Files
Unified_Dashboard_Suite_Go/cmd/image-to-text/image_to_prompt/prompts.py
T

62 lines
2.5 KiB
Python

"""
Preset prompts for different image analysis and prompt generation tasks.
"""
PROMPT_PRESETS = {
"prompt": (
"Analyze this image and write a detailed, high-quality prompt suitable for text-to-image "
"generation (such as Midjourney, Flux, or Stable Diffusion). Describe the subject in detail, "
"the artistic medium and style, composition, camera framing, lighting, color palette, "
"atmosphere, and background details. Output only the prompt without conversational intro or outro."
),
"flux": (
"Generate a highly descriptive prompt for Flux / Midjourney to recreate this exact image. "
"Break down the subject, pose, attire, facial expression, artistic style, camera lens/lighting setup, "
"depth of field, background elements, and overall mood. Output only the prompt description."
),
"detailed": (
"Provide a comprehensive, dense visual description of this image. Describe all visible elements "
"including subjects, positioning, background, textures, lighting, colors, and subtle details. "
"Organize the description clearly."
),
"short": (
"Write a concise, accurate single-sentence caption describing the main subject and action in this image."
),
"ocr": (
"Extract and transcribe all visible text, letters, numbers, signs, labels, watermarks, or logos from this image. "
"If there is no text, reply 'No text detected'."
),
"tags": (
"Analyze this image and generate a comma-separated list of descriptive keywords/tags. "
"Include subject, clothing, colors, setting, art style, mood, and camera angle. "
"Format: tag1, tag2, tag3, ..."
),
"story": (
"Write an imaginative, vivid short paragraph telling the story or setting the narrative mood behind this image."
),
}
DEFAULT_MODE = "prompt"
def get_prompt(mode: str, custom_prompt: str | None = None) -> str:
"""
Get prompt text according to mode or custom user prompt.
Args:
mode: Preset key from PROMPT_PRESETS or 'custom'
custom_prompt: Explicit custom prompt text (overrides preset if provided)
Returns:
The prompt string to pass to the Vision-Language model.
"""
if custom_prompt and custom_prompt.strip():
return custom_prompt.strip()
mode_lower = mode.lower().strip()
if mode_lower in PROMPT_PRESETS:
return PROMPT_PRESETS[mode_lower]
# Fallback to default
return PROMPT_PRESETS[DEFAULT_MODE]