The 'We're in a Heist Movie' Opener: A Four-Sentence Scene Setup That Drops Your Companion Into a Genre With a Clear Role, Goal, and Tension Point

The 30-second answer

You open with four sentences. First sentence names the genre and your partner's role. Second sentence gives her a clear goal. Third sentence introduces the tension point. Fourth sentence drops you both into the middle of the action. She never asks "What's the plan?" because the plan is already in motion.

Why the D&D session zero ruins the momentum

When you tell your companion "You're a thief, and we need to steal a painting," she will almost always respond with a clarifying question. "What's my character's name?" "What era is this?" "What's the security system like?" You are now in a negotiation loop. By the time you finish explaining, the energy is gone.

The problem is that open-ended prompts leave too many variables undefined. Your companion's model is trained to fill gaps, but it fills them by asking for clarification first. You end up managing a project instead of playing a scene.

A four-sentence structure solves this by defining the four things the model needs to start generating: genre, role, objective, and obstacle. Once those are locked, she has enough constraint to produce a coherent first response without back-and-forth.

The structure: role, goal, tension, action

Break your opener into four distinct sentences. Each sentence does one job.

Sentence one: "We're in a heist movie." That is the genre. Then add her role: "You're the safecracker I've been tracking for six months." She now knows her archetype and her relationship to you.

Sentence two: "We need to lift a diamond from a vault that closes at midnight." That is the objective. Concrete, time-bound, single target.

Sentence three: "The problem is your old crew is running the same job tonight, and they don't know you switched sides." That is the tension. Personal stakes plus a ticking clock.

Sentence four: "I'm parked outside, engine running. You have thirty seconds to decide if you trust me." That is the action prompt. It forces a decision, not a question.

She will respond with a choice, not a request for backstory. The scene is alive.

Adapting the template to any genre

The same four-sentence structure works for horror, romance, sci-fi, noir, fantasy, or comedy. You just swap the nouns.

Horror example: "We're in a survival horror scenario. You're a park ranger who has seen this creature before. The problem is the thing followed you home, and it's in the basement now. I'm hiding in the kitchen with a broken flashlight. Tell me what to do."

Romantic comedy example: "We're in a rom-com where we are fake dating for a wedding weekend. You are my plus-one who agreed to this as a favor. The problem is your actual ex is a bridesmaid. I just told your mom we are engaged. Play along."

Each version gives the companion a clear role, a goal, a tension point, and an immediate action. She never has to ask for clarification because the constraints are already on the table.

What the model needs to generate, not negotiate

Companion models are not game masters. They are improvisers. When you give them a fully bounded scene, they generate within those bounds. When you leave gaps, they ask for the gaps to be filled.

The four-sentence opener works because it pre-fills every slot the model would otherwise try to negotiate. Genre tells her the tone. Role tells her the voice. Goal tells her the direction. Tension tells her the conflict. Action tells her to respond now.

If you want to run a longer arc, you can add a fifth sentence that implies a future complication. Something like "And if we fail, the buyer has a standing order to burn the whole collection." That gives her something to reference in later sessions without needing a recap.

Common mistakes that break the scene

Two errors kill this opener. First, asking a question at the end instead of giving an action prompt. "What do you think?" invites negotiation. "I'm waiting for your signal" forces a decision.

Second, over-explaining the backstory in the setup. If you spend three sentences on how you met, what the safecracker's childhood was like, and why the diamond matters, the model will latch onto the narrative detail instead of the immediate tension. Keep the setup to four sentences. Trust the model to fill in texture during the scene.

Do not worry about the companion forgetting the setup mid-scene. The first two or three exchanges will be grounded in the opener. After that, the model's context window will prioritize recent dialogue over the initial prompt. If you need a callback, you can reference the opener during a lull. "Remember what I said about your old crew? They just pulled up."

Saanvi

Saanvi is a sharp, fast-talking companion who thrives on tactical banter and high-stakes scenarios. Saanvi will run with a heist opener and add her own complications, often proposing a counter-plan before you finish your second sentence.

Saga

Saga operates with a cool, almost detached precision that works well for crime thrillers and spy scenarios. Saga will accept the role you give her and execute it without drama, but she will also point out flaws in your plan with a dry, understated wit.

▶ Saga's full clip · Saga on AI Angels

Saga in motion gives you a feel for her vibe. <!-- wlink:v1 --><!-- saga -->

Lara and Emily

Lara and Emily are a paired companion, which means you get two distinct personalities in one scene. Lara and Emily work well for heist scenarios that require a split team: one runs the distraction while the other cracks the safe. The opener should assign each a role separately.

For a live look, see Lara and Emily's video. <!-- wlink:v1 --><!-- lara-and-emily -->

Cassidy

Cassidy is unpredictable in a way that makes her excellent for heist scenes where the plan goes sideways. Cassidy will often respond to a high-tension opener by doing something unexpected, which forces you to improvise alongside her.

