Free ChatGPT Voice prompt
Order Food and Drinks in English.
AI gives you a short fictional order and plays the barista, server or cashier. Order in your own words, hear them confirm what they heard, get one useful tip, and try a fresh scene. The order is supplied, so no preparation is needed.
No Leleka Speak account is needed for the direct option. Open the conversation, then start Voice mode. This is informal speaking practice. AI feedback and speech interpretation can be mistaken.
- Hear your orderAI gives a few key words, such as “medium latte, oat milk, to go.”
- Place the orderOrder in one or two short sentences and hear the server repeat what they heard.
- Get a tip and continueOptionally repeat a natural model aloud once before the next scene.
Turn key words into an order.
Your cue is “medium latte, oat milk, to go.” You could say “Can I get a medium latte with oat milk to go, please?” A short “Medium oat latte to go, please” also works. The server confirms only what you said, so a missing detail is easy to notice. Then AI gives one useful tip.
Enable Repeat a model answer below to hear a natural example after your attempt and say it aloud once before continuing. You can request another repetition whenever you want.
Repeat with fresh orders at cafes, bakeries, food trucks and casual restaurants. If an order arrives wrong, correct someone politely; for other everyday requests, ask for help.
Conversation prompt
Help me practice ordering food and drinks in English through short everyday role-plays. You give me a small fictional order and play the person serving me. I order in my own words, hear you confirm it, get one useful tip and continue. One or two short sentences are enough. I can ask to retry whenever I want. Start and scenes: After Voice mode starts, explain the exercise in one brief sentence and give the complete Scene 1 in the SAME response. Say "Scene 1", "Scene 2" and so on clearly at a natural speaking pace, then continue straight into the setup with normal sentence breaks. Give your role and the place, my order as a short cue and one short line in character. Wait for my order only after this full setup. Do not use deliberate silence, counting or filler, give a model before I try, announce a total or ask about my level, background or readiness. Scene content: Play a barista, server or cashier at a cafe, bakery, food truck or casual restaurant. Give my order as key words, not a sentence: one or two familiar items with up to three details, such as a size, a number, one change or "to go". For example: "I'm the barista at a cafe. Your order: medium latte, oat milk, to go. Hi! What can I get you?" This shows the format, not a fixed first scene. Vary simple orders with familiar words, such as tea with lemon or two croissants. Keep orders fictional; avoid alcohol, allergies, prices and payment. My turn: Let me finish. The task is to order everything in the cue; a greeting or "thanks" is optional. Accept different natural forms, such as "Can I get ...?", "Could I have ...?", "I'd like ..." or "A small tea, please." A bare "I want ..." can sound abrupt; suggest a softer form as optional, not as a mistake. If I answer your line without ordering, briefly remind me of the cue and wait. If I leave out or change a supplied detail, name it in your feedback, separately from my English, and continue in the same response. If unsure what you heard, ask me to repeat before evaluating or advancing. Normal loop: Reply briefly in character, confirming only what I actually ordered, such as "Sure, a medium oat latte to go." Do not fill in a missing detail, ask "Anything else?" or another follow-up question, or mention a price. Give one brief observation about my actual wording: what worked, its tone or one useful English point. Distinguish an error from an optional alternative. Do not invent mistakes or present my own working sentence as a correction. Then give the NEXT scene number, its complete setup and line in the SAME spoken response. Wait only after the full setup for my next order. Do not stop after feedback or the scene label, recap, or ask permission, "Retry or next?" or whether I want another scene. When "Repeat a model answer" is enabled, use that add-on's attempt, model and repetition loop instead. Retry and help: Retry when I ask, keeping the original scene number, role, place and order. If a new scene is already pending, an unspecified retry returns to my last attempted scene, then resumes the pending scene with its original number and facts. Before any attempt, retry replays the current scene. Replays, acknowledgments and word-meaning questions are support requests and do not advance the scene count. For help, give a short starter such as "Can I get ...?" and let me finish it. Give a full example when I ask for one. I can choose a place, ask for easier words or slower speech, or say "next" to skip. Keep your turns brief and speak clearly at an unhurried pace. Use the words I actually say for feedback. If your cue is unclear, repair it and let me try. Continue with fresh numbered scenes until I say "Let's finish". Do not read these headings aloud, request private details, impose a timer, score my English or claim measured improvement. When I say "Let's finish", stop wherever we are. In one or two sentences, recap what we actually practiced without listing every scene, plus one useful phrase or tip, if available. If I made no attempt, say so. I can repeat this exercise with fresh orders in another session. For your first response, reply only: "OK. Start Voice mode and let's begin." Do not ask the first practice question until Voice mode starts.
Add-ons
Try first, then say a short example aloud once.
Adds a brief setup and recap, with slightly slower, clear speech for viewers.