the official Fanvue AI creator program*
🔒 Heads up — finish the previous lesson first. The course works in order.

Phase V · The Vault · ⏱ 15 min

Scene Swap

When you're done

By the end of this lesson you will have:

Why this matters

Scene swap is the only Bopgpt tool whose input is a whole photograph. Every other tool builds a scene from your text and cards; this one copies a scene from a photo you hand it. That makes it useful in two specific jobs, and risky in two specific ways.

The two jobs. Repair: you have a keeper whose composition you love, but her face drifted or her signature trait dropped. Swap her locked identity back into that exact frame instead of rebuilding the scene from scratch. Reuse: one generic scene you generate once can serve her today and a second creator later.

The first risk is legal. Because the input is an actual photo, a careless reference choice can copy a real person's likeness or someone's paid content. The rights rules in What you need to know close that door, and they are the most important content in this lesson.

The second risk is editorial. Every swap borrows someone else's lighting, location, and styling. A few well-aimed swaps read as a girl with taste. A feed that is mostly swaps stops looking like the one recognizable woman you built in lesson 1-1 (Pick Her Niche & Persona) and lesson 1-2 (Create Your AI Model), because the feed's look is coming from other people's reference photos. The grid test in step 6 is how you catch that drift before fans do.

You met the ownership rules in lesson 3-2 (Research: Find What Works), where you remade patterns from your research saves. Here they bite harder, because this tool does not take a pattern. It takes the photo itself.

What you need to know

Everything in this lesson happens in Bopgpt: the Create tab (Scene swap mode), the Models tab (stand-in models), and her Library. Scene patterns you want to remember go into the same platform saves your research runs on (lesson 3-2 (Research: Find What Works)).

Terms used below:

What this lesson spends:

Generation Cost
One Scene swap output coming soon
One card-shoot image, for comparison 5 credits (the lesson 2-1 (Your First AI Photoshoot) price)

Until the swap's cost is printed here, do not assume it matches the card-shoot price. If your balance cannot cover a swap, top up or wait. A thin balance is never a reason to reach for a riskier reference.

The rights rules

Every reference must clear all three. Breaking them risks a real person's likeness, a copyright owner's license, or another creator's paid product, and any of those can end her accounts and this business.

The editorial filters

A reference that clears the rights rules still has to pass these two before you spend a credit:

A reference must come from one of the two sources in step 1 and clear all five filters, or it does not get uploaded. The whole check takes about ten seconds once it is habit.

The steps

  1. Pick the reference from one of the two allowed sources. In Bopgpt: either generate a generic scene yourself on a stand-in model (open Models, then + New model, build a model you will never publish, and shoot the setup you want), or pick one of her own Library keepers for a repair job. If you are at your plan's creator cap, skip the stand-in and build the new scene directly on her with cards using the lesson 3-3 (The Content Factory) recipe; the swap earns its keep on repairs and reuse, where the reference already exists.

  2. Run the five filters out loud. On your computer, before uploading: the three rights rules, then the two editorial filters. Say each one and answer it. If any answer is no, the reference is out, whatever it looks like.

  3. Run the swap. In Bopgpt:

  4. Open the Create tab and pick Scene swap (lesson 2-1 (Your First AI Photoshoot) walked this screen).
  5. Upload the reference photo. Well-lit, one person, clean composition swaps best.
  6. Confirm your model is selected, then generate.

The output is a single image of her in the reference's scene, with the reference's outfit and setting and her face and identity. Run it again for variations.

  1. Judge the output with the keeper rule plus one extra check. In Bopgpt, on the output: run the 10-second keeper rule from lesson 2-1 (Your First AI Photoshoot) (hands, background, face consistent with her reference). Then the check card shoots never needed: confirm her signature trait is still there. Scene swap keeps the reference's outfit, so if her trait from lesson 1-1 (Pick Her Niche & Persona) is something she wears, the swap can replace it with whatever the reference person had on. A swap that erases the trait is a photo of a stranger who looks like her; reroll it, or pick a reference where the trait would naturally show.

  2. Learn your swap keeper rate over the first five swaps, and obey the stop rule. In Bopgpt: notice how many of your first five swaps you kept. There is no log to fill; her Library shows what you kept against what you generated. That fraction is your planning number. If it is one keeper in three, a remix slot that needs one swap costs about three runs, so budget credits by it. The stop rule ends every reroll spiral: two misses on the same reference means you retire the reference. A third attempt on a bad reference spends credits on hope.

