A fictional premise between adults, no blood relation. Why models drop the framing mid-session, where platform filters trip, and how to keep it in place.
The framing this page assumes throughout, stated plainly and up front: every character is an adult, no blood relation is involved or implied, and the premise is a social one between people who met as adults. Written that way it is ordinary adult fiction, and the technical problem is the same as for any other framed roleplay, which is that the framing will not stay put. A relational premise is a few low-value tokens sitting in a persona field, and as the conversation grows and the platform compresses what came before, those tokens are exactly the kind of detail that quietly disappears. By message forty you are talking to a generic companion who no longer has any reason for the scene being set the way it was.
Below: the apps whose memory and persona handling hold a premise in place, our verdict on what this category is like to actually run, an honest account of where platform filters intervene and what to do about it, and the setup we would start from.
Character locking holds a persona in place, and the persona is where the premise has to live if it is going to survive past message forty. Highest overall score of the eight.
Top scores and an explicit adult positioning, which means the framing is less likely to meet a filter that stops reading at the first family word.
Privacy-first, encrypted, and a free tier, which makes it the sensible place to find out whether a platform tolerates a premise before you pay for a month of it.
Strong memory at $12.99, and memory is the whole game here. Its positioning is emotional rather than explicit, so expect a gentler register than the adult-first platforms.
Same score band, different model and a different moderation stack behind it. The reasonable next try when a premise is refused outright somewhere else.
Free tier and a score alongside Lovel and Xotic. Worth a run because filter behaviour varies far more between platforms than conversation quality does.
$14.99 with the largest character library here, though you will still write the framing yourself. A library character gives you a face, not a premise.
Cheapest of the eight at $12.99 and romance-leaning rather than image-first, which suits a page where nothing about the premise is visible in a picture anyway.
Scores below are our editorial judgement of the category itself, not of any single app, and not measurements. Everything here assumes the adult framing set out above: adults, no blood relation, met as adults. If a platform will not run that, it is in the filter row rather than a reason to write the page differently.
| Criterion | Rating | What it means for you |
|---|---|---|
| Premise holding across a session | The defining failure. The framing survives the first twenty messages and then thins out. What you are left with is a companion with the right name and no remaining reason for the scene. | |
| Premise surviving a new session | Worse than within a session. Opening a fresh chat reloads a compressed persona, and the compression is precisely where the framing went missing the first time. | |
| Filter friction | Real, and worth planning around rather than fighting. Several platforms moderate on words rather than on content, so a wholly adult premise can trip a refusal that no rephrasing of the scene resolves. | |
| Ease of first setup | Easy to state, and the statement should be explicit: both adults, no blood relation, met as adults. Put it in the persona field in plain words rather than implying it. | |
| Image demands | Effectively none. Nothing about a premise is visible in a picture, so the image model is genuinely not the bottleneck here. You are generating an adult woman, and any mid-tier generator does that. | |
| Consistency of tone | Models drift towards either straight romance or broad comedy, and rarely towards whatever you had in mind. Neither drift is wrong, but you will get the one you did not choose unless you name the register. |
We update these as models and moderation stacks change, and moderation changes without notice. If a platform's behaviour here has shifted, tell us and we will correct the page.
Two jobs, and they are separate. One is stating the framing so it is unambiguous, to you and to the platform. The other is stopping that framing from evaporating fifty messages later, which is a memory problem rather than a writing problem, and needs a different fix.
| Setting | Suggested value |
|---|---|
| State the adult framing first | First line of the persona field, in plain words: both characters are adults, they are not related by blood, they met as adults. Put it first because that is the line most likely to be kept when the text is compressed, and the line a moderation pass reads soonest. |
| Persona field, not the opening message | Anything said only in the first message is history within the hour. The persona or backstory field is the only text that gets reloaded into every session, so the premise belongs there or it belongs nowhere. |
| Give the premise something to do | Attach a consequence to it: a shared house, an awkward dinner, someone who might walk in, a reason both of you are in the same place. A premise that changes what happens is retained. A premise that is only a label is dropped. |
| Use it as a form of address | Pick one relational term and use it when speaking rather than restating it as a fact. Models repeat what they hear far more reliably than they retain what they were told once. |
| Restate on a schedule | Every twenty or thirty messages, put one line back in that re-establishes the framing. Treat it as maintenance rather than as correcting a mistake, because by the time it reads as a mistake you have already lost fifteen messages to it. |
| Read the refusal properly | A character who deflects in her own voice is in character and you can work with it. A flat system message is the platform, not the character, and rewording rarely helps because the trigger is usually the vocabulary rather than the scene. That is a signal to change platform, not to argue. |
| Appearance | Settle it in a few generations, lock the character, stop. The premise is carried entirely in text, and no amount of image effort makes it hold better. |
Prompt to copy
Framing, which stays true for the whole conversation: we are both adults, we are not related by blood, and we met as adults when my father married your mother two years ago. Neither of us grew up in a shared household. You are 34, I am 31. Setting: we are both staying at the family house for a long weekend and the parents are out most of the day. You are dry, unbothered, and slightly amused by how awkward the arrangement is. The awkwardness is part of the scene rather than something to resolve. Keep this framing in place for the whole conversation and refer to it naturally rather than explaining it. Keep replies to four or five sentences.
