Prompts für Bildbeschreibung und Captioning

import anthropic, base64

client = anthropic.Anthropic(api_key='YOUR_API_KEY')

with open('image.jpg', 'rb') as f:
    image_data = base64.standard_b64encode(f.read()).decode('utf-8')

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=500,
    messages=[{
        'role': 'user',
        'content': [
            {
                'type': 'image',
                'source': {
                    'type': 'base64',
                    'media_type': 'image/jpeg',
                    'data': image_data
                }
            },
            {
                'type': 'text',
                'text': 'Describe this image in detail.'
            }
        ]
    }]
)
print(response.content[0].text)

structured_prompt = '''
Describe this image in detail, addressing each of the following aspects:

1. FOREGROUND: Main subjects and objects in the foreground
2. BACKGROUND: Setting, environment, and background elements
3. COLORS: Dominant color palette and notable color contrasts
4. MOOD: Emotional tone, atmosphere, and lighting
5. TEXT: Any visible text, signs, labels, or written content
6. PEOPLE: If people are present — count, approximate age, pose, expression

Organize your response using these exact section headers.
Be specific and descriptive. Avoid vague terms like "some" or "various".
'''

print(structured_prompt)

# For image captions in a product catalog
short_prompt = '''
Write a 1-sentence product image caption (under 15 words).
Focus on the product, its key feature, and setting.
'''

# For accessibility alt-text
alt_text_prompt = '''
Write an image alt-text description for a visually impaired user.
Limit: 125 characters.
Include: what the image shows, any text visible in the image, the most important action or emotion.
'''

# For detailed analysis
detailed_prompt = '''
Write a detailed image analysis of 200-300 words.
Cover: composition, subjects, setting, colors, mood, and any notable technical or artistic elements.
Structure as a single flowing paragraph.
'''

print('Three length-controlled description prompts defined.')

# Medical imaging description
medical_prompt = '''
Describe the key visual findings in this medical image.
Focus on: anatomical structures visible, any abnormalities or anomalies,
location using standard anatomical terms (left/right, superior/inferior, medial/lateral),
and image quality or artifacts.
Note: this description is for informational purposes only, not diagnostic.
'''

# Architecture / real estate
architecture_prompt = '''
Describe this property image for a real estate listing.
Cover: room type, approximate size, key features (flooring, ceiling, natural light),
condition, notable fixtures or finishes, and overall style.
Tone: professional, appealing, factual.
'''

# Security / surveillance
security_prompt = '''
Describe this security camera image.
Note: number of people, approximate location in frame, clothing colors,
any objects being carried, direction of movement, and time of day if discernible.
'''

print('Domain-specific prompts defined.')

json_description_prompt = '''
Analyze this product image and return a JSON description:
{
  "product_name": "inferred product name or null",
  "category": "electronics|clothing|furniture|food|other",
  "colors": ["primary color", "secondary color"],
  "condition": "new|used|unclear",
  "background": "white|lifestyle|outdoor|studio|other",
  "people_visible": true | false,
  "text_visible": "extracted text or null",
  "quality_score": 1-10,
  "caption": "one sentence product caption"
}
Return only the JSON object.
'''

print(json_description_prompt)

accessibility_prompt = '''
Write an image description optimized for screen reader accessibility.

Guidelines:
- Start with the most important content (what is this image about?)
- Describe spatial relationships (the man on the left, the building in the background)
- Include all visible text verbatim
- Describe faces and expressions if relevant to the content
- Skip decorative descriptions unless they convey meaning
- End with: if this is a graph or chart, include the key data it shows
- Maximum 250 characters for alt-text. If more is needed, write a 1-sentence alt-text plus a longer caption.
'''

print(accessibility_prompt)

import anthropic, base64, json
from pathlib import Path

client = anthropic.Anthropic(api_key='YOUR_API_KEY')

DESCRIPTION_PROMPT = '''
Describe this image for a product catalog.
Return JSON: {"caption": str, "colors": [str], "category": str, "alt_text": str}
'''

def describe_image(image_path):
    with open(image_path, 'rb') as f:
        img_b64 = base64.standard_b64encode(f.read()).decode('utf-8')
    r = client.messages.create(
        model='claude-opus-4-5', max_tokens=200,
        messages=[{'role': 'user', 'content': [
            {'type': 'image', 'source': {'type': 'base64', 'media_type': 'image/jpeg', 'data': img_b64}},
            {'type': 'text', 'text': DESCRIPTION_PROMPT}
        ]}]
    )
    return json.loads(r.content[0].text)

print('Batch image description pipeline defined.')