Keishiki/TODO.md

# NOW
## Compositor
- Video!
  - Video import (research opencv libavcodec etc)
    - maybe https://superuser.com/a/1397578
  - Blender previews resource
    - assimp or call into blender executable?
- Data models
  - Dump and read back state, (de)serialization!!!
  - Undo's
  - Node groups

- Motion blur
  - Do multisampling, should be fast enough

- Text (idea: index-based evaluation in plugboard)
- Shapes ideas:
  - Backend: https://github.com/jdryg/vg-renderer
  - Frontend options:
    - AE-like interface
    - canvas-like interface
    - embed glisp maybe
      - inquire + still a lot of friction for simple shapes if we don't also get the glisp gizmos
- External data driving (csv, json or something else?) -- use a node to select source

- Pulses
  - use convolution with interps

## UI
- Copy & Paste
- Simple export

## Audio
- SDL_mixer can do all of wav ogg flac mp3 opus
  - don't care about output
    - maybe add option in ui for ffmpeg muxing via syscalls

# Later
## Compositor
- Investigate bgfx vulkan regression on 1109f3c5bf71f5e5c30fcaf6f899a25e3316adff
- Simple 3D engine
  - Particles -- shoot for performance in https://github.com/CHCTW/DirectX12-Framework-
- Flat sets and flat maps pending compiler support
- Random Idea: Array-eval nodes for instancing

## UI
- Adapt nodes for shader graph -- code editor will be fine for now
  - SRELL throws. Might be forced to switch to with PCRE or Re2.
- Blender control scheme
- Viewport tools & gizmos (Layer-specific & nodes -- can separate into different tools?)
  - Baku vec2 drag is good, color drag is not so necessary because imgui has a good picker
  - Layer types
    - Constrained AVLayer (find a different name for this LOL)
      - Transform gizmos + pan behind tool
    - Shape layer pending VG backend + figure out what to do forfrontend