What is Prompt Reverse Engineer (提示词反推器)?
Prompt Reverse Engineer (提示词反推器) is a Windows desktop application that converts visual evidence and creative requirements into structured image prompts, video prompts, motion instructions, keyframe descriptions, timeline plans, style specifications, and platform-oriented execution drafts.
Core capabilities
- Image prompt reverse engineering: extracts subjects, composition, camera, lighting, color, materials, style, and negative constraints from still images.
- Video prompt reverse engineering: extracts keyframes, shot structure, timelines, action chains, and continuity constraints from video.
- Motion prompt reverse engineering: describes action start points, change paths, speed, rhythm, camera motion, and end states.
- Strict separation between static image prompts and video motion instructions.
- Chinese and English structured prompt output.
- Start, intermediate, and end-frame planning with motion chains.
- Transferable visual Style DNA extraction.
- Platform-oriented prompt compilation.
- Local projects, history, assets, queues, presets, and generation feedback.
- User-configured OpenAI-compatible multimodal APIs and per-function model routing.
System requirements
| Operating system | Windows 10 or later |
|---|---|
| Runtime | Python 3.10 or later with Tkinter |
| Model access | An OpenAI-compatible API that supports image input |
| Video processing | ffmpeg and ffprobe; the installer attempts to download them automatically |
Privacy model
The application does not include project-operated telemetry. Projects, history, generated files, temporary frames, and metadata are stored locally. Media and text explicitly included in a model request may be sent to the API provider configured by the user.
On Windows, saved API keys are encrypted with DPAPI and bound to the current Windows user. See the Chinese privacy page for the complete data boundary.
Language scope
Repository documentation and generated prompts are available in Chinese and English. Version 1.2 desktop controls remain primarily Chinese; the project does not claim that a small number of translated controls constitutes a complete English user interface.
This is an independent community project. It is not affiliated with or officially endorsed by Jimeng, ByteDance, Kling, Kuaishou, OpenAI, or other model and platform providers.