System Prompt – Usage Instructions

1. Technical Setup

This system prompt was designed and tested in LM Studio.

Test model: qwen3-v1-8b-instruct q6_k.gguf (other models have not been tested and may produce inconsistent results).

2. Creating a Preset in LM Studio

Open LM Studio. Create a new preset. Copy the full contents of the .txt system prompt file.

Paste it into the System Prompt field. Save the preset.

3. Recommended Model Settings (Qwen3)

Use the following parameters for stable and consistent output:

Temperature: 0.7 : Top-K: 40 : Top-P: 0.95 : Repeat Penalty: 1.1

These values balance creativity with structure, ensuring the model follows the strict prompt rules while still generating varied outputs.

4. How to Trigger Prompt Generation

The system is activated using age keywords. Type one of the following and send it as your input: teenage boy, young adult man, adult man, middle-aged man, elderly man

Important: Use only one keyword per prompt. The model will generate a complete portrait prompt based on this input. Do not add extra text unless you intend to modify behavior.

5. Image Generation Model

This prompt is optimized for: chroma-unlocked-Q4_K_S v48-detail-calibrated https://civarchive.com/models/1807633?modelVersionId=2183592

Copy prompt from LM Studio and paste as positive prompt. Negative prompt: no. (why, no? Because the analog style can't be too clean, too perfect).

6. You can also add your own short descriptions by modifying the system prompt (overly long descriptions can cause behavioral changes, qwen will only select the first phrases or generate the same prompt over and over).

7. This style won't work with models like z-image-base, z-image-turbo, flux-2klein, etc., because these models generate images that are too clean, too digital.

p.s. use cgf = 3.5 for more interesting results