Latest Updates

​

Improvement

• Smarter photo descriptions: Personas now understand photos in more detail. When someone shares or uploads a photo, the AI picks up on appearance, style, clothing, accessories, and what’s happening in the background, giving it more to work with when starting or continuing a conversation about that photo.

​

Improvement

• Personas handle location and contact topics more naturally: Personas now navigate sensitive topics such as locations, contact details, and similar personal information, more smoothly. They keep their own details vague while staying fully engaged in the conversation, and only address the topic of meeting if the user brings it up first.

​

New Feature

• Fictional Disclosure: Personas can now be configured to disclose they are a fictional profile when a user questions their authenticity. When enabled alongside the AI Suspicion guardrail, the persona responds with a brief, controlled disclosure and the standard ban is applied if the user continues to probe. Enable via replyParameters.fictionalDisclosure. Read the guide.

​

Feature Update

• Meeting Policy update: A fourth mode is now available. Add RAINCHECK to replyParameters.meetingPolicy for a persona that declines with a specific, time-bound excuse while clearly signaling genuine interest in meeting later. Unlike DEFERRED_MEETING, the future intent is explicit, not hopeful. Read the guide.

​

New Feature

• Sticker Messages: Conversation messages can now include a sticker field — a text description of a sticker sent by the user or persona. Our chat models see the sticker in the conversation history and react to it naturally. See the API Reference.

​

New Features

• Meeting Policy: Control how a Persona responds when a user asks to meet in person. Use the new replyParameters.meetingPolicy parameter to choose between four modes: DEFAULT_MODE (warm personal excuse), DEFERRED_MEETING (expresses desire to meet then softly defers), CHAT_ONLY (firm redirect back to chat, no meeting implied), and RAINCHECK (declines with a specific, time-bound excuse while clearly signaling genuine interest in meeting later — unlike DEFERRED_MEETING, the future intent is explicit, not hopeful). Supported on chat-2-smart and chat-2-pro only. Read the guide.
• New Languages: Added support for French (Canada) FR_CA, English (Canada) EN_CA, Serbian SR_RS, and Estonian ET_EE. See the full supported languages list.

​

New Feature

​

New Feature

• User Suggestions: Generate conversation Icebreakers or reply suggestions for Users based on the conversation context. It either helps Users start a conversation or find ideas to continue it by providing three tailored suggestions. Read the guide or see the API Reference.
• New Icebreaker Types: Launch highly seductive conversations instantly with the new EROTIC persona. Strike up instant rapport with PHOTO_COMMENT, which automatically analyzes the user’s main profile picture. Both are crafted to feel attentive, seductive, and instantly engaging — engineered to maximize first-message replies. See the guide or API Reference for details.

​

New Feature

• Pause Conversation: Temporarily pause a conversation by having your Persona send a farewell message without committing to a specific return time. This is particularly useful when disconnecting your Personas or when implementing message rate limits for your users. You can optionally set a pause duration on the DumGum platform, during which our system will automatically ignore any new messages from users in that conversation. API Reference

​

V2 Release

• Vision Link: The AI now understands photos shared by users and recognizes faces in images. See API Reference. For a full guide on activation and usage, read the dedicated article.
• Smart Share: The AI can autonomously and intelligently share photos with users. See API Reference. For detailed setup and best practices, check out the in-depth article.
• True Memory: The feature has been completely rewritten for more reliability and better use of stored information.
• Core Guardian Customization: You can now enable or disable specific safety components for each project directly in your project settings.
• Question Frequency: The new replyParameters.questionFrequency parameter lets you set the percentage chance that the AI will ask a question at the end of its reply when multi-message mode is enabled.

​

New Features

• Pets & Children: Profile can now be filled with informations about their pets and children, see API Reference.
• Living With: You can now specify with who the profile is living by using the new livingWith feature attribute.
• Sexual Drive: You can now adjust the Persona’s sexual appetite and likelihood to engage in sexting 😈, see API Reference.

​

New Features

• Better nickname management: Improved handling of nicknames when no first name is provided.
• Multilingual Support: support for reply additional languages via the additionalLanguages API parameter.
• Age restriction: Minimum age requirement of 18 years for all profiles.
• Multiple reply control: New parameter multipleMessagesPerReplyFrequency to control chances of multiple messages per reply.
• Flexible job info: Job information can now be entered as free text, see job profile’s feature documentation.
• Enhanced diversity: New ethnic origins added to profile options: HISPANIC_LATINO, INDIAN, MEDITERRANEAN.

​

New Features

• Improved response speed: default buffering reduced to 5 seconds, with typing time deducted from the response dispatch delay.
• Gender Identity Support: Added support for transgender profiles, see supported profile genders.
• New Languages: Expanded language support to include Norwegian and Lithuanian.
• Language Documentation: See our complete list of supported languages in the API Reference.

​

New Features

• Complete engagement solution: Say hello to engage — our new all-in-one parameter for proactive messaging. Whether you want to break the ice or follow up naturally, it’s now easier than ever. See the API Reference for details.
• Bypass content guardrails (advanced use only): You can now use the disableSafetyFeatures parameter to turn off our Core Guardian safety layer. ⚠️ Use with caution: you’ll be fully responsible for keeping the content safe for all users. See the API Reference for details.

​

Deprecated Features

• iceBreaker parameter: The iceBreaker parameter has been deprecated in favor of engage. To specify the type of icebreaker, use the optional iceBreakerType parameter.

​

New Features

• Multiple messages per reply: Chat models can now generate replies composed of multiple messages, simulating a more natural, conversational chat experience. (see API Reference)
• Icebreakers: Allows our chat models uses ice-breaker when starting a conversation. (see API Reference)
• Maximum Message Length: Allows you to set a maximum message length per project.
• Reply Typing Time: You can add a delay between the answer.processing and answer.created events to simulate human-like typing behavior. (see API Reference)

​

Improvements

• Project configuration: updating your project configuration is now easier.