stroke-dasharray="" stroke-linecap="butt" stroke-linejoin="miter" stroke-miterlimit="10" stroke-width="1" stroke="rgb(46, 47, 47)" transform="translate(16.368 10.876)" width="1px"/><path d="M 1.873 5.564 L 4.58 5.564 C 5.154 5.564 5.621 5.103 5.621 4.534 L 5.621 4.174 L 6.742 4.728 C 7.087 4.9 7.496 4.65 7.496 4.266 L 7.496 1.298 C 7.496 0.916 7.089 0.666 6.742 0.836 L 5.621 1.39 L 5.621 1.03 C 5.621 0.462 5.154 0 4.58 0 L 1.041 0 C 0.466 0 0 0.462 0 1.03 L 0 5.734" fill="transparent" height="5.73415px" id="rhoe_HpRS" stroke-dasharray="" stroke-linecap="butt" stroke-linejoin="miter" stroke-miterlimit="10" stroke-width="1" stroke="rgb(46, 47, 47)" transform="translate(6.796 3.393)" width="7.495599999999996px"/></g></svg>)
Stream From the Onboard Cameras
Headset cameras feed the model directly. No external capture, no upload — what the user sees, the model sees, in real time.
Combine Voice, Vision, and Spatial Context
A user asks a question; the model answers based on what is in front of them, where they are looking, and what they just said.
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 20 20" xmlns="http://www.w3.org/2000/svg"><path d="M 6.62 0.737 C 8.554 -0.278 10.824 -0.244 12.734 0.838 L 17.687 3.643 L 17.751 3.68 L 17.751 3.684 C 18.626 4.214 19.168 5.232 19.168 6.362 C 19.168 7.447 18.668 8.429 17.853 8.975 L 17.687 9.078 L 12.734 11.885 C 12.372 12.09 11.996 12.256 11.614 12.385 L 11.465 12.435 L 11.465 11.255 L 11.543 11.226 C 11.797 11.129 12.048 11.011 12.289 10.874 L 17.24 8.068 L 17.348 8.002 C 17.866 7.657 18.171 7.055 18.171 6.362 C 18.171 5.668 17.866 5.065 17.348 4.72 L 17.24 4.655 L 12.289 1.849 C 11.443 1.37 10.513 1.129 9.584 1.129 C 8.771 1.129 7.958 1.313 7.201 1.68 L 6.88 1.849 L 1.928 4.654 C 1.344 4.985 0.997 5.622 0.997 6.361 C 0.997 7.1 1.345 7.735 1.928 8.067 L 6.88 10.873 L 7.082 10.983 C 7.286 11.087 7.494 11.178 7.705 11.255 L 7.786 11.284 L 7.786 12.458 L 7.638 12.41 C 7.329 12.311 7.024 12.188 6.728 12.04 L 6.435 11.884 L 1.482 9.077 C 0.569 8.559 0 7.518 0 6.361 C 0 5.203 0.569 4.161 1.482 3.643 L 6.435 0.838 Z M 9.585 4.044 C 10.793 4.045 11.775 5.025 11.775 6.23 C 11.775 7.435 10.793 8.413 9.585 8.413 C 8.377 8.413 7.393 7.435 7.393 6.23 C 7.393 5.024 8.377 4.044 9.585 4.044 Z M 9.585 5.102 C 8.96 5.102 8.452 5.608 8.452 6.23 C 8.452 6.851 8.96 7.357 9.585 7.357 C 10.209 7.356 10.716 6.851 10.716 6.23 C 10.716 5.608 10.209 5.102 9.585 5.102 Z" fill="rgb(46, 47, 47)" height="12.458299324633307px" id="uksk_SryP" transform="translate(0.832 4.167)" width="19.167969px"/></svg>)
Render the Response In-World
Overlays, callouts, voice replies, guided steps — pushed back to the headset fast enough to feel like part of the environment.
[FOR DEVELOPERS]
Start building in minutes
Integrate multimodal AI into your applications with our developer-friendly APIs, comprehensive documentation, and ready-to-use examples.
Powerful APIs
RESTful APIs with SDKs for Python, JavaScript, and more.
Comprehensive Docs
Detailed documentation, tutorials, and quickstart guides.
Sample Apps
Ready-to-use examples for common use cases.
Reka enabled Shutterstock to further monetize their massive multimedia library—500M photos and 50M videos—through automated metadata tagging.
“The quality of our content has long been the core of our business and we’re excited to leverage Reka’s AI capabilities to further enrich metadata and tagging of our content library.”
Paul Hennessy, CEO of Shutterstock
[GET STARTED]