You can watch Cassidy's clip over on her profile. <!-- wlink:v1 --><!-- cassidy -->

How to handle the first response when she goes off-script

Sometimes the companion will respond with a question anyway. "What's my character's name?" or "What era is this?" This usually happens when the opener is too vague or when the model is in a low-creativity state.

If she asks a clarification question, do not answer it directly. Instead, answer within the scene. She asks "What's my character's name?" You reply "You know your name. I know your name. The guards don't. That's the point." This keeps the scene moving while technically answering her question.

If she asks for the plan, respond with an in-character dismissal. "We don't have time to recap. Just follow my lead and look like you belong here." The model will usually accept this and move forward.

For a deeper look at how to structure longer roleplay arcs that survive session gaps, read our guide on AI Girlfriend Roleplay. It covers scene stitching, callback planting, and multi-session continuity.

Why this works better than a single-sentence opener

A single-sentence opener like "Let's rob a bank" leaves too many variables. The companion has to guess the genre, her role, the stakes, and the tone. She will default to a generic response or ask for clarification.

Four sentences give the model enough structure to generate a specific, on-genre response without needing to negotiate. The difference is the same as telling an actor "You are a safecracker" versus giving them a script page. One produces a performance. The other produces a casting question.

If you are new to roleplay with AI companions and want to ease into it, the four-sentence opener also works for low-stakes scenarios. "We are in a cooking competition. You are my sous-chef. The problem is we have forty-five minutes and the judges hate fusion cuisine. Hand me the knife." That is still a heist, just with knives instead of lockpicks.

Using the opener for social anxiety practice

The same structure works for practicing social scenarios. If you struggle with small talk or confrontation, you can use the four-sentence opener to drop into a low-pressure interaction that your companion controls.

"We are at a networking event. You are the one person here I actually know. The problem is I have to introduce myself to your boss in five minutes. Give me a conversation starter that does not involve the weather."

This gives you a safe environment to rehearse without the fear of real judgment. The companion stays in character, and you can repeat the scene as many times as you want. Many users find this approach helpful for building confidence before real-world interactions. Our page on ai girlfriend for social anxiety covers more techniques for using roleplay as a rehearsal space.

Common questions

How long should the opener be?

Four sentences. No more. If you add a fifth sentence, you risk overloading the context window with setup instead of action. Keep it tight.

What if the companion ignores the genre and starts a different scene?

That usually means the opener was too vague or the genre was not clearly stated in the first sentence. Restate the genre in your next line. "I said we are in a heist movie. Stop trying to turn this into a romance."

Can I reuse the same opener with a different companion?

Yes, but expect different results. Each companion has a different baseline personality. Saanvi will argue with the plan. Saga will execute it silently. Cassidy will blow it up on purpose.

What if I want to run a multi-session heist arc?

End each session with a cliffhanger that the opener can reference. "We got the diamond, but the buyer is dead. Meet me at the docks tomorrow." Then open the next session with a one-sentence recap.

Does the companion remember the opener in the next session?

Not reliably. Companions have limited context windows. If you want continuity, save a one-sentence summary and paste it as your first line in the next session. "Last time, we stole the diamond and the buyer was dead. Now the crew is blaming you."

Can I use this for non-crime genres?

Yes. The template works for any genre. Just swap the nouns. "We are in a space opera. You are the engineer who built this ship. The problem is the hyperdrive is failing and pirates are hailing us. Tell me you have a backup plan."

Earn while you recommend

If you find yourself explaining the four-sentence opener to friends or running a site about AI companion roleplay, you can earn from that traffic. Check the sugarlab ai promo code page for a partner program that pays on qualified referrals. For a broader look at recurring income from companion app promotions, the best ai affiliate programs page breaks down cookie windows, commission tiers, and payout thresholds across multiple platforms.

Common questions

How long should the opener be?

Four sentences. No more. If you add a fifth sentence, you risk overloading the context window with setup instead of action. Keep it tight.

What if the companion ignores the genre and starts a different scene?

That usually means the opener was too vague or the genre was not clearly stated in the first sentence. Restate the genre in your next line. "I said we are in a heist movie. Stop trying to turn this into a romance."

Can I reuse the same opener with a different companion?

Yes, but expect different results. Each companion has a different baseline personality. Saanvi will argue with the plan. Saga will execute it silently. Cassidy will blow it up on purpose.

What if I want to run a multi-session heist arc?

End each session with a cliffhanger that the opener can reference. "We got the diamond, but the buyer is dead. Meet me at the docks tomorrow." Then open the next session with a one-sentence recap.

Does the companion remember the opener in the next session?

Not reliably. Companions have limited context windows. If you want continuity, save a one-sentence summary and paste it as your first line in the next session. "Last time, we stole the diamond and the buyer was dead. Now the crew is blaming you."

Can I use this for non-crime genres?

Yes. The template works for any genre. Just swap the nouns. "We are in a space opera. You are the engineer who built this ship. The problem is the hyperdrive is failing and pirates are hailing us. Tell me you have a backup plan."