Hello there!

For the majority of April, I’ve been slowly working on a backlog of videos for my new experimental pet project show about frog creatures called Bellychubbies. For that to work, I’ve been improving filming experience in BBS a lot lately, while neglecting any game engine related features. There are no people who are interested in creating game with BBS right now anyways, so I decided to improve machinima tools.

Here is first two videos (machinimas to be precise):

There is also short on the same channel. There are at least another 7 videos scheduled throughout May, so if you’re interested to see more, feel free to check the channel every 2-3 days.

Behind The Scenes

I’ve been working on a new update, which offers a lot of QoL features to the BBS’ filming department. These videos heavily use following techniques: audio-syncing, multiple-pass recording, game controller, and post-recording editing. If you’ve heard of Blockbuster mod (for Minecraft, I made it by the way), then you might have a general idea, if you don’t, then let’s get into it!

Audio-syncing

Before I can record the video, I need to record audio voice lines. BBS supports waveform preview when sound is playing, so it’s possible to synchronize head bopping to the dialogue. I use Audacity to record voice lines from macOS-powered TTS for narrator and Mozilla’s TTS for Bellychubbies voices (with some minor voice pitching). Then I chop and arrange recorded voice lines according to the script, and export it to BBS as a .wav.

Multiple-pass recording

For every Bellychubby on the screen, I recorded each twice. First time I record just the walking and turning around, and then the second time I use game controller to control eyes and head movement, so that I could focus only on expression during second pass.

Here is a short video demonstrating the multi-pass recording:

Game controller

In essence, BBS’ model system supports models exported from Blockbench with animations. Cool thing about Blockbench is that it supports math expressions, so at any particular keyframe, a variable, instead of a static value, can be inserted. This way, I was able to pass current joystick values directly to the model, and from there, these variables can be used in many ways, like controlling the eyes, the head movement, the torso, arms, etc.

Here is a short video demonstrating using left and right triggers for basic arm movement, left stick for eyes movement, and right stick for rotating torso along with head and arms:

Post-recording editing

After the scene was recorded, the only things that are left: setup camera movement and add any additional effects/animations. Camera movement can be added in BBS’ clip-based camera editor. Camera movement can be synchronized with the audio and scene’s actors by adding audio and scene clips, and from there different camera movement can be configured using a wide range of different camera clips. My preferred method is to first create an idle clip of the length of the entire scene, cut it based on the voice lines, and finally change camera angles and add camera animations where needed.

The second step is adding pose animations. This can be done by editing the player recording directly in the camera editor. Player recording have special blocks called actions, which execute whenever the recording reaches them. At the moment, there is a form action, which allow to change actor’s appearance (form) to another model or even posed model.

Here is a short video demonstrating basic form action editing to add animations:

Once everything is done, all left to do is to record the video. Fortunately, BBS also can do this as well using external ffmpeg binary (there are plans to rewrite it to a software ffmpeg using OpenCV).

I hope that interesting! Thank you for attention! 😊

P.S.: I apologize for any cringe you must’ve experienced by watching my videos. 👊😔