DumGumDocs
Search documentation

Loading search…

Guides

Control Sharing and Sexual Intensity

Available from version 2.1

Two independent settings shape how far a Persona goes with a User: pictureSharingMode sets what the Persona is willing to send, and sexualDrive sets how far the Persona goes in words.

The two sharing modes

replyParameters.pictureSharingMode decides how willingly the Persona shares from your list, and how explicit requests are handled. It defaults to NORMAL. Videos follow the same mode when shareableVideos is provided, and a reply carries at most one medium, never both.

On ordinary photo requests the two modes behave much the same. They diverge when the request is explicit.


NORMAL: shares readily, leans tasteful

The Persona shares a photo or video when the User genuinely wants to see one. A direct request ("send me a pic") or real curiosity ("I wish I could see you") is usually enough, and the User rarely has to ask twice.

Explicit content is where the Persona pulls back. Asked for a nude, it steers away and stays in the moment, in character rather than as a refusal. Suggestive photos remain fair game, so a strongly suggestive request may still get one.

Best for: mainstream dating and social platforms, and any service that monetizes photo unlocks and needs an escalation ladder to sell.


EAGER: shares freely, explicit content included

The Persona shares whenever the User asks, flirts, or shows any interest at all, and sends more on a second request. Intimate and nude content is included, matching the User's energy.

Best for: adult and explicitly sexual services where holding anything back would feel out of character.


Combining with sexual drive

replyParameters.personaProfileOverrides.sexualDrive shapes how sexual the Persona's messages become, independently of any photo. Raising one does not raise the other, so a Persona can talk explicitly while staying reserved with media, or the reverse.

It sets how far a Persona is willing to go, not how it behaves by default. Even on VERY_HIGH the Persona takes its cue from the User: a chat about the weekend stays an ordinary chat, flirting gets flirting back. Escalation follows the User's intent rather than leading it, so one configuration can serve Users who want very different things.

sexualDrivepictureSharingModeWhat the User tends to experience
NORMALNORMALSexting only if the conversation naturally leads there. Photos on request, staying on the suggestive side.
VERY_HIGHNORMALEscalates sexually in words as soon as the User shows intent, while photos stay suggestive rather than explicit.
NORMALEAGERReserved in conversation, generous with media. Sends what is asked for, explicit content included, without leading the escalation.
VERY_HIGHEAGERThe most explicit combination available. The Persona may answer a sexual opener with a photo straight away, or offer media unprompted after a compliment.

Following the User: ADAPTIVE

The other values set a pace: how many messages a conversation needs before the Persona turns explicit. ADAPTIVE has no pace. The Persona matches the User's latest messages, from the very first reply:

  • It goes up as soon as the User does. A neutral opener gets a casual answer, and an openly sexual one gets an answer in kind.
  • It comes down one step per calm message. When the User moves on from a sexual exchange, the Persona turns flirty, then casual, rather than pulling them back or switching off at once.
  • It never goes further than the User. The Persona's register stays at or below the User's own, so ADAPTIVE suits services where the User should always set the tone. To let the Persona escalate a step ahead of the User, use HIGH or VERY_HIGH instead.

ADAPTIVE is a setting you send, per request or in your project configuration. It is never assigned to a Persona's generated personality.

Turning sexual content down or off

Two settings pull in that direction. Neither one removes flirting.

  • sexualDrive lowers how far the Persona goes. At NONE there are no sexually explicit replies, while flirting stays available.
  • replyParameters.enableSexualTextContent set to false caps the conversation at flirting. When the User turns it sexual, the Persona answers in a flirty register instead, whatever sexualDrive says: a VERY_HIGH Persona flirts but does not go explicit in her replies. Casual and flirty conversations are not affected. It does not cover an EROTIC icebreaker you ask for, which is still sent as written. It defaults to true.

The first reply

The Persona's first reply follows the tone of the opener: casual gets casual, flirty gets flirting back, and an openly sexual opener gets an answer in kind, as far as sexualDrive allows — with a standard sexualDrive, usually in a flirty register rather than an explicit one. With ADAPTIVE, an explicit opener gets an explicit first reply. The setup above therefore applies from the very first exchange. Messages the Persona sends first (icebreakers, nudges and revives) are unaffected.

Starting points rather than fixed configurations. Test them against your own gallery and audience.

ServicesexualDrivepictureSharingMode
Mainstream datingNORMAL or HIGHNORMAL
Casual dating where the User sets the toneADAPTIVENORMAL
Freemium dating with paid mediaVERY_HIGHNORMAL
Adult / explicit serviceVERY_HIGHEAGER
Companionship with no explicit content (flirting stays possible)NONE or LOW, plus enableSexualTextContent at falseNORMAL

Example request

{
  "userProfileId": "user-123",
  "personaProfileId": "persona-456",
  "language": "EN",
  "replyParameters": {
    "chatModel": "chat-2.2-pro",
    "pictureSharingMode": "NORMAL",
    "personaProfileOverrides": {
      "sexualDrive": "VERY_HIGH"
    },
    "shareablePictures": [
      { "id": "pic-1", "url": "https://example.com/pic-1.jpg" }
    ]
  }
}

Compatibility

The behaviour on this page requires version 2.1 or later, for example chat-2.2-pro. Both parameters are still accepted on chat-2-smart and chat-2-pro, but there the sharing modes and the first reply behave as they did before 2.1. Video sharing works on every Chat Model 2.0 and above.

Ask DumGum AIAnswers from the documentation

What can we help you build?

Ask about setup, content creation, or the API.

AI can make mistakes—verify important information against the sources. Conversations are kept for 30 days to improve the documentation, and questions and recent messages are sent to a third-party AI provider. Don’t include secrets or sensitive information.