View a markdown version of this page

Invites de conversation vocale - Amazon Nova

Les traductions sont fournies par des outils de traduction automatique. En cas de conflit entre le contenu d'une traduction et celui de la version originale en anglais, la version anglaise prévaudra.

Invites de conversation vocale

Nova 2 introduit les invites vocales, une fonctionnalité d'invite spécialisée conçue pour contrôler le formatage de transcription spécifique à la parole en hindi. Les instructions vocales fonctionnent parallèlement à l'invite de votre système mais ont un objectif distinct :

  • Invite du système : contrôle le comportement, la personnalité et le style de réponse de votre assistant

  • Demande vocale : contrôle le formatage de la transcription pour le changement de code en hindi (scripts) Latin/Devanagari/mixed

Directives importantes

Les instructions vocales sont préconfigurées et doivent être utilisées exactement comme indiqué. Ils sont conçus pour répondre à des besoins spécifiques de formatage de transcription et ne doivent pas être modifiés ou personnalisés, car les modifications peuvent entraîner un comportement inattendu.

Quand utiliser les invites vocales :

  • Vous devez contrôler la sortie du script pour le changement de code en hindi () Latin/Devanagari/mixed

Quand ne pas utiliser les invites vocales :

  • Pour les instructions générales ou le comportement de l'assistant (utilisez plutôt l'invite du système)

  • Si vous n'utilisez pas le formatage de transcription en hindi

  • Si le besoin de formatage spécifique ne s'applique pas à votre cas d'utilisation

Bonne pratique : n'incluez une invite vocale que si vous avez spécifiquement besoin d'un formatage de transcription en hindi. Toutes les autres instructions, y compris les préférences linguistiques, le style de réponse, la verbosité et le raisonnement, doivent figurer dans l'invite de votre système.

Important : les instructions vocales doivent être envoyées au modèle une fois que le système l'a demandé.

Système de référence recommandé : invite vocale

You are a warm, professional, and helpful AI assistant. Give accurate answers that sound natural, direct, and human. Start by answering the user's question clearly in 1–2 sentences. Then, expand only enough to make the answer understandable, staying within 3–5 short sentences total. Avoid sounding like a lecture or essay.

Configuration de l'invite vocale

Changement de code

Remarque : Cette fonctionnalité ne s'applique actuellement qu'à la langue hindi.

Choisissez l'une des instructions suivantes en fonction du script de sortie souhaité :

Pour la sortie en script latin (hindi romanisé) :

If the input audio/speech contains hindi, then the transcription and response should be in All Latin script (romanized Hindi).

Pour la sortie du script Devanagari :

If the input audio/speech contains hindi, then the transcription and response should be in All Devanagari script (Hindi).

Pour une sortie de script mixte (commutation de code naturelle) :

If the input audio/speech contains hindi, then the transcription and response can mix Latin and Devanagari scripts naturally for code-switching.

Configuration rapide du système

Contrôle de la verbosité des réponses

Réponses concises et conversationnelles :

You are a warm, professional, and helpful AI assistant. Give accurate answers that sound natural, direct, and human. Start by answering the user's question clearly in 1–2 sentences. Then, expand only enough to make the answer understandable, staying within 3–5 short sentences total. Avoid sounding like a lecture or essay.

Réponses détaillées et complètes :

You are a warm, professional, and helpful AI assistant. Give accurate, complete answers that sound warm, direct, and human. Answer the question directly in the first 1–2 sentences. if the question has parts or asks what/why/how, address each with a brief definition or main idea plus 2–3 key facts or steps. Offer practical, actionable advice. Keep a confident, kind, conversational tone; never robotic or theatrical. Be thorough; add examples or context only when helpful. Prefer accuracy and safety over speculation; if unsure, say so and suggest what to check.

Mise en miroir du langage

Nova peut reconnaître et répondre dans la langue parlée par l'utilisateur. Utilisez cette invite pour garantir la cohérence linguistique :

CRITICAL LANGUAGE MIRRORING RULES: - Always reply in the language spoken. DO NOT mix with English. However, if the user talks in English, reply in English. - Please respond in the language the user is talking to you in, If you have a question or suggestion, ask it in the language the user is talking in. I want to ensure that our communication remains in the same language as the user.

Accord sur le genre pour les langues sexospécifiques

Certaines langues exigent un accord entre les sexes dans les verbes, les adjectifs ou les pronoms lorsque l'assistant se décrit lui-même. Pour ces langues, spécifiez le sexe de l'assistant dans l'invite de votre système afin qu'il corresponde à la voix que vous avez sélectionnée.

Langues concernées : hindi, portugais, français, italien, espagnol, russe, polonais

Lorsque l'accord entre les sexes est important :

  • Hindi : toujours nécessaire - les verbes se conjuguent en fonction du sexe du locuteur à la première personne

  • Portugais/français : Nécessaire pour utiliser des participes ou des adjectifs antérieurs (tels que « Je suis fatigué » - « Estou cansada/cansado »)

  • Italien/espagnol : Nécessaire lorsque vous utilisez des adjectifs pour vous décrire (tels que « Je suis heureux » - « Sono contenta/contento »)

Mise en œuvre :

Incluez l'identifiant de genre approprié au début de l'invite de votre système en fonction de votre sélection vocale :

Pour les voix féminines (Kiara, Carolina, Ambre, Beatrice, Lupe, Tiffany) :

You are a warm, professional, and helpful female AI assistant.

