I've read all of documentation and the forum posts about the workflow of putting down keyframes to create smooth camera paths.
It seems there are two different approaches: zoom in & flight
Zoom in = It is difficult to use the 'Distance from last keyframe', as it doesn't seem to update correctly when clicking the rendered image. The number seems to fluctuate wildly. I have therefore found it difficult to create a smooth camera zoom animation. Maybe I understand it wrong.
Flight = I have found 'Distance from last keyframe' to be most useful when 'absolute distance mode' is turned on. But this proves useful only for flight animations. I still cannot achieve the smoothness of the example animation. It is also confusing that the 'Estimated viewpoint distance to the surface' number seems to grow instead of shrink.
So instead of getting a smooth camera path, my camera tends to speed up and then slow down for each keyframe.My aim is to create a smooth camera animation skin to the clip below. It is stunning in its magistic beauty. I would love to hear tips on how to archieve this. How many keyframes were required to make this? How far apart were the keyframes?
youtube.com/watch?v=bO9ugnn8DbEAny suggestions on workflow would be most appreciated! I saw a few other forum posts about this, but they did not go into depth on the exact process. How often are you laying down a keyframe? Willing to share an example? How do you make sure the distance between keyframes is consistent?
Absolutely brilliant software. I've been dreaming of this for years. It is such a joy. Especelly the automatic shadows and ambient occlusion.
On a seperate note
- What does 'Dynamic DE correction' actually do? Does it override 'DE step factor'?
- Found a consistent crash: If you accidentlly lay down a keyframe on frame 3, but keyframe 2 does not yet exist... then the software will crash. I've just learned to be cautious about it. It can happen when you decide to start clean slate and you've deleted all your keyframes, so your 'keyframe' number is whatever was selected last.
Just so we are all on the same page, here are some things I have collected from around the forum. This helped me to understand the deeper workings of the software... This assumes you have read the google website documentation.What is an average number for "Distance from last keyframe" for fluid flight?- It depends how close to the fractal surface you want to fly. Total size of Mandelbox is about 10. So if you go with step about 1.0, you will go through whole Mandelbox in about 10 keyframes. But if you want to fly between of some detailed structures very close to the fractal, step value 0.000001 should be to high (or to low). You have to feel it. There is no recipe for this. Flying speed can varying along the animation. If you the camera is far outside the fractal, you can move the camera by long distance. But during closing up the fractal surface you can decrease speed. You have to continuously adapt the flight speed upon the situation. It is unrealistic to always move the camera with the same speed.
- If you use mouse clicking on the rendered image and the "Mouse click close-up ratio" is set to 5, then you should create one keyframe for each "click".
- If you use arrows and Forward/Backward buttons, you have to observe "Estimated viewpoint distance to the surface" and "Distance form last keyframe" values. If you want to create smooth flight you should try to keep for each keyframe similar values for "Distance from last keyframe" - you will get almost constant flight speed. But if you want to make "zoom in" animation, you should observe distance to fractal surface. You shouldn't decrease distance more than 5 times per keyframe.
Suggested settings for an animation camera- Step for camera moving multiplied by DE: from 0.5 to 3.0
- Absolute distance mode: checked for flight animation, unchecked for closing up
- Absolute movement distance: depends on required flight speed and how close to fractal surface the camera is
- Rotation step in degrees: from 5 to 30
- Mouse click close-up ratio: from 3 to 5
- Enable zoom by mouse click: checked
- Go close to indicated surface: usually unchecked, but it is useful for flight-around animations
- Rotation without using gamma angle: checked
Optimizing render times- If "DE error" is high (above 1%) it means that raymarching algorithm is not accurate enough and you will observe noise in the render. When it is lower than 0.5%, the noise is almost invisible on images, especially where there is a lot of details. Always try to play with DE Step factor, because it has the biggest influence on rendering speed. You can use LQ (Low quality) or HQ (high quality) buttons to find "optimal" value of DE factor (it uses DE error value to find optimal DE factor). Then you can tune DE factor manually (check quality of image, not the "DE error").
- Try not to change Max. iteration. In most cases it has no influence on rendering speed. The difference is only visible when you are rendering some cross-sections (using limits). Normally "rays" never reach high iteration regions, because may-marching threshold is based on the distance (not on maximum number of iterations). You can watch it in histograms window. On left side there is number of iterations histogram (scale is up to 64 iterations). For mandelbulbs and mandelboxes average number of iterations is between 5 and 20 (peak on histogram is on the left or in the middle).
- About volumetric light it increases "DE error" value, because calculating of light rays uses not accurate ray-marching algorithm (for faster rendering). Fractal details will be still accurate. If image looks good, just ignore this. To get faster rendering of volumetric light you can decrease quality of this effect (default is 5) for instance to 1.0. rendering will be 5 times faster but there will be more noise on rays. On high resolution image noise is less visible.
- About the formulas, the fastest is "Tglad's folmula Mandelbox". The slowest are Xenodreambuie's formula and all hybrids. One remark about hybrids. When in hybrid sequence the most of iterations are Mandelbox or some kind of IFS formulas always use "Linear DE Mode". Then ray-marching is much more accurate and you don't have to reduce DE Step factor (rendering is faster). Sometimes you have to try both modes and compare the result.
Slowest things to render- volumetic light
- not DE shading mode - uses special algorithm for calculating normal vectors. It's very slow but surface is always smooth.
- ambient occlusion based on rays (not the one from Post effects tab) - speed depends on quality parameter - higher value = higher number of rays = much slower). Sometimes quality = 2 is enough.
- additional light sources. Rendering time = normal time * number of light sources (sometimes faster when lights are not so bright - there is some optimization)