prompts.

**CharacterBank**：存储角色外貌、服装、风格标签，自动附加到每个镜头的提示词

✅ **完整的创作工作流**: 剧本 → 角色 → 场景 → 分镜 → 成片全流程自动化

五大核心面板（剧本/资产/分镜/动态/合成）

专用 Agent（提示词优化/质量检查/风格迁移）

剧本 → 角色/场景/道具 → 资产图库 → 分镜脚本 → 分镜视频 → 最终视频

Prompt = Subject + Motion, Background + Motion, Camera + Motion

[Subject]: Who or what (singular preferred) [Motion]: What they're doing (plain language, specific speed) [Scene]: Where it happens [Camera]: Shot type + movement + lens feel [Style]: Visual aesthetic, lighting

A professional woman in her 30s, wearing a tailored navy blazer, sitting in a modern office. She looks directly at camera with a confident smile. Soft natural light from a large window. Slow dolly-in focusing on her face. Camera: Medium close-up, 85mm lens feel, slow dolly-in over 4 seconds Style: Corporate headshot, warm professional lighting, shallow depth of field

Create a city skyline at sunset with flying cars and neon signs.

# 图生视频（单图） dreamina image2video \ --image ./character.jpg \ --prompt "人物缓慢转身微笑，微风吹动头发，慢速推进镜头，金色黄昏光，电影级调色，避免抖动，避免变形" \ --model_version seedance2.0 \ --duration 8 \ --poll 60 # 多模态（角色图 + 场景图 + 配乐） dreamina multimodal2video \ --image ./character.jpg \ --image ./scene.jpg \ --audio ./bgm.mp3 \ --prompt "人物缓慢走向远处，镜头平稳跟随，场景色调与背景图一致，避免抖动" \ --model_version seedance2.0 \ --duration 10 \ --ratio 16:9 \ --poll 60

Cinematic slow-motion shot of a cyberpunk runner leaping across neon rooftops at night, dramatic rain, pulsing synth soundtrack

Wide establishing shot → slow zoom in, golden hour lighting, epic orchestral build-up

Epic fantasy battle: knight charging dragon across stormy battlefield, dynamic camera tracking, thunderous soundtrack

Load Checkpoint | v CLIP Text Encode ----+ | | v v Universal Determinism (strict, seed=42) | v Empty Latent Image | v Deterministic Sampler (seed=42) | v VAE Decode | v Save Image

# 1. 登录 Cloudflare Dashboard # 2. 进入 "Workers & Pages" # 3. 点击 "Create Application" → "Create Worker" # 4. 命名为 nsfw-api-proxy