The Importance of Image Sequences
Practise safe rendering.
Rendering is a pain.
For anyone working in post production — visual effects, motion graphics or any sort of animation that involves computers, rendering is all too often a bane.
Merely the act of committing to the rendering of anything is an open invitation for Murphy’s Law — anything that can happen, will happen and with rendering this inherently “bad news bears”.
No matter how advanced software and hardware has become over the years, rendering is far from being any more reliable than it has ever been. As it has always been prone to error and mistakes, it is hard to imagine that ever changing. Rendering is just as relatively expensive and complicated despite these improvements because the bar of expectations in quality has continually raised over the years as a result.
It is not uncommon for an artist to be working late at night, on a project with a tight deadline and rendering somehow, somewhere, screws up, it is for the most part unavoidable and unpredictable.
Rendering is expensive. It is the heavy lifting of a processor to generate images based on the instructions of an artist. Sometimes this is done realtime, other times via a biased or unbiased renderer, where a delay exists before receiving usable results from the CPU or GPU to what is known as a frame-buffer.
The cost of rendering is mostly in terms of time but of course this equates to real financial cost, and potentially worse if deadlines are not met. Render times can vary wildly, which depends on the expense of the calculations needed to be done to generate images on the available hardware, which may or not be appropriate.
Sometimes the resulting images are not what were expected, desired or something went wrong while being generated. This could have been caused by anything from hardware failure, software crash or human error with an incorrect setting that may not have been caught, before committing. These problems are often frustratingly silly, where perhaps images were rendered in the wrong format, a lower quality or resolution than what was needed. There are any number of reasons that can cause a render to fail or be otherwise unusable.
Because of this, rendering to a single video file is a bad idea. While doing this is often seen as the most logical thing to do and therefore attractive, to render and export at the same time, remember that rendering itself is expensive. Should anything go wrong in the rendering process, or even a mistake made during the creation of the work, the resulting video file is for the most part worthless. The artist would then have to fix whatever mistake occurred, or ensure a failure does not repeat and render the entire video again.
This is also a problem with versioning of projects. For example, if the production is mostly interview footage and a request comes in to change the logo at the beginning or change something as simple as a colour, it would call for rendering again.
It only takes one frame to render incorrectly.
It is important to split the process of rendering and exporting, and remain careful not to conflate the two. Specifically the rendering to a sequence of images, one image per frame on disk and then later to compile this sequence as an export process to a single video file.
Rendering to a sequence immediately opens up a lot of possibilities. It enables an artist to target a specific problem, should one occur and render only that which is needed, previously missed, or anything that otherwise requires change. However while outputting to a sequence does not make rendering any less fallible, it does offer a lot more freedom for fixing anything wrong and crucially saving time. Rendering to a sequence also enables the ability for multiple processors in different computers on the same network to each contribute to the same sequence, further decreasing the overall render time, provided the same or compatible render engines are used and dependencies are met. This is rarely possible on a video file.
Exporting differs from rendering because it is the process of creating the master version of a production by compiling an existing rendered image sequence to a single video file. With the heavy lifting done this process is usually — and certainly should be — fast. Vastly slower if rendering would be included during this same stage. This process typically requires very little processing since the rendered media should be streaming from disk.
As this export process should be where the master version is produced, it is not advisable to compress it any more than what is considered ‘lossless’, where little to no visible reduction of quality is made to the resulting master file. It is important to be aware that significant compression at this stage is very likely to slow the export process. From this master, all subsequent deliverables are made, whether compressed or not.
Setting up an export is not difficult and typically involves adding an image sequence to an NLE timeline, along with any necessary audio. Further work could be done at this stage, such as trimming or colour correction and it would still be faster to be done here than at the rendering stage.
On a technical level a sequence is little more than a collection of numerically named image files on disk, along with padding in the numerical value within the file name, to ensure that the frames remain contiguous, meaning that they would appear in the correct order when viewed as a sequence.
Commonly used image render formats include .exr .dpx .tiff .tga or .sgi, with many of these having their own technical caveats, but popular for their “lossless” compression, if any is applied. In terms of a sequence they provide exactly the same function without any major difference in workflow when creating and compiling sequences.
Setting up a render of an image sequence requires little additional work from the artist, only to be aware that they are outputting individual images, one per frame of the sequence, in a folder with an appropriate naming convention and frame numbering, and it is ideal to keep different sequences in separate folders.
This workflow succeeds in production regardless of scope, when working with any professional grade graphics and animation software, to make the production more reliable and efficient. Feature productions have handled it this way for decades, even dating back to a time when film was the more dominant media in motion picture work. While it may seem that separating the rendering and exporting tasks is additional effort, it will save invaluable time should something inevitably go wrong.
Remember to practise safe rendering.