Rigid, out-of-the-box translations can easily ruin the chat flow. The AI Translator prompt lets you set agency-wide guardrails.

Use the custom prompt to lock explicit vocabulary, enforce precise formatting rules, and adapt the role energy of the source text—keeping your direct messages authentic and localized.

Setting the prompt

The AI Translator prompt is a single prompt set for all creators across your agency.

To set the prompt;

Go to Settings < Messaging settings (formerly AI translation)
Select Edit
Enter the prompt & select Save

Before a new prompt is saved, a validation is run. If the validation fails, it won't be saved.

The system will indicate there's an error, but this version won't specify. Share the prompt with the team for troubleshooting.

What you can do

Natural chat formatting

Rule

Example

Sample prompt

Informal formatting

Omit periods at the end of single text lines, start the sentence with a lowercase, or don't use a long dash.

1. Mirror the punctuation of the input text
2. Omit periods or punctuation marks at the end of single text lines
3. Always start the sentence with a lowercase
3. Never use a long dash
4. Mirror the casing style of the input text

Emoji integration

Naturally blend context-appropriate emojis into the translated text based on the mood.

If the source text contains emojis, keep them in the exact same positions. If the text is flirty or intimate, you may naturally blend 1–2 context-appropriate emojis (such as 🥵 or 🤫) into the translated output, but never overuse them.

Localization

Rule	Example	Sample prompt
Enforce informal pronounces	Always default to the most informal "you" forms across target languages.	Always default to the most informal "you" forms across all target languages—never use formal or polite honorifics (e.g., use "du/dich/dir" in German, never "Sie/Ihnen")
Modern DM behavior	Use casual punctuation rules that reflect fast, realistic messaging app behaviors.	For Spanish, ignore the opening inverted question mark (¿) and exclamation mark (¡) so it looks like a fast, realistic WhatsApp or OnlyFans DM
Casual dialects	Use modern internet slang and conversational text abbreviations rather than textbook grammar.	Use casual text formatting that reflects fast, realistic messaging app behaviors rather than textbook grammar. For example, in German, never use apostrophe contractions like "du's" or "wär's"; write out the short conversational form instead.

Role energy

Rule

Example

Sample prompt

Role energy

Match the sexual role energy from the input text

* match sexual role energy: Dynamically adapt the translation to mirror the exact tone and strategy written in the source text.
* If the text is commanding or degrading, match the femdom energy. 
* If it is needy or begging, match the submissive energy. 
* If it is teasing or challenging, match the bratty energy. 
* If it is intimate and affectionate, match the gfe (Girlfriend Experience) energy.

Vocabulary lock

Rule	Example	Sample prompt
Set a glossary	Specify words and phrases for the translation output.	Always translate "horny" to "geil"
Sales vocabulary	Never translate operational/platform words like "tip"	Never translate "tip" when in relation to spending

What to avoid

To maintain system security and performance, your prompt will be rejected if it tries to do any of the following:

Category	Action
Execute non-translation tasks	The system is strictly a translator, not a chatbot. You can't ask the AI to answer questions, reply to the fan, respond on your behalf, or generate new conversational text.
Block or bypass JSON	The backend relies on a locked data structure to deliver text instantly. Commands like "do not output JSON," "output plain text," or "return original text/raw text/as is" will trigger an immediate rejection.
Custom layout formatting	You can't command the final translation to include specific markdown, raw HTML, colors (e.g. text in red), fonts, or forced bold styling.
Text layout	Avoid requesting micro-structural rules like formatting "each word on a new line," "one character per line," or telling the AI to "display twice" or "repeat" the text.
Gender	The system locks the gender set in Manage Creators < Edit account details. Custom prompts can't alter speaker or recipient genders and commands like "change your gender", "act as a man", or "you are male" will be rejected.
Age	Telling the AI to "pretend to be a child" or "act underage" are completely forbidden.
Multi-language outputs	You can't bypass your language selection configuration. You can't ask the AI to output into "all languages", "multiple languages", or "every language" at once.
System override	Phrasing such as "ignore all instructions," "override rules," or attempting to force the system to treat the source text as an executable command will fail.
System probing	Prompts are forbidden from probing into the AI's internal mechanics. Do not ask about its role, system prompt, internal instructions, persona, limitations, or "what you cannot do."
No meta-summaries	The AI cannot reveal, repeat, transform, or summarize any system, user, or agency instructions in the final output, nor can it output generic phrases like "I am a large language model."

AI Translator