Good evening everyone. It's been a couple of weeks since the last update. Time for another.
Previously I had a 15 minute DayZ bullshittery cut ready for animation.
I should be on 10 minutes complete out of 15.
I'm currently 11 minutes and 20 seconds complete.
Ahead of schedule and making good progress.
Heavily summarising the animation work:
Despite the predicted keyframing difficulty, editing speed is being helped by long segments of conversations without subjects in front of me in the game. For example, there's a 1:39 minute segment with 6-8 people in Teamspeak - but none of them were actually in the same room in Chernarus at the time. That simplifies keyframing rather dramatically.
Speed has also been helped by some efficiency improving innovations. Detailed below.
I would however like to rework the final 2 minutes of the video. I feel it can be improved. So will make an attempt at a funnier cut before attempting a full keyframe.
Let me tell you about some exciting workflow changes I've been using these last two weeks.
Special thanks to Duckslayer on this front.
Recently I've been aware of the fact that my editing process is very 'set in stone'. And I keep seeing new things appear around me in After Effects. But they're outside of my understanding and I need to set aside some training time.
So being mindful of that, I sought to try and solve a very niche problem that was consuming too much time. And wondering if better tools were now available.
When I start any given bullshittery I have to take the stored livestreams and sort them. These are kept as .mp4 files. And are usually saved alongside .txt files called "notes", from an OBS plugin called INFOWRITER.
These text files, written on the same day as the stream, are past-me telling curren- me where the jokes are:
From there, each stream moves from the "UNNESTED" folder to the main root folder full of footage. Meaning I've "nested" the clips in Adobe Premiere and they're available for me to make a compilation.
It also means I won't accidently waste time trying to make new nests (clips) from the same stream.
But some streams, usually a single digit number, enter a folder I call "UNNESTED AND UBSCRUBBED". Meaning there was something amusing in there. But the notes got lost, were accidently deleted, or were far too vague to make sense of.
In these instances I'm reluctant to discard them because I know there's something of worth. I just don't know where exactly.
My prior solution was to simply render the whole stream into a small .mp3 file and then speed it up by 20%. Putting it on in the background when cleaning around the house, walking the dog, etc.
But that is a disproportionately large amount of time being spent on what are usually a small number of livestreams. And it often falls to the wayside as the editing gets harder.
Duckslayer presented a solution.
He's given me a Python script thing and a tutorial on how to use it. And it can automatically sift through large .mp4 files in a folder and generate a simple .txt transcipt file of the dialogue being spoken. With a JSON file for the timestamps.
It's not perfect, but it doesn't need to be.
It just gives me a much stronger idea of the moments where we're probably bantering. Letting me snap to specific areas in the stream and narrow down where everything is. Alongside also letting me find supporting setup for jokes previously logged.
And from that I had a minor brainwave.
As a normal part of my process (since 2017) when cutting together a bullshittery I have to take each clip I've "nested" in Adobe Premiere and export it into AE. So this segment I've cut from a 2 hour stream? It looks like this with no text. Just some clips in Premiere. Trimmed from 2 hours to 27 seconds:
But then after right clicking it > "Replace with After Effects composition", it turns into this below.
It's just a copy of the .aep composition you're then editing in After Effects on your other monitor. Any changes you make in there, are reflected on your main timeline.
Unfortunately though, After Effect's sound mixing is absolute ass. It's not designed for it at all. So it smushes it all together in a way that becomes extremely confusing if you're trying to sync sounds and dialogue.
My solution has always been to just tell the Adobe Media encoder to churn out 40 ish .mp3 sound-only versions of every single clip in the bullshittery. Then just discard all of After Effect's rubbish attempts at playing the audio and tell it to read it straight from the .mp3 I'm giving it.
They're extremely tiny, fast to open and in perfectly sync with the dialogue being spoken. So I copy in my text templates and get straight to work. Typing out what is being spoken and when.
But the brainwave was...
"if I'm churning out 40 mp3 files anyway then why not churn out 40 .mp4 files at the same time? Then feed those through Duckslayers new transcribing tool to see the text?"
That produces 40 small transcript text files with a rough approximation of the spoken dialogue.
From there, I knew that people can import subtitles into their video projects with Adobe. Documentaries have been doing it for years.
So I found a simple script that would convert the contents of text files into text layers.
One sentence at a time.
A little messy and not perfect of course.
But it doesn't need to be. An approximation is enough.
From there I'm able to listen to identify which sentence belongs to which speaker. Delete the ones I do not need.
Then use another script to smush the multiple text layers into 2-3 dedicated layers, using my font templates (Dillinia UPC in Bold). Where the "Source Layer" for the text changes on a new sentence, preserving all font settings and remaining parented to the Nulls (invisible squares about to be keyframed).
And voila, what I've just described is the normal form of all my After Effects compositions.
But this time, all the text has been auto generated.
That is a sizeable improvement in setup efficiency.
Granted, the true meat of the editing isn't this. Keyframing takes hours. Masking still takes hours. And I then have to spent quite some time chopping up the sentences into a structure that flows far more naturally under the eyeline. A process that still involves rigorous rewatching and refining. I'm only describing the start of the process.
But keystroke-to-keystroke it's much faster to modify, correct and tweak existing text then it is to type it all from scratch.
So thank you very much Duckslayer. He's created a significant, perhaps even double digit percentage improvement in my editing efficiency. And one that will improve all future edits going forward.
So right now, there's one more week of expected keyframing. And either next Friday or sooner, I'll be crashing into the quality assurance/rework phase for final delivery.
Video expected early July. Thank you all.