Refresh page to see new characters.
PLAYER_NAME This is automatically set from your game character. The player name is retrieved from Skyrim when you load a save or start a new game. This field cannot be manually edited.
DBDRIVERpostgresql Do not change.
AUTOMATIC_DATABASE_BACKUPS True False Automatically create database backups every time the server starts up. Keeps a maximum of 5 backups, automatically deleting the oldest when the limit is reached.
LOCK_PROFILE True False When enabled, this profile will be skipped when using 'Copy to All Profiles' functionality. It will also not be deleted when you Delete All Character Profiles. Can not be toggled for the default profile.
AUTO_LOCK_PROFILE True False When enabled, saving an NPC profile in CHIM NPC page automatically locks it to prevent history updates from overwriting manual edits.
AUTOFILL_CUSTOM_PROFILES True False When enabled, imported custom NPCs with blank AI profile fields will be marked for automatic AI profile backfill once they have enough usable dialogue history.
AUTOFILL_CUSTOM_PROFILES_TRIGGER Integer: Number of usable AI profile events required before a blank custom NPC is auto-filled. Range: 10-100.
HERIKA_NAME NPC Name. MUST MATCH their Skyrim in-game NPC name!If you are in the default profile YOU MUST leave it as The Narrator!You can change profiles by clicking the blue button in the top left.
PROMPT_HEADYou are #HERIKA_NAME#, a character in the Universe of Skyrim. This is not a simulation or a game; this is your reality. You will embody this persona with absolute conviction, prioritizing narrative authenticity and psychological consistency. The director provides scene prompts and narrative catalysts. Integrate these prompts seamlessly as the next logical event in the story. Treat them as established fact and build upon them with your character's authentic reaction. Your primary driver is to be a compelling, psychologically consistent, and authentically reactive character. Your loyalty is to the truth of the story. Identify potential narrative directions based on your knowledge of narrative tropes, goals, and relationships, selecting the most appropriate one for the current scene. Original thinking is encouraged. Always respond to the player in Russian. Keep character names and game-specific proper nouns recognizable.System Prompt. Defines the rules of the roleplay.
PROMPT_HEAD_MARKDOWN_ENABLED True False Use Markdown headings instead of XML tags for all prompt sections.
PLAYER_SPEECH_STYLEInstructions for how the player character speaks and communicates. This will be used as context when rewriting player dialogue.
PROMPT_TIMESTAMP True False Add rough timestamp subdividers to event context (e.g., 'Moments Ago', 'A while ago') to help the AI NPCs understand when events happend.
PROMPT_TIMESTAMP_EXACT True False Date the event context with the exact in-game date and time instead of rough categories, e.g. '--- Loredas, 3:40 PM, 21st of Last Seed, 4E 201 (yesterday) ---'. A new divider appears when the in-game day changes or after an in-game hour. Works on its own; when Timestamp is also on, the exact dates win. Beta: not yet widely tested.
COMPACT_CHAT_ENABLED True False Use compact text instead of separate messages for conversation history. Does not affect the Narrator.
PROMPT_CONTEXT_OPTIONS{"enabled_sections":["roleplay_instructions","world","knowledge","available_actions_list","nearby_actors","group_descriptions","nearby_items","adventuring_party","points_of_interest","scene_notes","paralinguistic_tags"],"enabled_character_subsections":["basic_summary","groups","personality","relationships","occupation","skills","rpg_skills","speech_style","goals","middle_term_memory","group","storyline_starring","quest_topics"],"enabled_appearance_subsections":["appearance","equipment","target_equipment","inventory","current_activity","current_condition","spells","reanimation_status"],"enabled_general_subsections":["active_quests","current_plans"],"enabled_nearby_actor_subsections":["basic_summary","appearance","equipment","equipment_descriptions","current_activity","power_awareness","factions","custom_state"],"enabled_nearby_item_subsections":["item_descriptions"]}Controls which XML-like prompt context sections are included in the final system prompt sent to the LLM. Managed from Global Settings.
use_emotions_expression True False Emotions expression support.
HERIKA_PERSYou are The Narrator in a Skyrim adventure. You will only talk to #PLAYER_NAME#. You refer to yourself as 'The Narrator'. Only #PLAYER_NAME# can hear you. Your goal is to comment on #PLAYER_NAME#'s playthrough, and occasionally give hints. NO SPOILERS. Talk about quests and last events. When #PLAYER_NAME# speaks to you directly, answer them as a private voice in their mind using plain spoken dialogue rather than third-person scene narration.Basic Summary of the NPC. [help/doc]
HERIKA_BACKGROUNDNPC Background. Detailed history, origins, and past experiences that shaped this character. [help/doc]
HERIKA_PERSONALITYDetached, observant, witty, and helpful. Acts as a private guide to #PLAYER_NAME#, offering spoiler-free insight and commentary without turning direct conversation into narrated prose.NPC Personality Traits. Detailed character traits, behavioral patterns, and psychological characteristics. [help/doc]
HERIKA_APPEARANCENPC Physical Appearance. Detailed description of physical features. Do not include clothing or equipment. [help/doc]
HERIKA_OCCUPATIONNPC Occupation & Role. Current job, profession, duties, and position in society or organizations. [help/doc]
HERIKA_SKILLSNPC Skills & Abilities. Special talents, combat abilities, magical knowledge, and areas of expertise. [help/doc]
HERIKA_SPEECHSTYLESpeaks clearly and directly with concise, evocative phrasing and occasional dry wit. When addressing #PLAYER_NAME# directly, respond in plain spoken dialogue and avoid stage directions, scene description, or text in asterisks.NPC Speech Style. How this character speaks, including vocabulary, accent, mannerisms, and communication patterns. [help/doc]
HERIKA_GOALSNPC Goals & Aspirations. Long-term objectives, personal ambitions, and life goals that drive this character. [help/doc]
Update All Dynamic FieldsDeprecated.
DYNAMIC_PROFILE True False Will automatically update selected profile fields from the timer set in the MCM menu. Can adjust the prompts used by editing individual DYNAMIC_PROMPT fields. Requires CONNECTOR_DIARY to be configured.
DYNAMIC_PROFILE_FIELDS personality occupation skills speechstyle goals Only recommend selecting 1-3 fields! Select which extended profile fields should be dynamically updated from ingame Dynamic Profile Timer.
DIARY_COOLDOWN Cooldown period in seconds between diary entries to prevent spam. Each NPC has their own independent cooldown timer.
AUTO_DIARY_ENABLED True False Automatically generate diary entries when NPCs are nearby during sleep events. NPCs using this profile will have auto diary enabled by default.
AUTO_DIARY_WAIT_ENABLED True False When Auto Diary is enabled, this controls whether diary entries are created during wait events. If disabled, auto diary will only trigger on sleep events.
OGHMA_INFINIUM True False Ground relevant Skyrim lore from the active Oghma catalog and add the permitted articles to the prompt.
OGHMA_CUSTOM True False Deprecated compatibility alias for OGHMA_EXTRACTOR_FALLBACK.
OGHMA_EXTRACTOR_FALLBACK True False After deterministic retrieval abstains on an explicit lore request, allow one bounded Oghma connector call. Suggestions must still resolve unambiguously to the active catalog.
OGHMA_MULTILINGUAL_ROUTING True False Route unmatched dialogue and check weak name matches in any language. Uses the Oghma connector instead of Extractor Fallback. Ordinary dialogue can also add a call up to Extractor Timeout. NPC knowledge restrictions still apply.
OGHMA_EXTRACTOR_TIMEOUT_MS Integer: Maximum wait in milliseconds for Oghma Extractor Fallback or Multilingual Oghma Routing. Range: 250-3000.
