How to use rendered files for faster export in Final Cut Pro?

This is an old problem that has existed for years. I have previously discussed it with Apple escalation support. The FCP caching system is complex and creates MD5 hashes of rendered frames/segments based on several things, including effect parameters and content. The internal name for render cache is "segment store."

When FCP needs to render something, it checks if caching should be used by generating a unique MD5 hash for the content, queries the segment store for cached data and decides whether to use it. The decision is complex may involve dependency chains where caching of effects can depend on other inputs to the decision process.

One non-intuitive case involves Title effects, if non-default text is used (which is the common case). If a Title effect is used with Neat Video (or I think certain other built-in effects), the timeline can be rendered to cache and the timeline will playback smoothly. But if exported, all clips with those effects will be recomputed, yet the cached result will not be used to expedite the export and encoding phase. So in this case the cache validation for playback is using a different algorithm than the cache validation for export.

I need to resume investigating this and discuss it with Apple, but it takes days of research to gather enough technical information. But I'll put it on my "to do" list.

This isn't a problem specific to Neat Video, but it often is noticed there because of the obvious difference between cached and non-cached performance.

Workaround: Apply Neat Video NR to the required clips at an earlier edit phase, export those to "bake in" the NR, import them and replace the timeline clips with them.

It happens with the latest Neat Video 6.01. However, that version is significantly faster, so this lessens the problem a bit.

I do use Neat Noise in FCP and it is very hard on the computer. I find the best workflow is to turn background rendering off. I always do that anyway as it interferes with everything else and is not easy to control. I then render any individual clips which have had Neat Noise applied. As long as you don’t make any changes, they should stay rendered. However,  another option would be to apply Neat Noise and then export them after rendering and re-import them into a final timeline. This should not be necessary but it is an option. 

There is a new version of Neat Noise which they claim is faster and more efficient. I just upgraded and it does seem to be quite a lot faster. Might be worth checking out if you use it a lot. 

Re an organic, experimental editing style, I get that. But why do you need Neat Video activated on much of an 80-minute timeline when you are doing active content editing? Can't you disable NV on those clips until the edit is locked, then enable it?

I realize due to current FCP design that can be cumbersome. It really needs a way to bulk disable (without changing the parameters) of a given effect. But doing that solves the immediate perf. issue, then when the edit is locked you can re-enable Neat Video on all clips having it.

I have been forced to do that many times on various productions and it is tedious. But it's better than constantly paying the performance cost as you crank out many evaluation exports during earlier phases of editing. And it avoids the hassle of exporting clips with baked-in NV, then importing and replacing them on the timeline.

Re concern about updating to Neat Video 6.x, that does not delete the prior version. I believe it stays on your system and the 5.x version stays applied to your clips. So ultimately you'd have to add the 6.x effect to each clip, then remove or disable the 5.x effect from each one. However you could create a test library with a few NV 5.x clips, then install 6.x to try that. It doesn't solve the caching problem but it renders about 2x faster.

The render cache issue needs further investigation but it is a deep problem and extremely time consuming to investigate. I will try to do that, but it will take a week of non-stop research before I even have enough info to discuss with Apple. I have some other "real work" to do, but at my first opportunity I'll start that process.

I did some unscientific experiments. Note that I did not use Neat Video, which I don't own, nor did I use any other plugins, but I just did create a test project with several layered, scaled clips with different blend modes, so that rendering could have some impact, or not.

I tried exporting it unrendered, and then I rendered everything, and exported again.

So long as I used Video and Audio, and the choice for Video Codec was "Source (ProRes 422)", then exporting the rendered version took a little less than 2/3 of the unrendered version. I tried this twice, just in case, and the results were consistent.

Exporting to H264, both exports (rendered and unrendered) took about the same time.

This seems to suggest that there is an advantage to rendering when exporting using the Source codec, but not when transcoding. I also noted that when exporting the rendered version, FCP tends to use less CPU than when exported the unrendered. (Note: I think that most of the heavy lifting is done in the media engine, not the CPU).