oofoe

Annnnnd... It's done. Final delivery of the final shots for "Lost Hallway" happened yesterday at noon. I burned the last two shots (479 and 130 frames respectively) to DVD (the portable hard drive had crapped out) and sent it off with the director to be Avid-ivated. I saw the semi final cut on Saturday with the final music and it looks good! It's a five minute short about lost umbrellas, lost dogs and lost innocence. I want to write up a detailed post-mortem later (so I don't forget the results of my hard won experience! (and because I'm narcissistic)), but for now just a few thoughts... In total, I was responsible for about forty-five shots ranging in length from 90 to 1242 frames. (In retrospect, I was an *eedeeott* for taking on that much, but it seemed like a good idea at the time.) CGI was rendered to 1/2 or 3/4 HD, outputting to full HD. Pipeline was: Sony HDCAM 60i from the camera, converted to 8bpp TIFF (we had enough data management issues already, thanks... 16 bit was right out) Virtual sets and prop replacement in Animation Master v15, rendering to OpenEXR format for RGBA and depth, PNG for matte passes. Matting and compositing with Vision (worked great with AM:15 using OpenEXR! See my other posts about depth buffers), output to TIFF. Final edit with Avid. [*] Data: 49GB incoming footage, 62GB total layers (renders, etc), 46GB outgoing footage. 900MB model, texture, animation and composite data. [*] Most regretted statement: "Oh, I can just put a garbage matte around that!" AM 15 was a real champ. It performed very well on my main machine (I often had AM, Vision, PhotoShop, Xemacs, FireFox and Thunderbird all churning at the same time!) with no crashes! The models were not terribly sophisticated, but they were *big*. Each standard hallway segment was sixty feet long (built to scale with the greenscreen set). The set for one shot was a hallway over a thousand feet long. The hallway sections were lit with model lights which worked really well -- except in the real-time preview (DirectX /OpenGL limit you to eight lights in a scene -- since each hallway section had twelve lights and I often had multiple hallway sections, I usually got to see black). I had big plans for ambient occlusion, shadows and reflections, however, due to time and hardware constraints, most shots were rendered without them. My one point of reference for most of the greenscreen shots was the doorknob that we had bolted to the greenscreen stage door. I modeled that precisely in AM and built the surrounding door based on my measurements from the stage. Best part about AM (aside from great workflow and easy modelling) was being able to reference and seamlessly import models, choreographies and light rigs. Worst part: Who do I have to kill to get network rendering back!?! ;-) Big huge props to Jason Young and Carlton Munday. Jason rendered shots for me in ENGLAND and figured out how to get the frames to me! Carlton did some rendering, and more importantly, got me a loan of a fast Shuttle PC for rendering locally with. Couldn't have done it without you guys! And a very special thanks to Martin and crew who put out some great software that let me take this project off the kitchen table and right into the editing suite. If anybody has any questions, I'll be happy to answer after I've had some sleep... I don't think I've slept more than an hour a day for the last three weeks, so please forgive any disconnectedness in this post. Thanks. P.S. Are there any Toronto Hash users interested in attending the first screening? If so, PM me -- if there's any interest, I'll see if we can get you in!

Hiya, What, I didn't say "compositing" often enough? ;-) I'm using a program called "Vision", which is the baby brother to Fusion, designed for video work. We use Fusion at my office for feature film work so I'm very familiar with it already. It comes with several video keyer nodes (and a completely useless demo version of Primatte), but none of them will do the full key right out of the box. It's more a "build it yourself chroma key", which works pretty well if you know what you're doing. Less well if you don't know what to do with the pieces (me). Currently, I'm blurring the grain out, then pulling a hard/soft key of each actress which I use to cut them out of the background. I was using a ColourCorrect node to do spill suppress, but it couldn't deal with the harsh green, so I'm using a nifty trick described by Steve Wright to "synthesize" a new green channel based on a combination of red/blue. Much better I think... (See attachment.) Thanks for the suggestions!

Hiya, I'm going to need to do something... Not sure what. The shots are way overlit with lots of green spill and glare. What you see there is my current "best effort" at getting the stuff out. Obviously not good enough... ;-) Vision is a nice program, but I haven't done green screen work in years, so I'm spending a huge amount of time re-learning to deal with it. At work, we use nicely rendered 3D elements that don't need to be keyed... Thanks!

Hi, I just wanted to drop a quick note about a short film production I'm currently working on with a buddy (I'm waiting on a render). The film is called "Lost Hallway". I'm handling CG and compositing for it. It's a very green-screen heavy job with a very tight deadline. I'm using Animation Master 15 for the CGI and Eyeon Vision for keying and compositing. The pipeline is Sony HD video camera at 60i to Avid, then export of each shot to individual TIFF files, which are then scaled down by Vision to make rotoscope images for loading into AM15. Modeling and rendering is done in AM, and animated to match the actresses (we used green screen tracking markers to get references for the pseudo-environment during camera moves). The frames are rendered to OpenEXR format with depth channels. These are merged with the keyed talent to make the final image. Grain, defocus and blurred reflections are done in composite. Being able to get the OpenEXR output from AM lets me do some nice 2D comp tricks that save a lot of time on the render side. Production scripting is handled by XEmacs, Python and NewLisp. Since both Animation Master and Vision use ASCII file formats, it's pretty easy to move things around as needed. I'm rendering to 1440x810, then scaling up to full 1920x1080 for the final. There are no shadows and only two levels of reflection. The target render time for the most complex shot is no more than 5 minutes per frame. I'm attaching one of the (hopefully) final images. It's not overly spectacular, but hopefully it gives you an idea of the sort of thing I'm doing (and the director approved it for release... ;-). I don't know if we'll make it to anything big, but hopefully this film will show up at a festival near you!