🐠01漫剧项目与流程@炮老师的小课堂

If you're used to ComfyUI being a canvas of latents and pixels, this pack will give you a jolt: the 🐠 漫剧 ("manga-drama") suite never generates an image or a video. It writes every piece of text that a vertical short-drama episode needs - script, visual setting, storyboard, image prompts, video prompts - so a separate tool like Seedance or MiniMax H3 can do the actual pixels. DapaoDramaProject is the config root of that text pipeline, the "advanced" branch's node 01.

Reach for it when you're building the 11-node advanced flow (or just want a project canvas that the guided three-node flow hides from you). It's the only node in the suite that stamps the project's identity: episode number, how many episodes you're planning, target seconds, production form, visual style, aspect ratio, prompt language, and the video-dialect you'll be generating into. Then it passes a self-describing DAPAO_DRAMA_BUNDLE down the chain. Nothing else works until this exists.

How it works - and why it's cheap

This node makes zero API calls. It's pure local validation plus a bundle constructor. Internally it builds the bundle (schema: dapao.drama/1) with your config, and it checks two rules before handing anything over: the current episode number can't exceed your planned total, and "⏳ 单镜最短秒数" can't be greater than "⌛ 单镜最长秒数". That's it - instant, free, and the last free thing you'll get, because every LLM stage after it costs money per call.

Here's the part that trips people: every document downstream is fingerprinted against this config (a SHA-256 of it, plus fingerprints of the upstream docs it was built from). Change the style or target seconds after your script and storyboard exist, and everything gets flagged stale. Downstream nodes refuse to run on stale material until you re-generate - and re-generating bills again. So think of this node as signing a contract you'll live with for the episode.

The inputs that actually matter

• 🎨 制作形态 - pick from six: 二维动态漫剧 (2D animated), 三维动画, 静态漫剧 (static), 水墨国漫 (ink), Q版动画, or 真人短剧 (live-action). This choice ripples: static form skips the video stage entirely later.
• 🎬 视频提示词方言 - 通用, Seedance 2.0, Seedance 2.5, or MiniMax H3. It only changes how the video-prompt node writes; the LLM doing the writing is chosen separately, on each stage node.
• 🖼️ 参考方式 - defaults to "挂图计划（自行出图）" (hang-image plan, you generate the images yourself). The suite will then use PLAN- identifiers that say "hang this image here later" - it never claims images exist. Switch to "明确文生视频" only if you really will run text-to-video.
• ⏳/⌛ single-shot min/max seconds - 0 means "model doesn't constrain me"; fill real numbers if your video model has limits. The storyboard node enforces them.
• Optional 📂 已有文档类型 + 📄 已有文档 - paste a finished script or visual setting to jump straight into a later stage without re-analyzing.

Installing

Same as the rest of the pack - it ships inside paolaoshi/ComfyUI-dapaoAPI:

cd ComfyUI/custom_nodes
git clone https://github.com/paolaoshi/ComfyUI-dapaoAPI
cd ComfyUI-dapaoAPI
pip install -r requirements.txt

Or use ComfyUI Manager and search "dapaoAPI". Then restart ComfyUI and hard-refresh the browser (Ctrl+F5) - the suite registers a new category menu and new front-end bits. No model files to download for this node; it's pure Python.

Gotchas

The pack's README never mentions this drama suite - it's a marketing changelog for the author's API image/video nodes. The real documentation lives in the repo's docs/short_drama_suite.md (Chinese) and the example_workflows/漫剧_*.json files. If you change the config after generating docs, expect "上游文档过期" (stale upstream) errors everywhere downstream; re-run from the stage you want to fix, and budget for the LLM calls that costs.