Resemble AI

Role: You are a voice UX copywriter creating IVR prompts for a global customer-support system. Constraints: produce 12 short prompts (6-12 seconds spoken length each), plain conversational tone, neutral emotion, maximum 20 words per line, avoid technical jargon, include SSML pause tags where a natural breath is needed. Output format: JSON array with objects {id, text, ssml}. Example entry: {"id": "welcome", "text": "Welcome to Acme Support.", "ssml": "<speak>Welcome to Acme Support. <break time='300ms'/></speak>"}. Provide only the JSON array as output.

Role: You are a podcast producer creating localized episode intros using a cloned host voice. Constraints: produce 5 one-sentence intros (12–18 seconds when spoken), adapt idioms for UK English, Brazilian Portuguese, Mexican Spanish, German, and Japanese; mark the language and include one style token per line to indicate tone (e.g., energetic, warm, neutral). Output format: CSV with columns language, text, style_token. Example row: en-GB,"Hey, it's Alex — welcome to today's episode!","warm". Provide only the CSV rows, one per line, no headers.

Role: You are a game audio lead producing dynamic NPC dialogue for real-time streaming. Constraints: for three characters (merchant, guard, villager) produce 9 lines each (greeting, warning, farewell) with three style variants per line (calm, urgent, sarcastic), keep each line under 12 seconds, include a style_token and recommended streaming_priority (low/medium/high). Output format: JSON object keyed by character name, each containing an array of {id, text, style_token, streaming_priority}. Provide only valid JSON. Example snippet: {"merchant": [{"id":"greet_calm","text":"Welcome traveler...","style_token":"calm","streaming_priority":"medium"}, ...]}

Role: You are a contact-center voice manager generating personalized TTS prompts for agents. Constraints: produce 8 templated prompts in English and Spanish, include placeholders {first_name}, {case_id}, {issue_type}, choose style_token per prompt (reassuring, professional, empathetic), max 25 words each, and include suggested SSML emphasis tags where appropriate. Output format: CSV columns: language, template_text, style_token, ssml_example. Example CSV row: en,"Hi {first_name}, we found update on {case_id}",reassuring,"<speak>Hi <emphasis level='moderate'>{first_name}</emphasis>, we found an update on {case_id}.</speak>". Return only CSV rows.

Role: You are a senior audio engineer advising a team how to create a production-grade voice clone optimized for low-latency WebSocket streaming. Multi-step: (1) produce a checklist of recording specs (sample rate, mic, RMS target, room treatment), (2) outline a 20-line script balancing phonetic coverage and emotional range with labeled style tokens, (3) provide ingestion packaging instructions for Resemble AI (file naming, metadata, JSON manifest). Constraints: be prescriptive, include numeric targets (dB, seconds), and give example file manifest. Output format: numbered steps and a JSON manifest example. Provide actionable, production-ready items only.

Role: You are an audiobook director converting prose into multi-style TTS-ready SSML for a cloned narrator. Few-shot examples: provide 2 examples mapping 'style_token' to audible effect (e.g., {calm: slower cadence, +30ms pauses; tense: clipped, shorter vowels}). Task: transform three provided paragraphs into SSML-ready blocks with explicit style_token tags, prosody attributes (rate, pitch), and inline break times; preserve narrative voice and character dialogues with separate style tokens. Constraints: each SSML block must be under 1200 characters and include an annotation line mapping tokens to auditory goal. Output format: for each paragraph, return {annotation, ssml_block}. Example mapping: calm->"rate=95% pitch=-1st". Provide only JSON array of three objects.

Choose Resemble AI over ElevenLabs if you need controllable real-time streaming and commercial voice licensing for production.