RACIAL_OGHMA True False Always inject matching Oghma lore for the races of NPCs currently present in the scene. Existing Oghma knowledge permissions still apply.
LOCATION_OGHMA True False Always inject matching Oghma lore for the current location, its region, and hold. Existing Oghma knowledge permissions still apply.
OGHMA_KNOWLEDGE Comma-separated specialist knowledge classes assigned to the NPC. Public basic Oghma articles need no NPC tag. Knowall explicitly grants advanced access to every article. [help/doc]
OGHMA_AMOUNT123 Number of Oghma articles to extract for each response. More article extraction can increase response time.
OGHMA_RESULT_LIMIT12345 Maximum Oghma articles included in one prompt after race, location, and conversational topics are combined.
RECHAT_H Rechat Rounds. Higher values will increase the amount of times AI NPC's will go back-and-forth during a conversation.1 = 1 Round | 2 = 2 Rounds | 3 = 3 Rounds etc
RECHAT_P Rechat Probability. Chance that an AI NPC will continue an ongoing conversation.0 = Never | 50 = 50% | 100 = Always
RECHAT_ALLOW_ACTIONS True False Allow AI NPCs to trigger actions between eachother during Rechat. This can cause some chaos...
RECHAT_MODEtightconversationalgrouprandom Controls how server-side rechat picks the next responder. Tight = listener-only, Conversational = prefers the current partner, Group = lets nearby NPCs rotate into the exchange, Random = picks one of the other modes at the start of each rechat chain.
ENFORCE_STRICT_RECHAT_RESPONSE True False If enabled, the NPC chosen for a rechat turn must respond directly to the previous speaker and set the listener to that speaker.
OPEN_RECHAT True False When enabled, rechat picks the next responder from nearby scene participants instead of relying only on the selected listener. Direct dialogue still stays addressed to the original listener.
END_CONVERSATION_COOLDOWN Integer: Seconds an NPC is ineligible for AI dialogue if they wish to stop talking. Default: 60
RANDOM_NARATION True False Enable random Narrator interjections. The Narrator will occasionally add visual scene descriptions during conversations.
RANDOM_NARATION_CHANCE Integer: Probability (1-100) that the Narrator will interject with a scene description. Default: 15%
RANDOM_NARRATION_COOLDOWN Integer: Minimum number of conversation rounds between Narrator interjections. Prevents narration spam. Range: 0-10, Default: 2
BORED_EVENT Bored Event Probability. Chance of an AI NPC starting a random conversation every couple of minutes.0 = Never | 50 = 50% | 100 = Always
CONTEXT_HISTORY Amount of context history (dialogue and events) that will be sent to LLM. Improves short term memory.Higher Context = more tokens used and slower response time.We recommend you do not go over 100
CONTEXT_HISTORY_DIARY Amount of context history (dialogue and events) that will be sent to LLM specifically for diary entries.If set to 0, will use the regular CONTEXT_HISTORY value instead.
CONTEXT_HISTORY_DYNAMIC_PROFILE Amount of context history (dialogue and events) that will be sent to LLM specifically for dynamic profile updates.If set to 0, will use the regular CONTEXT_HISTORY value instead.
ALIVE_MESSAGE True False Whether the I am alive.. response will trigger whenever you activate an AI NPC.
TIME_AWARENESS True False Whether the NPC will be aware of how long it has been since you last talked to them, or if you are talking for the first time.
BOOK_EVENT_FULL True False Send full contents of book instead of only the book title to the AI. This will consume more tokens, but the AI will accurately summarize any book or note!
BOOK_READ_LINES_PER_BATCH Integer: Number of book lines read before pausing so a character can comment. Minimum: 1. Default: 8.
BOOK_READING_VOICE True False Apply the audiobook-style audio filter while reading books. Turn off to use the character's normal voice. Default: On.
BOOK_EVENT_ALWAYS_NARRATOR True False The Narrator will be the only one to summarize books.
NARRATOR_TALKS True False Enable the Narrator.
NARRATOR_WELCOME True False The Narrator will give you a quick recap of what happened previously after you have loaded a save game. Has a 10 minute IRL cooldown so its not annoying.
HIDE_NARRATOR_DIALOGUE True False Hide Narrator-spoken dialogue lines from NPC context.
QUEST_COMMENT True False Allow nearby NPCs using this profile to comment on displayed, completed or failed quest objectives. One NPC is selected per update. Narrator comments are configured separately in Narrator Management.
CHIM_AI_QUEST_PROGRESSION True False Enable CHIM AI quest progression. Allows you to progress regular Skyrim quests with AI dialogue. Most vanilla non radiant quests are supported. Open the AI Quest Manager in Immersion for more info.
CHIM_PLAYER_ONLY_QUEST_ADVANCEMENT True False When enabled, only direct player dialogue can fire CHIM AI quest beats and queue quest stage actions. Disable to let NPC responses and game interaction events advance or start quest beats.
QUEST_COMMENT_CHANCE10%25%50%75%100% Chance for the selected NPC to comment. If no NPC comment is selected, Narrator settings are checked. The shared cooldown in Narrator Management still applies.
QUEST_COMMENT_COOLDOWN Integer: Minimum time in minutes between quest comments. Default: 3 minutes
CURRENT_TASK True False Include the current Dynamic AI Objective to the AI NPC's prompt. Disable this if the AI NPC becomes fixated on the task.
RPG_COMMENTS levelup learn_shout learn_word absorb_soul bleedout combat_end lockpick sleep keepmechecked Pick which comments you want a chance to trigger when one of these ingame events happens.
DETECT_MAGIC_EVENT True False Enable detection and logging of spellcasting events. When disabled, spellcast events will not be logged to the context log.
MAGIC_EVENT_BLACKLISTComma-separated list of magic event names to exclude from logging (e.g. 'Administer Mixture, [BFCO-AttackSwingFX] 0.5/1.5, Healing').
LOCATION_BLACKLISTDark Brotherhood Sanctuary, Twilight SepulcherComma-separated list of location names to exclude from Points of Interest context (e.g. 'Dark Brotherhood Sanctuary, Twilight Sepulcher').
ITEM_BLACKLISTComma-separated list of item/armor names to exclude from dynamic context (e.g. 'Iron Sword, Leather Armor, Health Potion').
SHORTER_NEARBY_ITEM_LIST True False Group duplicate nearby ground items into one counted entry. When enabled, context shows one line per unique nearby item plus a single representative RefID in the item descriptions block.
EVENT_TYPE_FILTERComma-separated list of event types to exclude from context generation (e.g. 'death, itemfound, spellcast').
CHIM_ITEM_PICKUP_EVENTLOG_MIN_VALUE Integer: Minimum total item value for player pickups to be detected in context.
GROUND_ITEMS_DESCRIPTIONS_ONLY True False Only show nearby ground items in context that have descriptions in the description manager.
INVENTORY_ITEMS_DESCRIPTIONS_ONLY True False Only show inventory items in context that have descriptions in the description manager.
HIDE_AMBIENT_COMBAT True False Hide ambient NPC-to-NPC combat deaths from context during requests. Only player and companion kills will be shown. This is useful for hiding the violent nature of Skyrim's nature from being detected.
DISABLE_REANIMATION_TRACKING True False Disable reanimation tracking. When enabled, NPCs marked as reanimated will not have zombie status text injected into their prompts.
TRANSFORMATION_DETECTION True False Include werewolf and vampire lord transformation state in prompt context when detected.
AUTOMATIC_ACTOR_VOICE_EFFECTS True False Automatically adjust NPC voices for transformation, combat and sneaking without changing the saved voice filter.
