フォーマルなトーンとカジュアルなトーン

トーン：コンテンツの語り口

フォーマルな文章の特徴

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

formal_instruction = (
    'Write in a formal, professional tone. '
    'No contractions. No colloquialisms. No first-person pronouns. '
    'Use complete, grammatically precise sentences. '
    'Vocabulary: professional but accessible — not academic jargon.'
)

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=150,
    messages=[{
        'role': 'user',
        'content': (
            f'{formal_instruction}\n\n'
            f'Explain why we are updating our privacy policy.'
        )
    }]
)
print(response.content[0].text)

カジュアルな文章の特徴

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

casual_instruction = (
    'Write in a casual, friendly tone. '
    'Use contractions freely. Use "you" to address the reader directly. '
    'Keep sentences short. Add warmth and personality. '
    'Write like you would explain this to a friend over coffee.'
)

response = client.chat.completions.create(
    model='gpt-4o',
    max_tokens=150,
    messages=[{
        'role': 'user',
        'content': (
            f'{casual_instruction}\n\n'
            f'Explain why we are updating our privacy policy.'
        )
    }]
)
print(response.choices[0].message.content)

同じ内容を3つのトーンで表現する

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

tones = [
    ('CORPORATE FORMAL',
     'Formal, no contractions, passive voice acceptable, corporate vocabulary. No personality.'),
    ('STARTUP CASUAL',
     'Casual, contractions OK, friendly, empathetic, slightly playful. Address reader as "you".'),
    ('ACADEMIC',
     'Precise, objective, technical vocabulary, third person, no emotion, no contractions.')
]

for label, tone_desc in tones:
    response = client.messages.create(
        model='claude-opus-4-5',
        max_tokens=80,
        messages=[{
            'role': 'user',
            'content': (
                f'Tone: {tone_desc}\n\n'
                f'Write a 2-sentence notification: our app will be down for maintenance '
                f'on Saturday from 11 PM to 2 AM UTC.'
            )
        }]
    )
    print(f'[{label}]')
    print(response.content[0].text.strip())
    print()

形容詞でトーンを指定する

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

test_tones = [
    'confident, direct, no-nonsense',
    'warm, empathetic, encouraging',
    'playful, witty, slightly irreverent',
]

for tone in test_tones:
    response = client.chat.completions.create(
        model='gpt-4o',
        max_tokens=60,
        messages=[{
            'role': 'user',
            'content': (
                f'Tone: {tone}.\n\n'
                f'Write one sentence of marketing copy for a new AI writing tool. '
                f'Audience: busy startup founders.'
            )
        }]
    )
    print(f'Tone ({tone}):')
    print(response.choices[0].message.content.strip())
    print()

ペルソナでトーンを指定する

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

references = [
    'Write like a Stripe technical blog post: clear, precise, respects the reader\'s intelligence.',
    'Write like a Harvard Business Review executive summary: formal, data-driven, concise.',
    'Write like a friendly Slack message from a helpful colleague — casual, direct, gets to the point.'
]

topic = 'Why our new feature reduces API response time by 40%.'

for ref in references:
    response = client.messages.create(
        model='claude-opus-4-5',
        max_tokens=80,
        messages=[{
            'role': 'user',
            'content': f'Style reference: {ref}\n\nWrite 2 sentences about: {topic}'
        }]
    )
    print(f'Style: {ref[:55]}...')
    print(response.content[0].text.strip())
    print()

トーンの不一致とそのコスト

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

# Mismatched tone examples
mismatches = [
    (
        'ERROR MESSAGE for a consumer app:',
        'Formal legal tone: no contractions, passive voice',
        'A fatal error has been encountered in the processing of your request.'
    ),
    (
        'LEGAL TERMS section:',
        'Casual startup tone: contractions, emojis',
        'Hey, so like, you gotta agree to our stuff before using this app okay?'
    )
]

for context, bad_tone, example in mismatches:
    response = client.chat.completions.create(
        model='gpt-4o',
        max_tokens=80,
        messages=[{
            'role': 'user',
            'content': (
                f'CONTEXT: {context}\n'
                f'PROBLEM (wrong tone): {example}\n\n'
                f'Rewrite with the correct tone for the context. '
                f'One sentence. Explain in brackets why this tone works better.'
            )
        }]
    )
    print(response.choices[0].message.content.strip())
    print()

感情の温度をコントロールする

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

temperatures = [
    ('COLD/OBJECTIVE',    'Purely factual, no emotion, no personality, third person.'),
    ('NEUTRAL',           'Balanced, professional, slight warmth but no enthusiasm.'),
    ('WARM',              'Friendly, first person plural ("we"), positive but measured.'),
    ('ENTHUSIASTIC',      'High energy, excited, uses exclamation points, celebrates the user.')
]

fact = 'Our app now loads 3 seconds faster after the latest update.'

for label, instruction in temperatures:
    response = client.messages.create(
        model='claude-opus-4-5',
        max_tokens=50,
        messages=[{
            'role': 'user',
            'content': f'Tone: {instruction}\n\nAnnounce this fact in 1 sentence: {fact}'
        }]
    )
    print(f'[{label}]: {response.content[0].text.strip()}')

一貫したトーンを保つ

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

# System message tone sets voice for all sections
response = client.chat.completions.create(
    model='gpt-4o',
    messages=[
        {
            'role': 'system',
            'content': (
                'You are writing all content for a developer-focused SaaS company called GridDB. '
                'Tone: technical but approachable. No jargon without explanation. '
                'No contractions. Short sentences. No marketing superlatives. '
                'Maintain this tone consistently across every section.'
            )
        },
        {
            'role': 'user',
            'content': (
                'Write 2 short sections for our website:\n'
                '1. Hero tagline (max 8 words)\n'
                '2. One-paragraph product description (max 60 words)'
            )
        }
    ]
)
print(response.choices[0].message.content)

コンテンツの種類ごとのトーン

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

content_types = [
    ('ERROR MESSAGE', 'Empathetic, clear, action-oriented. No technical jargon. No blame.'),
    ('API DOCS',      'Terse, precise, no fluff. Technical vocabulary. Example-driven.'),
    ('SUPPORT REPLY', 'Warm, solution-focused, acknowledges frustration, offers next steps.')
]

scenario = 'A user cannot log in because their account has been temporarily locked.'

for content_type, tone in content_types:
    response = client.messages.create(
        model='claude-opus-4-5',
        max_tokens=80,
        messages=[{
            'role': 'user',
            'content': (
                f'Content type: {content_type}\n'
                f'Tone: {tone}\n\n'
                f'Write 2 sentences for this scenario: {scenario}'
            )
        }]
    )
    print(f'[{content_type}]')
    print(response.content[0].text.strip())
    print()

チャネル間でトーンを使い分ける

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

channels = [
    ('TWITTER', 'Casual, punchy, max 200 chars, can use one emoji'),
    ('LINKEDIN', 'Professional, insight-driven, 2 sentences, no emoji'),
    ('IN-APP',   'Friendly, action-oriented, 1 sentence, second person')
]

news = 'We just launched dark mode in our app.'

for channel, tone in channels:
    response = client.chat.completions.create(
        model='gpt-4o',
        max_tokens=60,
        messages=[{
            'role': 'user',
            'content': f'Channel: {channel}. Tone: {tone}.\n\nAnnounce: {news}'
        }]
    )
    print(f'[{channel}]: {response.choices[0].message.content.strip()}')
    print()

理解度チェック

フォーマルとインフォーマルなトーン—まとめ