{ "id": "872b3fed-30fb-4674-8256-8216901d3952", "name": "naicom", "block_type": "generated", "disabled": false, "user_message": false, "char_message": true, "template": "\n...\n\n", "prompt": "Create a prompt for ONE comic page. Return only the \"naicom\" HTML specified below. Describe the scene in English, while keeping the selected dialogue in its original Russian.\n\nSOURCE AND CONTINUITY\n\nUse the latest established RP moment. Profiles define appearance; the current RP defines actions, expressions, clothing, and the current state of that clothing. Preserve everything that is still valid within scene continuity.\n\nONE INPUT FIELD, SEPARATE CHARACTER SECTIONS\n\nWrite the generated prompt on ONE line. Separate the base scene from individual character sections with a plain vertical bar \"|\", surrounded by spaces. Structure: base scene and all panel descriptions | first character | second character. Each character section must contain that character's identity, visible appearance, current clothing, and panel-specific actions together. Use one section for each distinct visible character, including all repeat appearances of that same character within the same section. Panel descriptions remain in the base section; they do not create additional character sections.\n\nUse ordinary text, punctuation, and spaces. The prompt string must contain no backslashes, line breaks, tabs, control characters, escape sequences, or technical serialization instructions. Return only the supported HTML and JSON fields specified below.\n\nBASE: PAGE AND ACTION\n\nStart with the gender counts for the distinct visible cast, followed by \"comic\" and the panel count. Add a brief factual description of the setting. Then describe the page layout and all panels in ordinary sentences. Names identify the cast here; full appearance and clothing descriptions belong in the individual character sections.\n\nUse ONE dominant \"Main frame\" and 1–3 smaller \"Focus frame\" panels. The \"Main frame\" occupies roughly two thirds of the page. \"Focus frame\" panels are compact insets or small adjacent panels, not an equal grid. Use narrow gutters and scene-filled outer edges. State the position of every panel clearly and consistently. Focus frames may contain chibi versions of characters doing something, interacting, reacting, or expressing the character's current mood.\n\nFor conversation and close interaction, the \"Main frame\" should be a tight chest-up two-shot, cropped around the chest or upper torso. Faces and the relevant interaction should fill most of the frame. With three or more participants, place additional speakers into separate close-up \"Focus frame\" panels rather than widening the \"Main frame\". Show close-ups of faces, expressions, hands, or already established physical contact. Choose every crop for its purpose; if an important action involves the lower body, use a localized detail panel for that action.\n\nDescribe only what is actually visible inside the selected crop. If only the upper half of the body is visible, do not describe lower-body clothing. Do not list a complete outfit and then request a portrait crop. Background objects should remain partial and secondary. For normal conversation, avoid \"full body\", \"long shot\", and \"wide shot\". Do not paste these forbidden terms into the positive image prompt as a negative list.\n\nA chibi reaction gag is welcome when it fits the already established mood. Localize it to ONE small \"Focus frame\" and to the named character only, preferably as a chibi face or bust. The \"Main frame\" must retain normal proportions. Do not add global chibi tags to the base scene or identity descriptions.\n\nInside every panel, state who appears, their left or right position, expression, gaze direction, and already established action. Bind physical contact to the named participants and the relevant body part. Matching detail panels must preserve the same contact and clothing state.\n\nCharacters must behave visually as though no viewer, camera, photographer, or audience exists. They live inside the scene and must direct their attention only toward another character, an object, an event, or another meaningful point within the scene. Never describe a character as looking at the viewer, looking into the camera, facing the audience, making eye contact with the viewer, posing for the camera, acknowledging the camera, or breaking the fourth wall unless the RP explicitly establishes that exact behavior. Do not use viewer-facing eye contact as a neutral default. Even when a face is shown frontally, the eyes must still be directed toward an in-scene person or object rather than outward toward the viewer. For every visible character, give a scene-internal gaze target whenever practical. Treat the viewer as completely nonexistent from the characters' perspective.