dramaclaw
▶ Play demoYouTube · loads on clickWhat they did
A script-to-finished-film engine for dramas, ads and product videos, built as one end-to-end pipeline.
Worth knowing
- Price: Open source (NOASSERTION)
- License: NOASSERTION
- Status: sourced. We have not run this ourselves yet (no "Tested" badge).
- Quality bar (computed by code): 4/5 (primary source ✓, reproducible ✗, new ✓, example output ✓, safe ✓)
The best new AI video workflows and prompts, weekly.
From the source
Setup. Self-hosted with Docker and docker compose on macOS, Windows (Docker Desktop + WSL2) or Linux; about 2 vCPU and 4 GB RAM recommended. No GPU is needed for the standard pipeline, because all inference goes through an OpenAI-compatible gateway (the official RelayClaw service or the bundled dramaclaw-gateway that routes to your own model providers). Web UI on port 8080, REST API on 8780. Licensed under Elastic License 2.0: commercial use is allowed with a small 'Powered by DramaClaw' credit, but not as a hosted SaaS.
Inputs. A manuscript or screenplay (Fountain supported). From there the Series pipeline builds episodes, characters and scenes, writes scripts, draws storyboards and first frames, and assembles voice-over, video and subtitles. The XiaHua canvas offers 18 node types for free-form image, video, audio and 3D-set work. A local MCP server exposes it to Claude Code, Codex and other MCP clients.
What the source says about results. The README links short dramas its team says it produced on this pipeline, and claims stable character identity across episodes through its asset library. Each stage is a separate resumable task.
Summarised from github.com/dramaclaw/dramaclaw, read 2026-10-11. Claims are the source's, not our tests.