Tool textures - Valve Developer Community Solid to players and NPCs but not other objects. In Left 4 Dead and Left 4 Dead 2 this clips players as both special infected and survivors. Autotune tpain preset.

Contents • • • • • • • • • • • • • • • • • Introduction Many people choose to capture their own homes as their first photogrammetry projects - without realising that modern, interior architectural scenes can be some of the hardest to scan well. Featureless, blank painted surfaces, reflections, specularity, the exacting perfection needed to make clean, manufactured surfaces not look distorted and misshapen - even the best source photography will fail to capture everything. If you're looking for an initial project, try something lumpy, organic and full of detail - paradoxically, some of the things hardest to model and texture conventionally are the easiest to capture. The tutorial should help you there, and should also act as a good starting point for some of the techniques used here - should you be willing to take the plunge into something much trickier, this tutorial will document the scan and cleanup of the Valve Lobby scene available in SteamVR Home. Capture The camera I used was again a Canon EOS 7D, this time with the EF-S 17-55mm f/2.8 lens at 17mm.

To ensure maximum image sharpness and consistency, a tripod was used - I started with a manual exposure of f/11, 1.6 seconds at ISO 100. Bada apps store. Image stabilisation was switched off, to stop the lens getting confused into attempting to correct non-existent wobbles. Both the zoom ring and focus ring were taped into position to prevent any accidental shifts - the zoom ring at its widest, and the lens set to manual focus at an approximate of 1m for maximum depth of field. (Keeping the focus the same across shots also avoids any changes due to.) I took just over 600 photos, with the camera in portrait orientation - making it easier to capture floor-to-ceiling shots.

For various areas (the large metal valve assembly in the centre of the lobby in particular) I took shots from different heights - looking from below, above and in the middle, orbiting around objects to get decent coverage from all angles. Using a wide angle lens helped ensure at least some coverage for everything - with such a large scene as this, it is easy to accidentally miss things. Try not to be too clever about how you capture - it's better to take too many photos than too few. If in doubt, always take more - especially when transitioning between different areas. Example photo in Lightroom, showing most of the development settings. I took a fair number of shots from outside the publicly accessible areas, looking back in through the glass - these proved fairly tricky to align relative to the rest of the bulk of the photos due to reflections from the glass boundaries and limited shared coverage. As with the outdoors scenes, the camera was set to shoot in RAW format, thus providing a bit more dynamic range to work with and the ability to further tweak shots in Lightroom - processing there was limited to setting a good white balance, tweaking exposure, pulling in shadows and highlights and correcting chromatic aberration and lens vignetting.

The same develop settings were applied to all images for uniformity, with the only shot-specific adjustments being to correct for an accidental change in exposure around half-way through - the command wheel on the back of the camera got nudged to 2 seconds. This was easily corrected out using Lightroom's 'Settings: Match Total Exposures' function - and will be prevented in future with the use of the little command wheel 'Lock' switch on the back of the camera. (So that's what it's for!) Processed photos were then exported to 8-bit, full-sized TIFFs - minimally compressed JPEGs would also have worked, but I wanted to make sure I'd get the absolute best detail in the low-contrast, near-featureless painted walls and ceilings. Tip: Much of the advice in this article should apply to other 3D modelling and photogrammetry software as well. The main feature required is the ability to re-import cleaned up meshes into the photogrammetry software for reprojecting new textures - Agisoft PhotoScan can do this too. Getting a good camera alignment was initially quite difficult - limited shots in a few places meant I had to add control points to various photos in order to align separate components together.

After getting around 550 of the 600-odd photos aligned, I set it to build a high-detail mesh - this, unsurprisingly, took a fair amount of time. Having a decent GPU, CPU and plenty of memory certainly helps things along. I ended up with a mesh approaching 400 million triangles - definitely too much to render in VR - then decimated it down to three million. While a bit higher than I'd normally need, I expected to trim quite a lot from this scene, but still wanted detailed objects to retain a fair triangle count. Rough geometry enclosing the scene, which I deleted first.