In this video I’ve added something I’ve been thinking of doing since the first one, which is to have little animations when a unit acts or a SOUL ability is being used. These appear as little animations on the bottom of each side acting.
The process to create these was fairly straightforward. First step was the place the unit art on a “green screen” background, since the model is not capable of generating transparent backgrounds. Then I used the open weights Wan2.2 model on ComfyUI, and prompted it to perform the animation that fits the act. Finally I added the finished result in kdenlive with a chroma key effect for green which allowed the unit to have a transparent background again.
Now the animations are not incredible as the model is not that capable of complex movements, especially for weird designs like these, but they give a little extra action on the side which doesn’t interfere with the main game-play.
I did try to make some animations involving multiple units (you’ll see like 3) by combining two units in the same image. But it was both very complex to arrange (I had to ask the model to first make the character face each other for example and then use the final frame from that as the next frame for the next), and also the result were pretty bad. Most of the time, one character wouldn’t animate at all, and the whole thing was very basic. So ultimately I scrapped that idea and stayed with a single unit. If a more capable open-weights model appears in the future I’ll give this another go.
This post will detail the creation steps, but first and without further ado, here’s my first battle report.
I made a quick Youtube channel just to host these, so as to keep things nicely separated. I wanted to post on pixelfed as well, but I couldn’t find a relevant instance for this content with open registrations. If you have one suggestion like this, lemme know.
Now that’s out of the way, lemme write down the process for making this video.
The first step is of course the recording. I bought a cheap table mount where I could put my smartphone on, and it’s a bit short but fortunately just enough for the Maleghast playmat I have. I also bought some Lavalier microphones but sadly those didn’t seem to work right this time so it ended up using my phone mic.
For recording software I used Open Camera which provides a lot of control on recording. Unfortunately since I didn’t quite know what I’m doing, I ended up recording is way too high bitrate, getting massive files. It also recorded in portrait. But nothing a quick ffmpeg command can’t fix:
Once I had my video recorded, I then had a good idea on how to edit it. The main idea is to display the units interacting and their stats, as well as the abilities they’re using and any conditions they have accumulated.
Initially I tried using G.I.M.P. to create the various templates I wanted, but manually tweaking all the elements was taking entirely too much time and it was also not looking that great. I decided to pivot and with some LLM help I had the initial skeleton for a python script which could compose a unit card from its component elements. Of course the LLM choked once the script became even mildly complex so after a point I just took over on my own.
The way it works is that I use kdenlive to go through the video. Every time a unit ability is activated, I set a guide at that timestamp. The name of the timestamp then contains all the info I need to generate the unit card at that point. The ability used, the current health, and any tokens assigned to that unit. Example: “bolides_3_vitality3+weak” will generate a Holy Body unit card with 3 health, and with 3x Vitality and 3x Weak condition tokens assigned.
At the same time I’m also trimming “thinking stretches” from the timeline, and generally trying to shorten the video if possible. For my first video I wasn’t very strict about this though, so it ended up 1.5 hours long. Given that it was a beginner game and we both had a few rules questions, I’m expecting this time to rapidly drop in newer videos. I’ve managed to play games in less than 1 hour already. I’m not quite sure if in future videos I will try to completely cut away all silence/thinking times as it allows the video to be a more chill-affair to have running in the background, which could be nice for ADHDers like myself. Tell me what you think.
Anyway, once I have all the Guides places, I used the very useful function of kdenlive to export all guides in this format “{{category}} - {{timecode}} - {{comment}}” and place them as a text file along with my script which then parses the guides to create all the images that I’ll use for the video and names them according to their timestamp. Then it’s just a matter of drag and dropping them into the timeline on the two tracks I added for this, adjusting their duration according to how long the action took, and finally adding the custom effects I created which made them fade in/out and place them on the left/right of the screen.
In python I’m utilizing badgepy to create badges for the tokens to attack to the unit cards. Simple, albeit not particularly pretty. I’m planning to add maybe some icons to the badges, but if you have some ideas on how to make this look better, send them my way.
Finally I go through the video at real-time to see if all looks correct. If any unit card is wrong (say, wrong HP or tokens), all I need to do is edit the guide at that timestamp, export guides and rerun the script and since kdenlive is loading the files dynamically from the hard disk, this is enough to update all the unit cards in my existing video. This will also mean that if in the future I come up with a different unit card design, it should be trivial to update everything, or even go back and update my old videos.
Whenever I notice mistakes in the play, I then use the subtitles track of kdenlive to quickly mention the mistake so that people watching know.
And that’s pretty much it. After that I only need to upload to youtube and figure out a thumbnail. For the first video I just used the Gemini GenAI as I couldn’t figure out a way to create an image containing the two primary necromancers from the factions and the images contained in the book where way out of my skill-set to shop into a good-looking image (and frankly I had had enough “cooking”, and I was just eager to publish). I’m looking how to do something similar using local diffusion models in the future, so when I do I’ll post that process here as well.
Hope you found this process useful, and let me know what you think of my first Battle Report!
Close
Ad-blocker not detected
Consider installing a browser extension that blocks ads and other malicious scripts in your browser to protect your privacy and security. Learn more.
