Once every year, researchers from Microsoft's labs around the world gather in Redmond to show their latest projects. The event, TechFest, is a glimpse into the minds of the people responsible for pushing the bounds of technology inside the company. In some cases, it can also serve as a preview of future Microsoft products and features.

About 40 projects were on display for the media yesterday, out of more than 100 total this year, so we didn't get to see everything. There could be some more jaw-dropping stuff for employees behind closed doors. But out of the projects we saw, these were some of the highlights.

Automated Receptionist: Also known as "Situated Interaction" this project from researcher Dan Bohus explores the ways humans and machines can interact. A  lifelike on-screen avatar can help complete basic tasks, such as scheduling a shuttle.

The receptionist (pictured above) shifts her gaze when speaking with different people in a group, and tries to determine whether people are Microsoft employees or visitors depending on what they're wearing (more formal attire usually means a guest, casual means a worker).

The project has been shown publicly in the past, but new capabilities this year include the ability to host a trivia game. If one person keeps getting stumped, the avatar turns and suggests another person help.

Microsoft Research's Eric Horvitz says one possible future application would be next-generation elevators that could hold the door open when people are approaching, and tell them when they missed their floor.

Commute UX: This project aims to improve on existing in-car "infotainment" systems by making it easier for people to communicate with the system by voice and by touch.

For example, when someone tries to respond to a text message using voice commands in the car, the system uses a set of 15,000 common text replies to help determine what the driver is saying. Microsoft Research's Ivan Tashev demonstrated the system for me at right.

One novel application: A driver can call up specific pages in the owners' manual by voice commands and see the pages on the in-car display.

Writing in Air: This one got lots of attention as a possible way of interacting with the Xbox in the future. It uses a basic webcam to recognize characters and words people write in the air with their fingers or an object. Wilson Rothman at Gizmodo has a good video of this one.

NanoTouch: This evolution of Microsoft Research's Lucid Touch project aims to let people interact with devices by pressing and moving their fingers against the back. It's especially useful on small screens because using fingers on the front can block what people see. Researcher Patrick Baudisch cited potential applications including jewelry, letting people put a small screen on an object on a necklace and interact with it by pressing on the back.

SecondLight: One of the most crowded booths was for this project, which was first shown publicly last year. It adapts Microsoft's Surface tabletop computer to project images and detect gestures in mid-air above the screen. For example, you can hold a piece of tracing paper above the tabletop and see a different image than appears on the computer itself.

To get a sense for what it's like, see this video I took at Microsoft's Professional Developers Conference last year.

Real-time stitching of mobile-generated videos: This project takes video streams from multiple mobile phones and constructs a higher resolution video out of them. The technology could be used at public events, for example, where multiple people are capturing the same scene. Ina Fried of CNet News.com has a longer post and video of this project.