MJL Posted October 7, 2011 Posted October 7, 2011 why would I want to do 1, or 2 or 4 or 7 passes? Quote
John Bigboote Posted October 7, 2011 Posted October 7, 2011 It is a different renderer... with different strengths, and weaknesses. For instance, if you need a nice fast test render... use 1 pass. The more passes you add will 'comb' thru the antialiasing better and better...you can see it getting better. Compare the standard A:M renderer to a 16 pass MP. The standard renderer does not render your displacement maps- and I don't think it can make use of the FakeAO plug-in, but it can be a fast renderer too and in certain situations be faster than MP. They both do motion-blur completely different too...and DOF. Quote
MJL Posted October 7, 2011 Author Posted October 7, 2011 Thanks, John. I thought it was something like that. BTW be on the look out for when I post this latest clip, it is right down your alley. Quote
John Bigboote Posted October 7, 2011 Posted October 7, 2011 I have an alley? I think I know what you mean...MJ. Watch for mine too... Quote
John Bigboote Posted October 7, 2011 Posted October 7, 2011 OH--- and another feature that only works with multipass is the hair feature called 'jitter'... Quote
Hash Fellow robcat2075 Posted October 7, 2011 Hash Fellow Posted October 7, 2011 Multi-pass is... many passes! for 16 passes, for example, the camera renders the image at 16 points on a 4x4 grid within the space of one pixel and averages them together. This is to reduce aliasing. Because it does this for all pixels it is slower than the regular renderer which is smart enough to know which pixels need full anti-aliasing or not. A patch of blue sky need no antiialiasing. MP also can show tru motion blur because it can render the object as it moves thru space at slightly different positions within the time of one frame. Regualr renderer calcultes teh direction and velocity of a pixel a just the instant of the frame and adds a quick blur in that direction to simulate motion blur. It is very fast compared to MP. because MP can do more than 16 passes it has the potential for better Anti-aliasing than the regular render if you dont' mind the time expense. And for some, reason some features work only in MP. Not sure why. Quote
Admin Rodney Posted October 7, 2011 Admin Posted October 7, 2011 The Tech Reference should have a good write up (with pictures!) that states the following: Multi-Pass Rendering By default, “Final”, render preset uses an A-buffer renderer. The A-buffer incurs some initial overhead for edge quality and adaptive computations, but generates images much faster than its "Multi-Pass" render preset equivalent, and can be efficiently used with global illumination solutions like radiosity or skylight-rigs. “Multi-Pass” rendering means that multiple render passes of the complete scene will be individually computed and averaged with each other. Multi-Pass is used for selected scenes where the light and camera jittering that happens on each pass can improve the quality of the shading. The number of passes determine the level of antialiasing: 4X4, or 16 passes, is the minimum acceptable if a “Blur” post effect is not applied. Advantages of A-Buffer Rendering 1) Much faster for the same Multi-Pass equivalent of antialiasing. 2) Practical for global illumination, like radiosity or light rigs. 3) Better bitmap fidelity. Advantages of Multi-Pass Rendering 1) Light jittering adds some global illumination, including penumbral shadows. (The A-buffer has “penumbral” klieg lights for a similar effect, and light rigs also provide this look). 2) Physically correct motion blur, (A-buffer motion blur is adequate in most cases). 3) Camera jittering blurs crisp geometry. Camera jittering can also be used for physically correct depth-of-field if enough passes (64) are used, (A-buffer depth-of-field is adequate in most cases). Methods to Take Advantage of Multi-Pass Rendering Since the entire scene is being rendered multiple times (1 time per pass) in Multi-Pass rendering, it is important to take as much advantage of this computationally expensive rendering method as much as possible. Here are a few tricks to giving you a higher quality rendered image at no extra cost. 1. Motion Blur - Use the motion blur option when rendering animations. With each rendered pass, time will be advanced slightly based on the motion blur "Percent" property. The Motion Blur percent actually represents the camera's shutter and symbolizes how long the shutter is open exposing the image during rendered. The default motion blur is 20%. It may take a substantial number of passes to remove the strobe looking motion blur causes by inadequate number of passes. 2. Depth of Field - Focusing the camera on objects of importance and allowing the foreground and background objects to be out of focus can be a very desired filming technique. Multi-Pass rendering depth of field adds no render time expense, but the same applies true with inadequate number of passes using depth of field as does with motion blur. With each pass the camera is slightly moved to render the scene from a different angle causing near and far objects to become blurry. To render your scene using depth of field, turn on the Depth of Field render settings on the camera or in the Options Dialog Rendering tab. When the Depth of Field property is set to "ON", the camera will display the Depth of Field manipulators when selected. Use the Focus Distance manipulator to set the focal point, and the Near and Far Focus manipulators to set where the out of focus starts. 3. Light Width - When multi-pass rendering the light width can be used to simulate area lighting, global illumination, and penumbral shadows. For each pass the light will jitter inside the area defined by the light's width property giving a different lighting angle. The benefits from this are endless. Shadows will have crisp edges closer to the object casting them and fall off as the distance increases. Surfaces will be lit from multiple angles giving more accurate specular highlights. 