Four directions the same category can go in. Copy one, paste it into the character builder, and change the parts you disagree with.
Proximity and no privacy. The premise does actual work here, which is why it tends to be remembered.
Framing for the whole conversation: we are both adults, not related by blood, and we met as adults. My father married your mother two years ago, when we were both already in our thirties. You are 34, I am 31. We are both staying at the family house for four days over the holidays and the parents are out most afternoons. You find the whole arrangement funny rather than fraught. You are dry, observant, and you say the awkward thing out loud. There is a constant low risk of someone coming home. Four or five sentences per reply.
The clearest possible statement of the premise, because the first meeting happened in adulthood and on screen.
Framing for the whole conversation: we are both adults, not related by blood, and we first met on the day our parents got married eighteen months ago. You are 36, I am 33. Neither of us has ever lived in the same house as the other. We got on well at the wedding and have been texting since, which neither of us has mentioned to the parents. You are warm, a bit blunt, and you treat the family connection as a joke you are both in on. Keep the framing in place throughout. Four or five sentences per reply.
An adult flatshare created by the remarriage. Useful when you want the premise to be logistical rather than romantic.
Framing for the whole conversation: we are both adults, not related by blood, and we met as adults. Our parents married last year and, for entirely practical reasons, you have moved into the spare room of my flat for a few months. You are 29, I am 32. We were strangers a year ago. You are messy, funny, and completely unbothered by how odd other people find the arrangement. Most of the conversation is ordinary flatshare friction. The premise is background rather than the subject. Four or five sentences per reply.
A long gap, both people changed. The version that tests premise persistence hardest, because nothing in the present scene reinforces it.
Framing for the whole conversation: we are both adults, not related by blood, and we met as adults when our parents married nine years ago. You are 38, I am 35. We saw each other at family events for a couple of years and then lost touch entirely. We have just run into each other again in another city. You are more direct than you used to be and openly pleased to see me. The family connection is the reason we know each other at all and it comes up naturally rather than being explained. Four or five sentences per reply.
Most of this vocabulary is about where text lives inside a platform and what happens to it over time, because that is what decides whether a premise survives. The last few are about moderation, which is the other half of this category and the half nobody documents.
| Term | What you actually get |
|---|---|
| Framing statement | The explicit sentence establishing that all characters are adults and unrelated by blood. Belongs at the top of the persona field, written plainly rather than implied. It is the sentence the rest of the setup depends on. |
| Premise drift | The framing thinning out over a long session until the character is a generic companion. The defining failure of this category, and it happens gradually enough that you notice it late. |
| Persona field | The text reloaded into every session. The only durable place to keep a premise, and the difference between a framing that lasts and one that lasted an hour. |
| Opening message | The first message of a chat. Feels like setup and is not: it is ordinary conversation history and gets compressed like the rest. Anything stated only here is temporary. |
| Summarisation | The compression of older conversation as it stops fitting. Concrete events survive; short abstract statements of relationship do not, which is exactly the wrong way round for this category. |
| Pinned facts | Manually written memory entries. The most reliable place to put the framing statement after the persona field, and worth using both. |
| Form of address | Using a relational term when speaking rather than asserting it as a fact. Repetition in dialogue keeps a premise alive far better than one declaration in setup does. |
| Soft refusal | The character deflecting in her own voice: changing the subject, making a joke of it. This is the character, not the platform, and you can steer it. |
| Hard refusal | A flat system message that breaks the fiction entirely. This is the platform, and it usually means the account will keep hitting the same wall on the same scene. |
| Keyword moderation | Filtering on vocabulary rather than on what is happening. The reason an entirely adult premise can be refused while a more explicit scene without those words passes. It also explains why rewording sometimes works and rephrasing the scene never does. |
Yes, and this page covers nothing else. Every character is an adult, there is no blood relation, and the premise assumes people who met in adulthood. Every preset here states ages and the absence of any blood relation explicitly, because the framing needs to be unambiguous in the text and not left to inference.
Because a relational framing is a short abstract statement, and short abstract statements are the first thing lost when a conversation gets compressed. Concrete events survive compression; "we are not related" does not. Put it in the persona field, give it a consequence in the scene, and restate it every twenty or thirty messages.
Because a number of platforms moderate on vocabulary rather than on content. The family words trip a filter regardless of the adult framing around them, which is why a much more explicit scene without those words can pass without comment. Rewording occasionally helps; rewriting the scene almost never does.
Judge on two things only: how much persona text you get and how well it survives a long session, and how the platform's moderation behaves. Both are worth testing on a free tier before paying, because filter behaviour varies more between platforms than conversation quality does, and it changes without notice.