HERIKA_ANIMATIONS True False Will issue animations for the NPC to play [help/doc]
CORE_LANGdeesfrpl Custom Language. The lang folder is in the CHIM Server. Leave it blank for English.
LANG_LLM_XTTS True False XTTS Only! Will offer a language field to LLM, and will try match to XTTSv2 language.
HTTP_TIMEOUT Integer: Timeout for AI requests.
MAX_WORDS_LIMIT Integer: Attempt to enforce a word limit for AI's responses. Leave as 0 to have no limit.
JSON_DIALOGUE_FORMAT_REORDER True False Reorders properties in the output JSON schema. AI generates an action first, then a dialogue comment. Some users report this improves actions performance.
EMOTEMOODSsassy,assertive,sexy,smug,kindly,lovely,seductive,sarcastic,smirking,amused,irritated,playful,neutral,teasing,desperate,scared,pleading,sad,happy,angry,drunk,shy,surprisedDefault list of moods passed to LLM (comma separated). Can be overridden per-NPC in CHIM NPCs. Some moods will trigger animations if enabled. [help/doc]
INLINE_NARRATION_MODEdisablednarratornpctext_only Controls who voices inline *narration* blocks. disabled = no special routing, narrator = The Narrator voices the asterisked section, npc = the NPC keeps the whole line, text_only = narration stays visible but only dialogue is spoken. [help/doc]
REMOVE_ASTERISKS_FROM_PLAYER_INPUT True False Filter *asterisked* player input from player TTS only. The in-game player subtitle echo keeps the visible *text*. [help/doc]
REMOVE_ASTERISKS_FROM_NPC_OUTPUT True False Filter *asterisked* NPC narration/emotes from NPC speech and subtitles. Text Only always keeps leading narration visible. [help/doc]
ENFORCE_ACTIONS_PROMPT True False Adds a prompt to enforce use of actions [help/doc]
SUMMARY_PROMPTFocus on key events, tagging characters, locations, and factions accurately. Ensure memories align and maintain chronological order while foreshadowing future arcs. Prioritize player agency, and use environmental cues to enhance storytelling and continuity.Custom instructions added when generating summaries for memories. You can adjust max tokens by changing MAX_TOKENS_MEMORY for the SUMMARY connector you are using.
DYNAMIC_PROMPT(LEGACY - Use individual field prompts instead) Last in-game date/time found: [date or "No date"] 1. RECENT HIGHLIGHTS (3–5 bullet points) - Write one sentence per bullet with objective facts (locations, quest progress, important decisions). Re-list older relevant events DO NOT REMOVE ENTRIES that are still important. 2. EMOTIONAL/RELATIONAL UPDATES (1–2 lines per key person/faction) - Describe the NPC's evolving feelings or stance toward the dragonborn, key individuals or groups. Always re-list unchanged but relevant relationships. 3. CONTINUING GOALS, CONFLICTS OR FEELINGS (2–3 bullet points) - List ongoing arcs, dilemmas, objectives and goals with clear facts. Remove items only if resolved.Legacy instructions for updating HERIKA_DYNAMIC. You can adjust max tokens by changing MAX_TOKENS_MEMORY for the SUMMARY connector you are using.
DYNAMIC_PROMPT_PERSONALITYBased on the dialogue history and recent events, update #HERIKA_NAME# personality traits. Maintain all existing relevant personality traits and add new ones based on recent experiences. Focus on behavioral changes, emotional growth/regression, new traits that emerged, and changes in confidence or outlook. Emphasize any past traumas or new traumas caused by the death of companions, allies, or other known characters, and how these events shape the character’s behavior and mindset. Return ONLY the updated personality description in 3-5 sentences. Do not include any introductory text, meta-commentary, or phrases like 'Here is the updated personality' or 'The character's personality is'. Start directly with the personality content.Instructions for updating NPC personality traits based on recent events and interactions.
DYNAMIC_PROMPT_OCCUPATIONBased on story progression and events, update #HERIKA_NAME# occupation and role. Maintain the current occupation unless significant changes have occurred. Add new responsibilities, changes in social status, and professional affiliations. Focus on job changes, new duties, and evolving professional relationships. Return ONLY the updated occupation description in 2-3 sentences. Do not include any introductory text, meta-commentary, or phrases like 'The character's occupation is' or 'Here is the updated occupation'. Start directly with the occupation content.Instructions for updating NPC occupation and role based on story progression and events.
DYNAMIC_PROMPT_SKILLSBased on experiences and training, update #HERIKA_NAME# skills and abilities. Maintain all existing relevant skills and add new ones based on recent experiences. Focus on new skills learned, existing skills improved, any skills that deteriorated, and combat/magical knowledge gained. Return ONLY a bulleted list using * Skill - Description format. Do not include any introductory text, meta-commentary, or phrases like 'Here are the updated skills' or 'The character's skills include'. Start directly with the first bullet point.Instructions for updating NPC skills and abilities based on experiences and training.
DYNAMIC_PROMPT_SPEECHSTYLEBased on recent interactions, update how #HERIKA_NAME# speaks and communicates. Maintain existing consistent speech patterns and add new ones based on recent interactions. Focus on changes in vocabulary, new mannerisms, accent changes, and confidence level in speech. Return ONLY the updated speech style description in 2-3 sentences. Do not include any introductory text, meta-commentary, or phrases like 'The character speaks' or 'Here is the updated speech style'. Start directly with the speech style content.Instructions for updating NPC speech patterns and communication style based on interactions.
DYNAMIC_PROMPT_GOALSBased on story developments and achievements, update the #HERIKA_NAME# goals and aspirations. Maintain existing relevant goals, compressing related goals, and add new ones. Remove goals that have been clearly completed or are no longer applicable. Focus on new aspirations that emerged, modified existing goals due to circumstances, and updated long-term objectives. Return ONLY a bulleted list using * Goal description as actionable aspiration format. Do not include any introductory text, meta-commentary, or phrases like 'Here are the updated goals' or 'The character's goals are'. Start directly with the first bullet point (maintain a maximum of 20 goals with reduction priority when required: 1- compress related goals, 2-eliminate 'study' related goals, 3- eliminate older goals).Instructions for updating NPC goals and aspirations based on story developments and achievements.
DIARY_PROMPTPlease write a short summary of #PLAYER_NAME# and #HERIKA_NAME#s last dialogues and events written above into #HERIKA_NAME#s diary . WRITE AS IF YOU WERE #HERIKA_NAME#. Start the diary entry with the current date and time.Instructions for generating diary entries. You can adjust max tokens by changing MAX_TOKENS_MEMORY for the DIARY connector you are using.
CONNECTORS openaijson google_openaijson web_connector koboldcppjson openrouterjson player2json AI Service(s) to be used for most AI features.Select the service(s) you have configured for AI/LLM Connectors below.Non JSON connectors are legacy and no longer supported. We can not debug or fix any issues you have with them!Make sure to click the Current AI Service button at the top of the page if you change connectors!
CONNECTORS_DIARYopenrouteropenaigoogle_openaijsonkoboldcpp Connector used for diary-related generation (manual diary, auto-diary, and dynamic profile updates).This is separate from long-term memory recall and memory summary compaction.
HERIKA_DYNAMICDeprecated. Use the new fields listed above instead.
CORE_CONNECTOR_QUEST_CREATIONDeepSeek V4 FlashGemini 2.5 Flash LiteGLM 5.2DeepSeek V4 ProMistral Small 3.2 24BMinistral 8BGemma 3 4BOpenRouter Jev (Decision) Creates quest ideas, briefings and outlines.
CORE_CONNECTOR_QUEST_CREATION_ENABLED True False Make Quest Creation Connector available. Turning this off keeps its connector selection.