4. Soft Antialiasing - The multi-pass render can be set up to have a different averaging mechanism used for combining each pass. This averaging technique slightly softens the image but gives the appearance of smoother antialiasing especially at lower passes. To invoke this softening effect use the Blur post effect on the camera with a softness value of 0. Note any softness value larger than 0 will invoke a true blur on the rendered image. Blur Post Effect See Post Effects for information on how to use them. Blur Softness Use a softness of 0 to only invoke the Multi-Pass softening pass averaging technique. There is more but this is from the 2003 write-up that tried to capture many of the benefits of Multipass rendering. The takeaway is that once you turn Multipass on there are quite a few things you can get in a rendering without it costing you more in time rendering frames. Quote
MJL Posted October 7, 2011 Author Posted October 7, 2011 Thanks Robert, Once again, I kinda thought that was the way it is. I know about some features (bump maps, etc.) don't work without multi-pass. As I get deeper into using this wonderful software, I want my understanding to hopefully grow as well. So I thought I'd ask. Multi-pass 1, small resolution= trial renders>short render times, Multipass 16?, whatever resolution = final product>much longer render time (much better-product) EDIT: Thanks again to you , Sir Rodney. You posted as I was typing. I'd read that tech manual more often if I could get my eyes to quit glazing over about halfway down the page. Quote
Admin Rodney Posted October 7, 2011 Admin Posted October 7, 2011 Martin wrote this concerning Multipass (and pretty much any post effects added to a render during or after the rendering) and it's worth noting: However, it does produce blurred images - that's why it isn't always on. In the case of animation, where motion blur is desirable, the directional nature of this blurring is positive, but static elements will be blurred. Almost all post-processing is image blurring: bloom, film-grain, depth-of-field, lens flares, motion blur, HDRI, etc., etc., etc. People like blur, that's the reason there's so much interest in separating out the render elements (diffuse, specular, mirrors, etc.) - so they can be degraded to varying degrees before recompositing. That's the art. Martin Hash Quote
Admin Rodney Posted October 7, 2011 Admin Posted October 7, 2011 Multi-pass 1, small resolution= trial renders>short render times, I'm trying to find a reference but haven't found it yet. It refers to some effects that don't kick in until 4x4 passes is set. I see a minor reference to that above but it isn't the quote I was after. But... that information is why I started to always use at least 4 passes set if I turn on Multipass. Quote
NancyGormezano Posted October 8, 2011 Posted October 8, 2011 (edited) I know about some features (bump maps, etc.) don't work without multi-pass. Bump maps in general work with both multipass and without multipass. There had been a specific situation (ver 14) where bump maps wouldn't show up if decals & materials were on the same surface and one was rendering without multipass. I am not aware if this is still the case, nor can I remember exactly what the exact conditions were. It might also have been related to using ray traced, or z buffered kliegs. I'm trying to find a reference but haven't found it yet. It refers to some effects that don't kick in until 4x4 passes is set. The option to Soften = ON/OFF kicks in at 5 pass rendering Here's a comparison chart that I did for myself awhile back to evaluate tradeoffs of increasing # passes, using soften feature, and rendering without multipass. My feeling is that there are render situations where 16 pass is overkill. In other situations, maybe not. (rendered in ver 16.0, original frame was 960x540 at 16pass soften=ON) Edited October 8, 2011 by NancyGormezano Quote
Hash Fellow robcat2075 Posted October 8, 2011 Hash Fellow Posted October 8, 2011 My own personal observation is that "soften" is a very slight effect. It seems to be most useful for improving the look of tight thin parallel lines such as one might see on radiator fins. My other advice is always go with regular render first for final work, it's faster and looks as good as 16 passes. Multi-pass for those one-pass tests and when some odd item doesn't exist in regualr render like EXR buffers or displacement maps. Quote
Admin Rodney Posted October 8, 2011 Admin Posted October 8, 2011 My own personal observation is that "soften" is a very slight effect. It seems to be most useful for improving the look of tight thin parallel lines such as one might see on radiator fins. I believe that is where this statement comes into play: To invoke this softening effect use the Blur post effect on the camera with a softness value of 0. Note any softness value larger than 0 will invoke a true blur on the rendered image. Usually people don't think to use the Blur Post Effect along with the Softness settings. Not unlike Softening when used by itself, without Multipass rendering I find that Blur alone isn't very effective. They seem to need each other to pull off ideal blurring effects. Quote