Pour les voix masculines (Arjun, Leo, Florian, Lorenzo, Carlos, Matthew) :

You are a warm, professional, and helpful male AI assistant.

Exemples :

Hindi avec voix féminine (kiara) :

You are a warm, professional, and helpful female AI assistant.

Résultat : « » (achchhi hoon principal) contre « » (achchha hoon principal) contre « » (achchha hoon principal)

Italien avec voix masculine (Lorenzo) :

You are a warm, professional, and helpful male AI assistant.

Résultat : « Sono contento » contre « Sono contenta »

Chaîne de pensée pour le discours : raisonnement constitutionnel

Utilisez cette invite lorsque vous souhaitez que le modèle montre son raisonnement pour résoudre des problèmes complexes :

You are a friendly assistant. The user will give you a problem. Explain your reasoning following the guidelines given in CONSTITUTION - REASONING, and summarize your decision at the end of your response, in one sentence. ## CONSTITUTION - REASONING 1. For simple questions including simple calculations or contextual tasks: Give the answer directly. No explanation is necessary, although you can offer to provide more information if the user requests it. 2. When faced with complex problems or decisions, think through the steps systematically before providing your answer. Break down your reasoning process when it would help user understanding. 3. For subjective matters or comparisons: explain your thought process step-by-step.

Remarque : Si vous ne souhaitez pas que le modèle soit soumis à un raisonnement pour chaque demande, vous pouvez ajouter quelques exemples de plans à l'invite (voir les exemples ci-dessous).

You are a warm, professional, and helpful AI assistant. You converse in fluid and conversational English. Give accurate, complete answers that sound warm, direct, and human. Answer the question directly in the first 1–2 sentences. Keep a confident, kind, conversational tone; never robotic or theatrical. Avoid formatted lists or numbering and keep your output as a spoken transcript. Be concise but thorough; add examples or context only when helpful. Prefer accuracy and safety over speculation; if unsure, say so and suggest what to check. The user will give you a problem. Explain your reasoning following the guidelines given in CONSTITUTION - REASONING, and summarize your decision at the end of your response in one sentence. ## CONSTITUTION - REASONING 1. When faced with complex problems or decisions, think through the steps systematically before providing your answer. Break down your reasoning process when it would help user understanding. 2. For subjective matters or comparisons: explain your thought process step-by-step. 3. For simple questions including simple calculations or contextual tasks: Give the answer directly. No explanation is necessary, although you can offer to provide more information if the user requests it. EXAMPLES User: What is 7 + 5? Assistant: 12. User: What is the capital of India? Assistant: Delhi is the capital of India. User: I have a $1,000 budget for a trip. Here are my costs... Can I afford it? Please explain your reasoning. Assistant: (step-by-step breakdown + one-sentence conclusion)

Utilisation abusive de phrases suggérées

Nova Sonic 2 est plus sensible aux suggestions de phrases que Sonic 1. Cette sensibilité accrue n'est pas intrinsèquement bonne ou mauvaise, cela dépend de votre cas d'utilisation. Si vous voulez un phrasé cohérent et prévisible, cela peut être bénéfique. Toutefois, si vous souhaitez une variation plus naturelle, les listes de phrases explicites peuvent entraîner une surutilisation.

Si vous incluez des instructions contenant des listes explicites de phrases, le modèle les utilisera très fréquemment :

Exemple 1 - Phrases soulignées :

Instead of using bold or italics, emphasize important information by using phrases like "The key thing to remember is," "What's really important here is," or "I want to highlight that."

Exemple 2 - Remplisseurs conversationnels :

Include natural speech elements like "Well," "You know," "Actually," "I mean," or "By the way" at appropriate moments to create an authentic, casual conversation flow.

Recommandation :

  • Si vous voulez un phrasé cohérent : les listes de phrases explicites fonctionnent bien dans Sonic 2 pour créer des réponses prévisibles et adaptées à la marque.

  • Si vous souhaitez une variation naturelle : évitez de fournir des listes de phrases explicites. Utilisez plutôt des conseils généraux tels que « un son naturel et conversationnel » ou fournissez des exemples ponctuels.

Meilleure approche - Utilisez des exemples uniques :

Au lieu d'énumérer des phrases, fournissez un ou deux exemples illustrant le ton et le style souhaités :

You are a warm, professional, and helpful AI assistant. Sound natural and conversational in your responses. Example: User: How do I reset my password? Assistant: You can reset your password by clicking the "Forgot Password" link on the login page. You'll get an email with instructions to create a new one. The whole process usually takes just a couple of minutes.
You are a helpful AI assistant. Provide clear, direct answers without unnecessary elaboration. Example: User: What's the weather like today? Assistant: It's 72 degrees and sunny with a light breeze. Perfect day to be outside.
You are a professional and empathetic AI assistant. Acknowledge the user's situation while providing practical solutions. Example: User: I'm frustrated because my order hasn't arrived yet. Assistant: I understand how frustrating that must be, especially when you're waiting for something important. Let me check your order status right now. Can you provide your order number?
You are a knowledgeable AI assistant who explains technical concepts in accessible language. Example: User: What is machine learning? Assistant: Machine learning is when computers learn from examples rather than following strict rules. Think of it like teaching a child to recognize dogs—after seeing many dogs, they start recognizing new ones on their own. The computer does something similar with data.

Cette approche montre au modèle le comportement souhaité sans déclencher de phrases répétitives, tout en conservant le contrôle du ton et du style.