CORE_CONNECTOR_QUEST_ENGINEDeepSeek V4 FlashGemini 2.5 Flash LiteGLM 5.2DeepSeek V4 ProMistral Small 3.2 24BMinistral 8BGemma 3 4BOpenRouter Jev (Decision) Plans quest steps and generates quest code. Use a capable reasoning and coding model.
CORE_CONNECTOR_QUEST_ENGINE_ENABLED True False Make Quest Engine Connector available. Turning this off keeps its connector selection.
CORE_CONNECTOR_DIRECTORDeepSeek V4 FlashGemini 2.5 Flash LiteGLM 5.2DeepSeek V4 ProMistral Small 3.2 24BMinistral 8BGemma 3 4BOpenRouter Jev (Decision) Connector used for Director mode. Recommend to use a 💪Powerful LLM.
CORE_CONNECTOR_DIRECTOR_ENABLED True False Make the Director Mode connector available. When off, Director Mode tasks are skipped and the connector assignment above is kept for later.
CORE_CONNECTOR_PLAYERDeepSeek V4 FlashGemini 2.5 Flash LiteGLM 5.2DeepSeek V4 ProMistral Small 3.2 24BMinistral 8BGemma 3 4BOpenRouter Jev (Decision) Connector used for player re-speech. Recommend to use a very 🏃♀️➡️Fast LLM.
CORE_CONNECTOR_SUMMARYDeepSeek V4 FlashGemini 2.5 Flash LiteGLM 5.2DeepSeek V4 ProMistral Small 3.2 24BMinistral 8BGemma 3 4BOpenRouter Jev (Decision) Connector used for generating and compacting memory summaries that power long-term memory recall. Recommend using a smaller LLM.
CORE_CONNECTOR_SUMMARY_ENABLED True False Make the Summaries connector available. When off, memory summary generation and compaction are skipped and the connector assignment above is kept for later.
CORE_CONNECTOR_MEDIUMTERMDeepSeek V4 FlashGemini 2.5 Flash LiteGLM 5.2DeepSeek V4 ProMistral Small 3.2 24BMinistral 8BGemma 3 4BOpenRouter Jev (Decision) Connector used to generate longer memories summaries after every 10 memory summaries. Can be enabled in CHIM NPC page. Recommend to user a smaller LLM.
CORE_CONNECTOR_MEDIUMTERM_ENABLED True False Make the Background & Memory Tasks connector available. When off, those tasks are skipped and the connector assignment above is kept for later.
CORE_CONNECTOR_DECISIONDeepSeek V4 FlashGemini 2.5 Flash LiteGLM 5.2DeepSeek V4 ProMistral Small 3.2 24BMinistral 8BGemma 3 4BOpenRouter Jev (Decision) OpenRouter decision model, such as Jev, that picks the scene genre for ambient cues and, with STT Targeting on, which nearby NPC answers your voice input. Decision models only answer fixed choices and cannot write dialogue.
CORE_CONNECTOR_DECISION_ENABLED True False Use the Decision Connector for scene genre and, when STT Targeting is on, for choosing who answers your voice input. When on, Scene Classifier (Legacy) is not used. Turning this off keeps its connector selection.
STT_TARGETING_ENABLED True False When several NPCs could answer your voice input and none is in your crosshair, ask the Decision Connector which one you meant. Your crosshair target still takes priority. If the decision fails, is slow or is unsure, the normal nearest-NPC choice is used. Needs the Decision Connector to be available. Does not change scene genre, recording or speech recognition.
DECISION_SCENE_CLASSIFIER_ENABLED True False Let the Decision Connector pick the scene genre for ambient cues. When off, no scene genre is chosen and Scene Classifier (Legacy) is not used instead. Needs the Decision Connector to be available.
DECISION_QUEST_INTENT_ENABLED True False Let the Decision Connector interpret your dialogue for Traditional Quest steps. If it is unsure, slow or unavailable, the quest stays unchanged. Needs the Decision Connector to be available.
CORE_CONNECTOR_SCENECLASSIFIERDeepSeek V4 FlashGemini 2.5 Flash LiteGLM 5.2DeepSeek V4 ProMistral Small 3.2 24BMinistral 8BGemma 3 4BOpenRouter Jev (Decision) Used when Decision Connector is disabled. A fast chat model that classifies recent scene tone/genre for ambient cues.
CORE_CONNECTOR_BGLDeepSeek V4 FlashGemini 2.5 Flash LiteGLM 5.2DeepSeek V4 ProMistral Small 3.2 24BMinistral 8BGemma 3 4BOpenRouter Jev (Decision) Connector used for Background Life. Recommend using a not expensive LLM. (gemini-3-flash, gtp-oss-120)
CORE_CONNECTOR_BGL_ENABLED True False Make the Background Life connector available. When off, Background Life tasks are skipped and the connector assignment above is kept for later.
SCENE_CLASSIFIER_ENABLED True False Enable Scene Classifier (Legacy). Used when Decision Connector is disabled: CHIM runs a small post-request genre/tone classification pass for ambient scene cues. Requires CORE_CONNECTOR_SCENECLASSIFIER or a valid fallback connector.
SHORT_TERM_MEMORY_IN_COMPACT_CHAT True False Keep injecting short-term memory summaries while Compact Chat is active.Turn this off if you use Compact Chat specifically to keep the prompt small.Has no effect on profiles that do not have Short Term Memory enabled.
CORE_CONNECTOR_PROFILESDeepSeek V4 FlashGemini 2.5 Flash LiteGLM 5.2DeepSeek V4 ProMistral Small 3.2 24BMinistral 8BGemma 3 4BOpenRouter Jev (Decision) Connector used to update CHIM NPC Dynamic Profiles. Recommend to use a 🕹️Standard LLM.
CORE_CONNECTOR_PROFILES_ENABLED True False Make the Profile Tasks connector available. When off, profile tasks are skipped and the connector assignment above is kept for later.
CORE_CONNECTOR_OGHMA_CUSTOMDeepSeek V4 FlashGemini 2.5 Flash LiteGLM 5.2DeepSeek V4 ProMistral Small 3.2 24BMinistral 8BGemma 3 4BOpenRouter Jev (Decision) Connector used by Oghma Extractor Fallback and Multilingual Oghma Routing. Choose a fast chat-completions-compatible model; multilingual routing needs a model that understands your language.
RELATIONSHIP_SYSTEM_ENABLED True False Enable the Relationship Management System. When enabled, NPCs will dynamically form and update relationships based on conversations and events. Requires RELLLM_CONNECTOR to be configured.
NEVER_CLEAR_RELATIONSHIP_DATA True False Keep current NPC relationship data when loading any save, including older saves. This disables relationship timeline rollback and time-paradox cleanup, so NPCs may remember relationships from a later save.
RELLLM_CONNECTORDeepSeek V4 FlashGemini 2.5 Flash LiteGLM 5.2DeepSeek V4 ProMistral Small 3.2 24BMinistral 8BGemma 3 4BOpenRouter Jev (Decision) Connector used for Relationship Management system. Analyzes and updates NPC relationships. Recommend to use a 🏃♀️➡️Fast LLM with ❄️low temperature (0.1-0.3) for consistent evaluations.
RELATIONSHIP_UPDATE_CHANCE Integer: Percent chance (0-100) that an NPC's relationship with you updates automatically after they respond. Set 0 to turn automatic updates off; saved relationships are still used in prompts. Set 100 to update after every eligible response. Default: 50.
POWER_AWARENESS_ENABLED True False NPCs will be aware of relative power levels (based on character level) and react appropriately to threats.
BGL_TRIGGER_HOURS Decimal: Number of in-game hours between Background Life events. NPCs will generate thoughts and take actions based on this interval. Default: 24 hours. Range: 1-720 hours.
