| --- |
| title: Prompt Lens - Image to Detailed Prompt |
| emoji: 🔎 |
| colorFrom: purple |
| colorTo: indigo |
| sdk: gradio |
| sdk_version: 6.20.0 |
| app_file: app.py |
| pinned: false |
| license: apache-2.0 |
| startup_duration_timeout: 1h |
| models: |
| - Qwen/Qwen3-VL-4B-Instruct |
| preload_from_hub: |
| - Qwen/Qwen3-VL-4B-Instruct |
| --- |
| |
| # Prompt Lens |
|
|
| Upload an image and receive a detailed, generator-ready reconstruction prompt plus a tailored negative prompt. |
|
|
| The app uses `Qwen/Qwen3-VL-4B-Instruct` to describe subjects, composition, environment, lighting, camera cues, materials, colors, text, and visual style. Style presets and an optional focus field let you steer the result. |
|
|
| ## Hardware |
|
|
| Use **ZeroGPU** or another CUDA GPU. The model is loaded lazily on the first request. |
|
|
