mirror of
https://github.com/Zippland/NanoComic.git
synced 2026-01-19 01:21:08 +08:00
92 lines
3.6 KiB
Markdown
92 lines
3.6 KiB
Markdown
# NanoComic 漫画分镜生成器 / storyboard generator
|
||
|
||
一个面向漫画创作的分镜生成项目：自动调研并输出逐格脚本。
|
||
Storyboarding for comics: auto research and frame-by-frame scripts.
|
||
|
||
|
||
## 功能概览 / Feature highlights
|
||
- 多步研究：生成查询 → 并行搜索 → 反思补充 → 分镜 JSON。
|
||
Multi-step research: queries → parallel search → reflection → JSON storyboard.
|
||
|
||
- 分镜预览：前端逐页渲染，并为每页触发图生成（需模型权限）。
|
||
Page cards on frontend; each card triggers image gen (requires model access).
|
||
|
||
- 模型可切换：Gemini 2.5 Flash / Flash Lite / 3 Pro Preview；语言可选（默认中文）。
|
||
Models: Gemini 2.5 Flash/Flash Lite/3 Pro Preview; language selectable (CN default).
|
||
|
||
- 一键开发启动：`make dev`。
|
||
One-command dev start: `make dev`.
|
||
|
||
|
||
|
||
## 目录结构 / Project layout
|
||
- `backend/`：LangGraph + FastAPI，含 `/generate_image`。
|
||
`backend/`: LangGraph + FastAPI with `/generate_image`.
|
||
|
||
- `frontend/`：Vite + React，展示对话/分镜卡片/图片状态。
|
||
`frontend/`: Vite + React UI for chat, cards, image status.
|
||
|
||
- `backend/scripts/test_generate_image.py`：独立图生成测试脚本。
|
||
`backend/scripts/test_generate_image.py`: standalone image test script.
|
||
|
||
## 环境与依赖 / Requirements
|
||
- Python 3.11+（后端）、Node 18+（前端）。
|
||
Python 3.11+ (backend), Node 18+ (frontend).
|
||
|
||
- 环境变量：`GEMINI_API_KEY`。
|
||
Env var: `GEMINI_API_KEY`.
|
||
|
||
- 图生成需 `gemini-3-pro-image-preview` 权限，否则无图/404。
|
||
Image gen needs access to `gemini-3-pro-image-preview`, else 404/no image.
|
||
|
||
## 快速开始 / Quick start
|
||
1) 克隆并进入项目，准备虚拟环境：
|
||
Clone repo and prepare venv:
|
||
```bash
|
||
cd backend
|
||
python -m venv .venv && source .venv/bin/activate
|
||
pip install -e .
|
||
echo "GEMINI_API_KEY=你的密钥" > .env
|
||
```
|
||
|
||
2) 前端安装依赖：
|
||
Install frontend deps:
|
||
```bash
|
||
cd ../frontend
|
||
npm install
|
||
```
|
||
3) 启动开发环境（根目录）：
|
||
Start dev (root):
|
||
```bash
|
||
make dev # 并行起前后端：Vite + langgraph dev
|
||
```
|
||
前端默认 http://localhost:5173，LangGraph/后端默认 http://localhost:2024。
|
||
Frontend at http://localhost:5173, backend at http://localhost:2024.
|
||
|
||
## 常用命令 / Useful commands
|
||
- 后端单独开发：`cd backend && source .venv/bin/activate && langgraph dev`
|
||
Backend only: `cd backend && source .venv/bin/activate && langgraph dev`
|
||
|
||
- 前端单独开发：`cd frontend && npm run dev`
|
||
Frontend only: `cd frontend && npm run dev`
|
||
|
||
- 前端交互生成：页面输入主题 → 分镜逐页渲染 → 每页自动触发图生成（需模型权限）。
|
||
Frontend flow: enter topic → cards render → each card triggers image gen (model access required).
|
||
|
||
- 脚本验证图生成：
|
||
Scripted image test:
|
||
```bash
|
||
cd backend && source .venv/bin/activate
|
||
python scripts/test_generate_image.py --prompt "Robot holding a red skateboard"
|
||
```
|
||
|
||
## 已知注意事项 / Notes
|
||
- 图生成依赖 `gemini-3-pro-image-preview`，未开通会 404/无图；先用脚本验证。
|
||
Image gen needs `gemini-3-pro-image-preview`; without access expect 404/no image—verify via script.
|
||
|
||
- 分镜 prompt 过长或含链接时可能被过滤，可简化后再触发生成。
|
||
Long prompts/links may be filtered; simplify and retry if images are missing.
|
||
|
||
- 确保 `.env` 被加载（`langgraph dev` 默认读取 `backend/.env`）。
|
||
Ensure `.env` is loaded (`langgraph dev` reads `backend/.env`).
|