  3. Keep swaps in their lane: remix slots only, and run the weekly grid test. On your computer, when you plan the day's production: Scene swap (lesson 2-1 (Your First AI Photoshoot)) is a remix tool for the patterns you find with lesson 3-2 (Research: Find What Works)'s method, and it earns a slot only when today's pattern is a scene type. Her card-built content stays the spine of the feed. On many days the right number of swaps is zero, because a hook, motion, or style pattern belongs to More-Like-This (lesson 3-5 (Carousels & More-Like-This)) or the reel tools (lesson 3-4 (Reels That Feel Real)). Once a week, open her profile grid and try to spot the swaps from the thumbnails alone. If you can spot them by lighting, palette, or styling, a fan already has, and next week's remix slots go back to cards until the grid reads as one girl again.

  4. Collect scene patterns inside the research block you already run. On Instagram, X, or TikTok, during the daily research 15 from lesson 3-2 (Research: Find What Works): when a scan turns up a scene type that would pass all five filters, save it to the same platform collection as every other pattern, in the same motion. Save the pattern, never the post as a reference. When a remix slot later calls for that scene, make the reference yourself from step 1's two sources. Seed the habit today by saving five scene patterns that fit her life.

AI prompts for this lesson

Paste these into Bopgpt's AI chat tab.

Here are my creator's details: [her name, niche, backstory and her want, voice words, her look, her signature trait]. List 15 GENERIC scene types she would plausibly be in, based on her backstory. Describe each as a scene only: setting, light, pose type (e.g., "café window seat, morning light, seated three-quarter"). Describe places and setups, never people: no names, no real people, no creators, no stock-site or search-engine phrasing. No locations anyone owns, nothing she couldn't afford or wouldn't do. Flag any scene that would hide her signature trait.
Here are one-line descriptions of her last 12 posts, in grid order, marked [card] or [swap]: [paste]. Act as a blunt photo editor. Does this grid read as one consistent person? Name any post whose lighting, palette, or styling breaks the set, and tell me plainly whether swaps are taking over. One-line verdict: keep swapping, or cards only next week.

Do it now

You're done when

One clean scene-swap keeper, signature trait intact, is in her Library, and five scene patterns that fit her life are in your saves. Your swap keeper rate finishes measuring itself over the next four swaps.

If you're stuck

The swap looks off: wrong proportions, weird blend

Reference quality drives swap quality. One person, clear lighting, simple background; busy or low-resolution references produce mush. After two misses on the same reference, retire the reference.

My swap keeper rate is way below my card shoots

Expected. A card shoot builds the scene around her face; a swap drops her face into light set up for someone else. Use your measured rate to budget runs, and favor references with soft, even light (window light, overcast outdoors), which give the tool less to reconcile.

Her signature trait keeps disappearing

The swap keeps the reference's outfit, so if her trait is jewelry or clothing, the reference person's styling wins. Two fixes: pick references where she would plausibly wear the trait (a neckline that shows a chain), or keep trait-forward posts card-built and use swaps for scenes only. A single post without the trait is fine; a feed without it is a different woman.

Can I swap her into scenes with other people in them?

No. One-person references only. Group scenes swap badly, and the other people in frame put you back in likeness territory.

Can I swap her into premium or NSFW scenes?

Two answers. First, whether Scene swap honors the SFW/NSFW rating toggle is still being confirmed:being confirmed. Second, the rules that hold either way. An explicit reference is almost always someone's paid or leaked content, and always a real person's body, so explicit references are banned outright. Premium content follows the ladder from lesson 5-2 (The Content Ladder): Premium and Top rungs are Fanvue unlocks only, never on socials, never in off-platform DMs, and never above her ceiling. Posting explicit content on a social platform gets that account banned. Premium sets get built with cards, never borrowed from someone else's photo.

Can I use real stock photos as references?

No. "Generic" describes the scene type, never the photo. A stock photo has a real model in it, and the swap keeps that photo's outfit and scene, plus the image carries a license that governs its use. Save the scene type to your collections, then build the reference yourself from step 1's two sources.

I found the perfect reference, and it's from a creator I follow

Then it is the wrong reference. Knowing whose it is proves the second rights rule fails: someone else would recognize it too. Save the scene type as a pattern, generate the setup yourself or pull one of her keepers that already has it, and swap that.

How many swaps should go on her feed?

There is no quota; the grid test governs. Swaps fill remix slots only on days when the pattern is a scene type. If the weekly grid test can spot them, run cards only until it cannot.

Next up

lesson v-2 (Upload Photos: Real-Model Creators) covers the other creation path: building a creator from a real, consenting person's photos, and the consent gate that makes it legal. Read the gate even if you never plan to use a real model; it is the most important rule in the Vault.