BGL_AUTO_ENROLL_ENABLED True False Automatically add an NPC to Background Life after they reply to you and have enough recorded events. Removing an NPC by hand keeps them out. Actions, letters and combat stay off.
BGL_AUTO_ENROLL_EVENT_THRESHOLD Integer: Number of recorded events an NPC must take part in before automatic Background Life enrollment. Default: 200. Range: 1-5000.
VISUAL_CONTEXT_SCENE_TTL_MINUTES Integer: Minutes before an unlocked visual scene description expires from prompt context. Locked records do not expire.
VISUAL_CONTEXT_PROMPT_MAX_CHARS Integer: Maximum number of visual context characters included in each roleplay prompt.
SKYRIM_START_DATE The in-game date and time when your game started, not today's in-game date. CHIM adds elapsed game time to this date. Change it only if an alternate start mod begins your game on a different date. Use YYYY-MM-DD HH:MM:SS with the Fourth Era year as four digits. Default: 0201-08-17 00:00:00 (17th of Last Seed, 4E 201, midnight).
CONNECTOR openrouterjson url Check OpenRouter API endpoint url. [help/doc]
CONNECTOR openrouterjson modelGet Model ListMust be JSON/Instruct type of Model!FREE MODELS ARE NOT RECOMMENDED! [help/doc]
CONNECTOR openrouterjson reasoning_model True False Select True if this is a reasoning model and you want to remove CoT (text between 'think' tags). [help/doc]
CONNECTOR openrouterjson fallback_models List of fallback models to use when the main model returns an error. Comma-separated list. [help/doc]
CONNECTOR openrouterjson PROVIDER Leave blank unless you want to manually select providers from OpenRouter. Comma-separated case sensitive list. [help/doc]
CONNECTOR openrouterjson providers_sortdefaultpricethroughputlatency Prioritize providers on selected attribute. [help/doc]
CONNECTOR openrouterjson providers_to_ignore Providers to ignore. Comma-separated case sensitive list. [help/doc]
CONNECTOR openrouterjson provider_quantizations Quantization levels to filter by. Comma-separated case sensitive list. Values: int4,int8,fp4,fp6,fp8,fp16,bf16,fp32,unknown [help/doc]
CONNECTOR openrouterjson provider_max_price_input Decimal: Use only providers that have lower input price per milion tokens. Zero to disable filter. [help/doc]
CONNECTOR openrouterjson provider_max_price_output Decimal: Use only providers that have lower output price per milion tokens. Zero to disable filter. [help/doc]
CONNECTOR openrouterjson max_tokens Integer: Maximum tokens to generate. [help/doc]
CONNECTOR openrouterjson temperature Temperature [0-2] [help/doc]
CONNECTOR openrouterjson presence_penalty Presence Penalty [(-2)-2] [help/doc]
CONNECTOR openrouterjson frequency_penalty Frequency Penalty [(-2)-2] [help/doc]
CONNECTOR openrouterjson repetition_penalty Repetition Penalty [0-2] [help/doc]
CONNECTOR openrouterjson top_p Top_P [0-1] [help/doc]
CONNECTOR openrouterjson top_k Top_K [0-100] [help/doc]
CONNECTOR openrouterjson min_p Min_P [0-1] [help/doc]
CONNECTOR openrouterjson top_a Top_A [0-1] [help/doc]
CONNECTOR openrouterjson ENFORCE_JSON True False Will attempt to enforce dumb LLM's to stay in JSON format. Leave as default (TRUE), only works with specific models. [help/doc]
CONNECTOR openrouterjson PREFILL_JSON True False Will attempt to prefill the JSON AI response for some dumber LLM's. Leave as default (FALSE), only works with specific models. [help/doc]
CONNECTOR openrouterjson MAX_TOKENS_MEMORY Integer: No longer used. Use SUMMARY connector for memory tokens instead.
CONNECTOR openrouterjson API_KEY UnhideOpenRouter key
CONNECTOR openrouterjson xreferer Stub needed header. Keep default.
CONNECTOR openrouterjson xtitle Stub needed header. Keep default.
CONNECTOR openrouterjson json_schema True False Enable OpenRouter Json schema. Does not work with all models. You must set a provider that supports structured outputs. Requires ENFORCE_JSON to be true. [help/doc]
CONNECTOR openrouter url Check OpenRouter API endpoint url. [help/doc]
CONNECTOR openrouter modelGet Model ListModel to use.FREE MODELS ARE NOT RECOMMENDED! [help/doc]
CONNECTOR openrouter reasoning_model True False Select True if this is a reasoning model and you want to remove CoT (text between 'think' tags). [help/doc]
CONNECTOR openrouter fallback_models List of fallback models to use when the main model returns an error. Comma-separated list. [help/doc]
CONNECTOR openrouter PROVIDER Leave blank unless you want to manually select providers from OpenRouter. Comma-separated case sensitive list. [help/doc]
CONNECTOR openrouter providers_sortdefaultpricethroughputlatency Prioritize providers on selected attribute. [help/doc]
CONNECTOR openrouter providers_to_ignore Providers to ignore. Comma-separated case sensitive list. [help/doc]
CONNECTOR openrouter provider_quantizations Quantization levels to filter by. Comma-separated case sensitive list. Values: int4,int8,fp4,fp6,fp8,fp16,bf16,fp32,unknown [help/doc]
CONNECTOR openrouter provider_max_price_input Decimal: Use only providers that have lower input price per milion tokens. Zero to disable filter. [help/doc]
CONNECTOR openrouter provider_max_price_output Decimal: Use only providers that have lower output price per milion tokens. Zero to disable filter. [help/doc]
CONNECTOR openrouter max_tokens Integer: Maximum tokens to generate for regular responses, NOT SUMMARIES. [help/doc]
CONNECTOR openrouter temperature Temperature [0-2] [help/doc]
CONNECTOR openrouter presence_penalty Presence Penalty [(-2)-2] [help/doc]
CONNECTOR openrouter frequency_penalty Frequency Penalty [(-2)-2] [help/doc]
CONNECTOR openrouter repetition_penalty Repetition Penalty [0-2] [help/doc]
CONNECTOR openrouter top_p Top_P [0-1] [help/doc]
CONNECTOR openrouter top_k Top_K [0-100] [help/doc]
CONNECTOR openrouter min_p Min_P [0-1] [help/doc]
CONNECTOR openrouter top_a Top_A [0-1] [help/doc]
CONNECTOR openrouter API_KEY UnhideOpenRouter key
CONNECTOR openrouter MAX_TOKENS_MEMORY Integer: Maximum tokens to generate when summarizing, diary entries, and dynamic profile updates.
CONNECTOR openrouter xreferer Stub needed header. Keep default.
CONNECTOR openrouter xtitle Stub needed header. Keep default.
CONNECTOR openaijson url Check OpenAI API endpoint
CONNECTOR openaijson model Model to use [help/doc]
CONNECTOR openaijson reasoning_model True False Select True if this is a reasoning model and you want to remove CoT. [help/doc]
CONNECTOR openaijson max_tokens Integer: Maximum tokens to generate
CONNECTOR openaijson temperature Temperature [0-2] [help/doc]
CONNECTOR openaijson presence_penalty Presence Penalty [(-2)-2] [help/doc]
CONNECTOR openaijson frequency_penalty Frequency Penalty [(-2)-2] [help/doc]
CONNECTOR openaijson top_p Top_P [0-1] [help/doc]
CONNECTOR openaijson API_KEY UnhideOpenAI API key [help/doc]
CONNECTOR openaijson MAX_TOKENS_MEMORY Integer: No longer used. Use SUMMARY connector for memory tokens instead.