Fuchur Posted October 8, 2011 Posted October 8, 2011 I know about some features (bump maps, etc.) don't work without multi-pass. Bump maps in general work with both multipass and without multipass. There had been a specific situation (ver 14) where bump maps wouldn't show up if decals & materials were on the same surface and one was rendering without multipass. I am not aware if this is still the case, nor can I remember exactly what the exact conditions were. It might also have been related to using ray traced, or z buffered kliegs. I'm trying to find a reference but haven't found it yet. It refers to some effects that don't kick in until 4x4 passes is set. The option to Soften = ON/OFF kicks in at 5 pass rendering Here's a comparison chart that I did for myself awhile back to evaluate tradeoffs of increasing # passes, using soften feature, and rendering without multipass. My feeling is that there are render situations where 16 pass is overkill. In other situations, maybe not. (rendered in ver 16.0, original frame was 960x540 at 16pass soften=ON) My personal experience is, that for Animations you can go with less passes while still-images need higher pass-values. A fast animation with 3 passes can look good enough because people dont see still images and dont have too much time to worry about some edge-problems while on a still images they have it and will very likely have a problem with less then 9 passes (depends on the displayed motives and the size of the image of course... Thanks Nancy! Nice comparission... gives a good idea of the renderhit! See you *Fuchur* Quote
John Bigboote Posted October 8, 2011 Posted October 8, 2011 Yes, Nancy's test is very illuminating... note the moire in the texture on her thigh(and noise in general on the upper legs) that disappears at 16+ passes... and is not there at all on the 'no pass' image... but there IS a lot of texture noise... Fucher is right about the lower passes for animation... I do supervised animation sessions where the client looks over my shoulder while I work. I explain that the initial sequences from A:M(quick renders) are 'low-res' and will be replaced with a higher pass count later on for better anti-aliasing... most of the time, the clients say- it looks fine just the way it is. Good discussion! I need to try that post-effect trick from the reference manual... set to zero, interesting! (below) 4. Soft Antialiasing - The multi-pass render can be set up to have a different averaging mechanism used for combining each pass. This averaging technique slightly softens the image but gives the appearance of smoother antialiasing especially at lower passes. To invoke this softening effect use the Blur post effect on the camera with a softness value of 0. Note any softness value larger than 0 will invoke a true blur on the rendered image. Blur Post Effect See Post Effects for information on how to use them. Blur Softness Use a softness of 0 to only invoke the Multi-Pass softening pass averaging technique. Quote
Hash Fellow robcat2075 Posted October 8, 2011 Hash Fellow Posted October 8, 2011 To invoke this softening effect use the Blur post effect on the camera with a softness value of 0. Note any softness value larger than 0 will invoke a true blur on the rendered image. I think this may be obsolete information or referring to something else. The current multipass "soften " is an ON/OFF setting with no value and does show up in the render without a post effect being used. Quote
NancyGormezano Posted October 8, 2011 Posted October 8, 2011 (edited) My other advice is always go with regular render first for final work, it's faster and looks as good as 16 passes. I have found that in general, the best compromise for me (quality versus render time), and in some cases, the better choice is 5 pass/soft=ON, for animation and as well as for stills. Render times for 5 pass/soft seem to equal NO pass. In my case, however: I tend to favor high contrast, high frequency patterns, textures. So I'm always looking for a balance between showing the pattern "noise" and blurring the pattern noise to reduce scintillation, stair stepping caused by movement, that tends to be most noticeable if lines become horizontal/vertical across image. Sometimes I like it to scintillate, as I'm a glitter type of gal. Decals get anti-aliased differently than hair textures, as well as differently than bitmap plus materials or procedural materials with noise or just the geometry edges. If one doesn't have textures, then 3 pass might be good enough, fastest solution for animation. Noise gets introduced by AO and FakeAO as well, looks good in stills, sometimes not as good with movement. I only use klieg's z-buffered shadows, usually with Global ambient intensity (color or Image based) for faster render times. Sun or bulb illumination, ray traced shadows will look different as well, have different kinds of noise. Blurring or softening or extra passes or pixel averaging are about getting rid of the noise, but then compromises sharp detail. In these two images, for example: I prefer how 5 pass looks on the hair (image emitters for both girl and lion) compared to the NO pass. I also prefer the 5 pass face texture (decal) on girl compared to 16 pass - gets washed out, yet her leg texture looks best on the 16 pass (less noise). The lions body texture (bitmap plus material) seems to look the same in all 3, so go with 5 pass. Antialiasing on whisker edges looks better in 5 pass versus NO pass. EDIT: changed images to show differences Edited October 8, 2011 by NancyGormezano Quote
*A:M User* Shelton Posted October 8, 2011 *A:M User* Posted October 8, 2011 Not to steal the topic, but I love those characters, Nancy. They are so much fun and the set and lighting is wonderful. Ok back to multi pass rendering. Steve Quote
NancyGormezano Posted October 8, 2011 Posted October 8, 2011 Not to steal the topic, but I love those characters, Nancy. They are so much fun and the set and lighting is wonderful. Thanks Steve Quote
Vertexspline Posted October 8, 2011 Posted October 8, 2011 Steve---my thoughts exactly looking at these ---great job Nancy ---so cute a character pair indeed!! and great lighting for sure. Plus the test was illuminating as well....and I think indeed for animation you can often get away with less passes as all that motion one does not have a chance to really focus on the subject so easily. Thanks for doing this ---so we all can decide for ourselves what makes sense. Quote
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.