Ton formel ou informel

Ton : la voix de votre contenu

Ce qui caractérise l'écriture formelle

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)

Ce qui caractérise l'écriture informelle

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)

Même contenu, trois tons

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()

Préciser le ton avec des adjectifs

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()

Préciser le ton avec une persona de modèle

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()

Décalages de ton et leur coût

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()

Contrôler la température émotionnelle

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()}')

Maintenir un ton cohérent

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)

Le ton selon les types de contenu

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()

Le ton selon les canaux

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()

Vérification des connaissances

Ton formel ou informel — récapitulatif