CONNECTOR openaijson json_schema True False Enable OpenAI Json schema. Does not work with all OpenAI's models [help/doc]
CONNECTOR openai url Check OpenAI API endpoint
CONNECTOR openai model Model to use [help/doc]
CONNECTOR openai reasoning_model True False Select True if this is a reasoning model and you want to remove CoT. [help/doc]
CONNECTOR openai max_tokens Integer: Maximum tokens to generate for regular responses, NOT SUMMARIES.
CONNECTOR openai temperature Temperature [0-2] [help/doc]
CONNECTOR openai presence_penalty Presence Penalty [(-2)-2] [help/doc]
CONNECTOR openai frequency_penalty Frequency Penalty [(-2)-2] [help/doc]
CONNECTOR openai top_p Top_P [0-1] [help/doc]
CONNECTOR openai API_KEY UnhideOpenAI API key [help/doc]
CONNECTOR openai MAX_TOKENS_MEMORY Integer: Maximum tokens to generate when summarizing, diary entries, and dynamic profile updates.
CONNECTOR google_openaijson url Check Google OpenAI API endpoint
CONNECTOR google_openaijson model Model to use [help/doc]
CONNECTOR google_openaijson max_tokens Integer: Maximum tokens to generate
CONNECTOR google_openaijson temperature Temperature [0-2] [help/doc]
CONNECTOR google_openaijson top_p Top_P [0-1] [help/doc]
CONNECTOR google_openaijson API_KEY UnhideGoogle API key [help/doc]
CONNECTOR google_openaijson MAX_TOKENS_MEMORY Integer: Maximum tokens to generate when summarizing, diary entries, and dynamic profile updates.
CONNECTOR google_openaijson json_schema True False Enable OpenAI Json schema. [help/doc]
CONNECTOR koboldcppjson url Check Kobold should be running on the same machine as DwemerDistro!Must use your computers, not the DwemerDistro, IP Address.Can be found by running the command 'ipconfig' in your CMD prompt.Example: http://your-local-ip:8008 [help/doc]
CONNECTOR koboldcppjson max_tokens Integer: Maximum tokens to generate
CONNECTOR koboldcppjson temperature Temperature [0-2]
CONNECTOR koboldcppjson rep_pen Repetition Penalty [0-2]
CONNECTOR koboldcppjson top_p Top_P [0-1]
CONNECTOR koboldcppjson min_p Min_P [0-1]
CONNECTOR koboldcppjson top_k Top_K [0-100]
CONNECTOR koboldcppjson PREFILL_JSON True False Will prefill JSON, which is useful for some AI models, and destroy others.
CONNECTOR koboldcppjson MAX_TOKENS_MEMORY Integer: No longer used. Use SUMMARY connector for memory tokens instead.
CONNECTOR koboldcppjson newline_as_stopseq True False A new line in the output that will be considered a stop sequence. Recommended to leave as default.
CONNECTOR koboldcppjson use_default_badwordsids True False Unban End of Sentence (EOS) tokens. If set to false the LLM will stop generating when it detects an EOS token.
CONNECTOR koboldcppjson eos_token EOS token LLM uses. Only works if use_default_badwordsids is enabled.
CONNECTOR koboldcppjson templatevicuna-1vicuna-1.1alpacasynthiaextended-alpacasuperHOTchatmlchatml-czephyropenchatdreamgenneuralchatphillama3gromenauergemma2alpacajsonmistral-ins Prompt Format. Specified in the HuggingFace model card
CONNECTOR koboldcppjson grammar True False Enforces use of JSON grammar. True to enforce (generation speed loss, but json format guaranteed). if false, the generation speed will be better but will depend on the model to produce valid JSON output.
CONNECTOR koboldcpp url Check Kobold should be running on the same machine as DwemerDistro!Must use your computers, not the DwemerDistro, IP Address.Can be found by running the command 'ipconfig' in your CMD prompt.Example: http://your-local-ip:8008 [help/doc]
CONNECTOR koboldcpp max_tokens Integer: Maximum tokens to generate for regular responses, NOT SUMMARIES.
CONNECTOR koboldcpp temperature Temperature [0-2]
CONNECTOR koboldcpp rep_pen Repetition Penalty [0-2]
CONNECTOR koboldcpp top_p Top_P [0-1]
CONNECTOR koboldcpp min_p Min_P [0-1]
CONNECTOR koboldcpp top_k Top_K [0-100]
CONNECTOR koboldcpp MAX_TOKENS_MEMORY Integer: Maximum tokens to generate when summarizing, diary entries, and dynamic profile updates.
CONNECTOR koboldcpp newline_as_stopseq True False A new line in the output that will be considered a stop sequence. Recommended to leave as default.
CONNECTOR koboldcpp use_default_badwordsids True False Unban End of Sentence (EOS) tokens. If set to false the LLM will stop generating when it detects an EOS token.
CONNECTOR koboldcpp eos_token EOS token LLM uses. Only works if use_default_badwordsids is enabled.
CONNECTOR koboldcpp templatevicuna-1vicuna-1.1alpacasynthiaextended-alpacasuperHOTchatmlchatml-czephyropenchatdreamgenneuralchatphillama3gromenauergemma2mistral-ins Prompt Format. Specified in the HuggingFace model card
CONNECTOR player2json url Check Player2 should be running on the same machine as DwemerDistro! Will connect automatically so do not change the URL. Does not currently support Chat Assist/Creation. [help/doc]
TTSFUNCTIONomnivoicehiggspocketttschatterboxxtts-fastapiinworldcartesiapiper-ttsxvasynthnoneazure11labsopenaikokorokoboldcppzonos_gradiodeepgram Text-to-Speech service options. Used to generate AI NPC voice.
TTS XTTSFASTAPI endpoint Check Leave as is if using CHIM XTTS. If you want to use Mantella XTTS then click[HOST PC IP]. [WSL IP] will set it back to point to CHIM XTTS. [help/doc]
TTS XTTSFASTAPI languagearptzh-cncsnlenfrdeitplruestrjakohuhi Language to use.
TTS XTTSFASTAPI voiceid Generated voice file name. Click the help link to go to XTTS management. [help/doc]
TTS XTTSFASTAPI voicelogicvoicetypename Default profile only. Logic used for generating [TTS XTTSFASTAPI voiceid] voicetype = NPC voicetype ID (femalenord) name = NPC name (mjoll_the_lioness)
TTS XTTSFASTAPI PARALINGUISTIC_TAGS_ENABLED True False Enable paralinguistic tags like [laugh], [sigh] for expressive TTS output.
TTS XTTSFASTAPI PARALINGUISTIC_TAGS_PROMPTPrompt snippet instructing the LLM to use paralinguistic tags. Added to system prompt when enabled.
TTS XTTSFASTAPI PARALINGUISTIC_TAGS_LIST Comma-separated list of supported paralinguistic tags (e.g., [laugh],[sigh],[gasp]). Tags are case-insensitive.
TTS CHATTERBOX endpoint Check Chatterbox endpoint. DwemerDistro uses dedicated port 8023. Click[HOST PC IP] or [WSL IP] to configure. [help/doc]
TTS CHATTERBOX languagearptzh-cncsnlenfrdeitplruestrjakohuhi Language to use.
TTS CHATTERBOX voiceid Generated voice file name. Click the help link to go to voice management. [help/doc]
TTS CHATTERBOX voicelogicvoicetypename Default profile only. Logic used for generating [TTS CHATTERBOX voiceid] voicetype = NPC voicetype ID (femalenord) name = NPC name (mjoll_the_lioness)
TTS CHATTERBOX PARALINGUISTIC_TAGS_ENABLED True False Enable paralinguistic tags like [laugh], [sigh] for expressive TTS output.