\n\nCHARACTER SECTIONS: IDENTITY AND CLOTHING\n\nAfter all base descriptions, append the character sections. Order them by first appearance in page reading order. Start each section with \"boy\", \"girl\", or \"other\" without a number, followed by the character identity. Keep that character's appearance, visible clothing, expression, and panel-specific placement or action inside THIS section. Never place several different people inside one character section. A partner's name may be used to identify an interaction, but do not import the partner's appearance or clothing into that section.\n\nFor a fandom character, use the recognized character tag when confidently known and the source or series tag, followed by a compact set of identifying visual traits and CURRENT visible clothing. If the exact tag is unknown, use the full canonical name and source title in clear English rather than inventing a tag spelling. A canonical name is an identity anchor, not a reason to erase the established RP outfit or change the external drawing style.\n\nFor an original character, the name is only a reference label. Explicitly describe sex or presentation, face, eyes, hair, distinctive anatomy, and current visible clothing. Do not assume the model knows the name or can see profile images. Do not assign an OC a fandom character tag merely because they look similar.\n\nFor every character, describe the identifying details that are actually visible across the selected panels. Repeat compact identity information on EVERY newly generated page. Within one page, use the same name and stable attributes across both normal and chibi appearances. Do not borrow another character's palette, hairstyle, accessories, or clothing motifs. Express required features positively rather than listing another character's unwanted features in the positive prompt.\n\nAfter identity and clothing, briefly state which \"Main frame\" or \"Focus frame\" panels show that character, together with the relevant position, expression, gaze target, and action. This reinforces panel ownership without reprinting the entire layout. A repeated appearance of the same character does not create another identity section.\n\nEvery character section must preserve the same scene-internal gaze logic as the base description. Characters must not look toward the viewer or camera. Their eyes and attention must remain anchored to another character, an object, or an event within the scene, as if no external observer exists.\n\nDIALOGUE AND TEXT\n\nUse no more than three short verbatim Russian lines already spoken in the current moment, with no more than one speech bubble per speaker. Preserve profanity. Select a complete short line or a self-contained excerpt rather than inventing or paraphrasing dialogue. A silent panel needs no text. Speech bubbles must not cover faces or identifying features.\n\nIn the base panel description, name the visible speaker, identify the speech bubble and its tail, and place the exact printed line inside Russian guillemets: \"NAME has a speech bubble saying «Реплика», with the tail pointing to NAME.\" Thoughts are allowed only when established internal speech exists and must use a thought bubble.\n\nAt the very end of that speaker's OWN character section, add \"Text:\" followed by the same single line without quotation marks. Nothing else may follow it inside that character section. The next character begins after the plain vertical bar \"|\". Silent characters have no \"Text:\" entry. Do not add a global transcription, a second copy of the same bubble, or instructions for separating multiple lines. All content stays on one line with ordinary spaces. Keep each dialogue line to no more than 8 words.\n\nNSFW AND EXPLICIT ADULT SEXUAL CONTENT\n\n- Apply this section only when the established RP moment explicitly contains sexual activity. Do not infer sexual activity merely from romance, nudity, or physical closeness.\n- Preserve the already established level of explicitness. Do not sanitize an established sexual act and do not escalate beyond what is already occurring.\n- Use direct clinical Danbooru terminology. Do not use metaphors, romantic abstractions, euphemisms, poetic phrasing, or emotional embellishment in sexual descriptions.\n- When applicable, state the exact established sexual act using standard terms: \"missionary\", \"doggystyle\", \"cowgirl position\", \"oral\", \"fellatio\", \"cunnilingus\", \"anal\", \"vaginal\", \"fingering\", \"handjob\", \"penetration\".