← Back to Work

Research entry

AnimStudio — AI Animation Pipeline

2026 · Independent Project

An early-stage pipeline for producing 2D anime-style YouTube content from plain-text screenplays — parsing character descriptions, generating storyboards, synthesising voice, applying lip sync, and assembling a final MP4.

Python AI ComfyUI ElevenLabs FFmpeg uv

Open live app / source →

Overview

AnimStudio is an experiment in fully automated 2D animation production. The input is a plain-text screenplay with character descriptions and dialogue. The output is a rendered MP4 with generated visuals, synthesised voice, and lip sync applied.

The motivation: producing even a short animated video manually is an enormous amount of work. Most AI video tools operate at the clip level and require significant human intervention between steps. This project attempts to wire the full pipeline end-to-end so that the only creative input required is the script.

Pipeline

Screenplay (plain text)
  → Parser        — extract scenes, characters, dialogue, emotion tags
  → Character Refs — generate reference sheet images per character
  → Storyboard    — generate scene images from descriptions
  → Animation     — generate animation clips (ComfyUI)
  → Voice         — synthesise dialogue per character (ElevenLabs)
  → Lip Sync      — apply lip sync to character clips
  → Assembly      — combine clips, audio, and transitions into final MP4

Screenplay Format

Screenplays are written in a plain-text format with structured headers:

TITLE: My Show - Episode 1

CHARACTER: ALEX
  Description: Young woman, short blue hair, anime style, brown eyes, red jacket
  Voice: <ElevenLabs voice ID>

SCENE 1: INT. COFFEE SHOP - DAY
  Description: A cozy anime-style coffee shop with warm lighting

  ALEX: (excited) I can't believe we finally made it!
  [Alex jumps up from her chair]
  ALEX: This is going to be amazing.

Emotion tags in parentheses (excited, angry, sad, whisper, laughing, etc.) drive ElevenLabs voice synthesis parameters per line of dialogue.

CLI

Each pipeline stage is independently runnable:

CommandDescription
animstudio parse <script>Parse screenplay to JSON
animstudio characters <script>Generate character reference sheets
animstudio storyboard <script>Generate storyboard images
animstudio animate <script>Generate animation clips
animstudio voice <script>Synthesise voice audio
animstudio lipsync <script>Apply lip sync
animstudio assemble <script>Assemble final video
animstudio run <script>Run full pipeline end-to-end

Status

Early stage — the pipeline scaffold and CLI are in place. Active development is focused on getting a full end-to-end run producing a coherent short scene before expanding to multi-scene episodes.