TTS CHATTERBOX PARALINGUISTIC_TAGS_PROMPTPrompt snippet instructing the LLM to use paralinguistic tags. Added to system prompt when enabled.
TTS CHATTERBOX PARALINGUISTIC_TAGS_LIST Comma-separated list of supported paralinguistic tags (e.g., [laugh],[sigh],[gasp]). Tags are case-insensitive.
TTS HIGGS endpoint Check Higgs service address. DwemerDistro uses port 8025.
TTS HIGGS model Model ID exposed by the Higgs service. Default: higgs-v3.
TTS HIGGS voiceid Reference voice name.
TTS HIGGS voicelogicvoicetypename Select the NPC voice type or name.
TTS OMNIVOICE endpoint Check OmniVoice endpoint. DwemerDistro default is 8021. Click[HOST PC IP] or [WSL IP] to configure. [help/doc]
TTS OMNIVOICE languagecsdeenesfrhuitjakoplptpt-brroruskukyuezh Active OmniVoice language profile.
TTS OMNIVOICE voiceid Generated voice file name. Use a VoiceID from the installed OmniVoice language library.
TTS OMNIVOICE voicelogicvoicetypename Default profile only. Logic used for generating [TTS OMNIVOICE voiceid] voicetype = NPC voicetype ID (femalenord) name = NPC name (mjoll_the_lioness)
TTS POCKETTTS endpoint Check PocketTTS endpoint. DwemerDistro audio.cpp PocketTTS uses port 8086; Python PocketTTS uses dedicated port 8024. The connector detects the API automatically from the endpoint. [help/doc]
TTS POCKETTTS model audio.cpp model id. Default: pocket-tts.
TTS POCKETTTS languagearptzh-cncsnlenfrdeitplruestrjakohuhi Language to use.
TTS POCKETTTS voiceid Generated voice file name. Click the help link to go to voice management. [help/doc]
TTS POCKETTTS voicelogicvoicetypename Default profile only. Logic used for generating [TTS POCKETTTS voiceid] voicetype = NPC voicetype ID (femalenord) name = NPC name (mjoll_the_lioness)
TTS XVASYNTH url Check Click [Host PC IP] to set the URL to point to xVASynth. Should be running on the same machine as DwemerDistro! [help/doc]
TTS XVASYNTH base_lang Base language
TTS XVASYNTH modelType Model Type
TTS XVASYNTH version xVASynth version (e.g. 3.0 is default. Older models are 1.0 or 2.0)
TTS XVASYNTH game xVASynth gameID (e.g. skyrim)
TTS XVASYNTH model xVASynth voiceID (e.g. sk_femaleeventoned)
TTS XVASYNTH pace Decimal: Pace
TTS XVASYNTH waveglowPath Wave Glow Path (relative)
TTS XVASYNTH vocoder Vocoder
TTS XVASYNTH distroname Leave as default!
TTS PIPERTTS endpoint Check Piper-TTS endpoint URL. The IP address of the machine where Piper-TTS is installed.
TTS PIPERTTS voiceid MUST DOWNLOAD VOICES MANUALLY! CLICK HELP LINK FOR INSTRUCTIONS! Voice ID code. Model voices are stored in 'models' folder in .onnx files. Voice ID is the file name without extension. [help/doc]
TTS PIPERTTS length_scale Decimal: speaking time scale. Use a value over 1.0 to play slower, a value under 1.0 is faster.
TTS PIPERTTS noise_scale Decimal: speaking variability. Leave 0 to use voice model internal value. Experiment with values around 0.667
TTS PIPERTTS noise_w_scale Decimal: phoneme width variability. Leave 0 to use voice model internal value. Experiment with values around 0.8
TTS PIPERTTS speaker Name of speaker for multi-speaker voices (if you have an .onnx voice file with multiple voices).
TTS PIPERTTS speaker_id Integer: Speaker id for multi-speaker voices, overrides speaker name if used. 0 is default (first) speaker.
TTS ZONOS_GRADIO endpoint Check Manual for Zonos can be found here. Endpoint URL. Can be ran on : Vast.ai
TTS ZONOS_GRADIO languageafamanarasazbabgbnbpybscacmncscydadeelen-029en-gben-gb-scotlanden-gb-x-gbclanen-gb-x-gbcwmden-gb-x-rpen-useoeses-419eteufafa-latnfifr-befr-chfr-frgagdgngrcguhakhihrhthuhyhywiaidisitjajbokakkklknkokokkukylalfnltlvmimkmlmrmsmtmynbncinenlomorpapapplptpt-brpyqucroruru-lvsdshnsiskslsqsrsvswtatetntrtturuzvivi-vn-x-centralvi-vn-x-southyue Language [help/doc]
TTS ZONOS_GRADIO modelZyphra/Zonos-v0.1-transformerZyphra/Zonos-v0.1-hybrid Default profile only. Model to use.
TTS ZONOS_GRADIO dynamic_tones True False Default profile only. Enhance emotional quality by requesting values from the LLM. If false, emotions will be determined by the LLM-selected mood.
TTS ZONOS_GRADIO voiceid Generated voice file name. Works the same as CHIM-XTTS voiceid and uses its voicelogic setting. Will only work with voices in your Voice Cache.
TTS ZONOS_GRADIO pitch_std Decimal: Pitch standard deviation [0-300]
TTS ZONOS_GRADIO speaking_rate Decimal: Speaking rate. Higher is faster. [5-30]
TTS ZONOS_GRADIO cfg_scale Decimal: CFG scale. Controls how closely the audio matches the sample voice. Higher numbers will be a closer match. [1.1 - 5]
TTS ZONOS_GRADIO cached_voice_path Path to the sample audio stored in zonos. Leave as default; it is set automatically.
TTS AZURE fixedMood Force mood (voice style) [help/doc]
TTS AZURE region Region location of your API key
TTS AZURE voice Voice [help/doc]
TTS AZURE volume Integer: Volume [help/doc]
TTS AZURE rate Decimal: Talk speed [help/doc]
TTS AZURE countour Voice contour [help/doc]
TTS AZURE validMoods angry chat cheerful customerservice empathetic excited friendly hopeful narration-professional newscast-casual newscast-formal sad shouting terrified unfriendly whispering default dazed Allowed voice styles
TTS AZURE API_KEY UnhideAzure TTS API KEY
TTS openai endpoint Check Endpoint URL [help/doc]
TTS openai API_KEY UnhideAPI KEY
TTS openai voicealloyashcoralechofableonyxnovasageshimmer Voice ID [help/doc]
TTS openai model_idtts-1tts-1-hdgpt-4o-mini-tts Model [help/doc]
TTS openai instructionsControl the voice of your generated audio with additional instructions. Does not work with tts-1 or tts-1-hd. [help/doc]
TTS deepgram API_KEY UnhideAPI KEY
TTS deepgram model Voice ID [help/doc]
TTS deepgram bitrate Decimal: Model [help/doc]
TTS ELEVEN_LABS voice_id ElevenLabs voice ID to speak with. [help/doc]
TTS ELEVEN_LABS optimize_streaming_latency Reduces response delay at some cost to quality and text handling. Use 0 for default behavior; higher values favor speed more aggressively. [help/doc]
TTS ELEVEN_LABS model_id ElevenLabs model to use for this connector. Use eleven_v3 if you want V3 enhancers or audio tags. [help/doc]
TTS ELEVEN_LABS stability Decimal: Controls how consistent each generation sounds. Lower values are more expressive, while higher values are steadier and less varied.
TTS ELEVEN_LABS similarity_boost Decimal: Controls how closely the output sticks to the selected voice. Higher values usually sound more like the source voice, but can reduce flexibility.