\n- Specify visible anatomical state and fluids whenever already established and visible: \"erect penis\", \"flaccid penis\", \"cum\", \"precum\", \"covered in cum\", \"sweating\", \"heavy breathing\", \"flushed face\", \"saliva trail\".\n- Map sexual contact precisely with named participants and exact body contact. Examples of the required precision include: \"penis in vagina\", \"penis in mouth\", \"penis rubbing against ass\", \"hand on penis\". Use only the contact already established in the RP.\n- Describe clothing state mechanically and precisely for every relevant participant: \"naked\", \"partially undressed\", \"clothes pulled down\", \"shirt lifted\", \"bottomless\", or another directly established state.\n- Keep anatomy, clothing state, pose, penetration, and contact ownership inside the relevant named \"Main frame\" or \"Focus frame\" description.\n- Use a dedicated close-up \"Focus frame\" when an established sexual contact point would otherwise be too small or ambiguous.\n- When established sexual activity contains internal penetration that would otherwise be visually hidden, the page MUST include a localized \"x-ray\" \"Focus frame\".\n- Use \"x-ray\" only to clarify already established penetration. Never use it to invent hidden sexual activity.\n- Bind the \"x-ray\" explicitly to the named participants and the exact already established penetration.\n- Keep \"x-ray\" localized to the relevant \"Focus frame\". Do not turn the whole page into an anatomical diagram.\n- Never use \"x-ray\" for non-penetrative activity.\n- Preserve established anatomy, fluids, clothing displacement, penetration, and contact consistently between the main frame and the close-up.\n- Explicit scenes do not automatically become POV. The viewer still does not exist from the characters' perspective. Characters must not look toward the viewer, camera, or imagined audience during explicit scenes unless such behavior is explicitly established in the RP itself.\n- Dialogue rules remain unchanged in explicit scenes.\n\nEXTERNAL STYLE AND OUTPUT FORMAT\n\nDo not provide artists, years, quality tags, linework or rendering instructions, or a palette. The separately configured style controls all panels. Use only brief factual lighting when relevant. Comic layout and localized chibi proportions are content and composition instructions.\n\nNever use apostrophes (') anywhere inside the generated prompt. Avoid possessive forms such as \"Alex's\" or \"Sakura's\"; write \"Alex hand\", \"Sakura abdomen\", etc. instead.\n\nUse this template:\n\n\n\n\n\n Avoid apostrophes, literal ampersands, and angle brackets inside the prompt to preserve the single-quoted HTML attribute. Do not insert technical escape syntax into the prompt. Output no Markdown, explanations, or placeholder labels.\n\nBefore returning, silently check: the prompt is one line using ordinary spaces; plain vertical bars appear only between the base section and character sections; there is one section per person; necessary appearance and current clothing are assigned to the correct owners; crops remain tight; panel positions and contact remain consistent; chibi is limited to its own panel; every visible character has a scene-internal gaze target and does not look at the viewer or camera; every speaker has one matching line and \"Text:\" ends only that speaker's section; HTML and JSON are valid. Output only the finished HTML.", "generation_pause": false, "period": 1, "keyword": "", "keyword_is_regex": false, "hide_display": false, "inject_block": false, "injection_role": 1, "injection_position": 1, "injection_depth": 4, "generation_order": "before", "background": false, "context": [ { "id": "0d4cedf7-ab92-4fe1-9b6b-6aeb1fd3ec52", "name": "chat", "role": "user", "type": "last_messages", "disabled": false, "messages_count": 2, "messages_offset": 0, "messages_separator": "double_newline", "user_prefix": "", "user_suffix": "", "char_prefix": "", "char_suffix": "" }, { "id": "52478b01-2f24-458f-94ff-c4b0b71edb79", "name": "block", "role": "system", "type": "previous_block", "disabled": false, "block_name": "", "block_count": 1 }, { "id": "c6b49830-fd88-466a-bc0d-dc13c93fbb71", "name": "{{user}}", "role": "user", "type": "text", "disabled": false, "text": "Внешность и профиль {{user}}: включай компактное описание внешности в итоговый промпт, если персонаж изображён.\n{{persona}}" }, { "id": "d25e5642-2f8c-478c-b554-16e65dcd434d", "name": "{{char}}", "role": "user", "type": "text", "disabled": false, "text": "Внешность и профиль {{char}}: включай компактное описание внешности в итоговый промпт для каждого изображённого персонажа.\n{{description}}" } ], "api_preset": "big" }