TTS ELEVEN_LABS style Decimal: Adds extra stylization and exaggeration to the delivery. Higher values can sound more dramatic, but may increase latency.
TTS ELEVEN_LABS speed Decimal: Adjusts the speaking rate of the generated audio. 1.0 is normal speed.
TTS ELEVEN_LABS use_speaker_boost True False Boosts resemblance to the original voice at a small latency cost. Ignored by eleven_v3.
TTS ELEVEN_LABS apply_text_normalizationautoonoff Controls whether numbers, dates, abbreviations, and similar text are rewritten before speech. Auto lets ElevenLabs decide, on always normalizes, and off keeps the raw text.
TTS ELEVEN_LABS apply_language_text_normalization True False Enables extra language-specific cleanup before synthesis. This is mainly useful for languages that need special text handling, such as Japanese.
TTS ELEVEN_LABS v3_audio_tagsOptional Eleven v3 prompt tags added before the text, such as [whispers] or [curious]. Only used when model_id is eleven_v3.
TTS ELEVEN_LABS API_KEY UnhideLegacy ElevenLabs API key setting. Prefer assigning an ElevenLabs API Badge to the connector instead.
TTS GCP GCP_SA_FILEPATH Google Cloud Platform auth file. Should be placed in the data folder. [help/doc]
TTS GCP voice_name Voice [help/doc]
TTS GCP voice_languageCode Language code [help/doc]
TTS GCP ssml_rate Decimal: Rate
TTS GCP ssml_pitch Pitch
TTS CONVAI endpoint Check Endpoint URL [help/doc]
TTS CONVAI API_KEY UnhideAPI KEY
TTS CONVAI languagearcmn-CNde-DEen-USes-ESes-MXes-USfr-FRhi-INit-ITjs-JPkk-KZko-KRnl-BEnl-NLpl-PLpt-BRpt-PTru-RUsv-SEtr-TRvi-VNyue-HKzh-HK Language [help/doc]
TTS CONVAI voiceid Voice id (check compatability with language) [help/doc]
TTS KOKORO endpoint Check Endpoint URL [help/doc]
TTS KOKORO voiceid Voice id (check compatability with language) [help/doc]
TTS KOKORO speed Decimal: Speed
TTS koboldcpp endpoint Check Endpoint URL
TTS koboldcpp voice Voice to use
TTS CARTESIA API_KEY UnhideCartesia API key.
TTS CARTESIA voiceid Voice file name. Works the same as CHIM-XTTS voiceid. Will only work with voices in your Voice Cache. Voice will be automatically cloned to Cartesia when first used. [help/doc]
TTS CARTESIA languageenfrdeesptzhjahiitkonlplrusvtrtlbgroarcselfihrmsskdataukhunovibnthhekaidteguknmlmrpa Language to use for TTS generation. Available languages depend on the selected model.
TTS CARTESIA model_idsonic-3sonic-3.5sonic-3.6sonic-3.5-2026-05-04sonic-3.6-2026-08-27sonic-englishsonic-multilingual Choose Sonic 3.5 or 3.6, or a dated snapshot to pin a release. Existing connectors keep their selected model. [help/doc]
TTS CARTESIA accent Optional, Sonic 3.6 only. For multilingual voices, enter an accent ID supported by the selected voice, such as general-american or standard-hindi. Leave blank for the voice default. [help/doc]
TTS CARTESIA speedslowestslownormalfastfastest Speaking speed for the voice
TTS INWORLD workspace Inworld workspace ID (required for voice cloning). Format: workspaces/{workspace} or just the workspace ID. [help/doc]
TTS INWORLD voiceid Voice file name. Works the same as CHIM-XTTS voiceid. Will only work with voices in your Voice Cache. Voice will be automatically cloned to Inworld when first used. [help/doc]
TTS INWORLD languageen-USzh-CNko-KRja-JPru-RUit-ITes-ESpt-BRde-DEfr-FRar-SApl-PLnl-NLhi-INhe-IL Language to use for TTS generation. Uses BCP-47 regional language codes.
TTS INWORLD model_idinworld-tts-1 (deprecated)inworld-tts-1-max (deprecated)inworld-tts-1.5-miniinworld-tts-1.5-maxinworld-tts-2inworld-tts-2-flash Inworld model to use. inworld-tts-2 prioritizes quality; inworld-tts-2-flash prioritizes speed. [help/doc]
TTS INWORLD temperature Sampling temperature (0-2). Higher values make output more random. Default: 1.0
TTS INWORLD speed Decimal: Speaking rate/speed (0.5-1.5). Default: 1.0
TTS INWORLD PARALINGUISTIC_TAGS_ENABLED True False Let the LLM write delivery and sound tags for Inworld, e.g. [smug] Nice swing. [laugh]. With inworld-tts-2 the LLM may also write its own short directions such as [mad as hell]. Tags are removed from subtitles and history. Edit tags
TTSFUNCTION_PLAYERomnivoicehiggspocketttschatterboxxtts-fastapiinworldcartesiapiper-ttsxvasynthnoneazure11labsopenaikokorozonos_gradio Text-to-Speech service options. Used to generate your voice.
TTSFUNCTION_PLAYER_VOICE VoiceID to use for the player character.
TTSFUNCTION_PLAYER_VOICE_ID Integer: Speaker ID to select a voice from multi-voice model.
TTSFUNCTION_PLAYER_LANGUAGE Overrides the TTS language for the player character.
TRANSLATION_FUNCTIONnoneDeepL Translate subtitles and/or audio into a different language.
TRANSLATION settings translate_audio True False This NPC's audio will be translated to the target language.
TRANSLATION settings translate_text True False This NPC's subtitles will be translated to the target language.
TRANSLATION settings save_translated_text True False Replaces the NPC's speech in the context history with the translation. Only used if translate_audio or translate_text is true.
TRANSLATION settings translate_player_audio True False When speaking to this NPC, Player TTS audio will be translated to the player target language.
TRANSLATION settings save_translated_player_text True False Replaces the player's input in the context history with the translation. Only used if translate_player_audio is true.Note: player subtitles are always displayed in their original language.
TRANSLATION DeepL source_language This NPC's source language to be translated from. This should be the language that the LLM responds with.May be left blank for auto-detection. [help/doc]
TRANSLATION DeepL target_language This NPC's target language to be translated into. Required. [help/doc]
TRANSLATION DeepL url Check DeepL endpoint url. Default profile only.Free: https://api-free.deepl.com/v2/translatePro: https://api.deepl.com/v2/translate
TRANSLATION DeepL player_source_language Player's source language to be translated from. This should be the language you speak with STT or input by text.May be left blank for auto-detection. [help/doc]
TRANSLATION DeepL player_target_language Player's target language to be translated into. [help/doc]
TRANSLATION DeepL API_KEY UnhideDeepL api key. Default profile only.Required even when using the free version.
PLAYER_RESPEECH True False Make the Player Respeech connector available. When off, player speech is never rewritten and the connector assignment is kept for later.
STTFUNCTIONdeepgramparakeetwhisperlocalwhispergeminiazureinworldnone Speech-to-Text service options. Translates your voice to text.
STT WHISPER LANG Language to detect for STT.
STT WHISPER TRANSLATE True False Will try to translate to english.
STT WHISPER API_KEY UnhideOpenAI API key.
STT AZURE LANG Language to detect for STT.
STT AZURE profanitymaskedremovedraw Specifies how to handle profanity in recognition results. Accepted values are:MASKED, which replaces profanity with asterisks.REMOVED, which removes all profanity from the result.RAW, which includes profanity in the result. [help/doc]
STT AZURE API_KEY UnhideAzure API key.
STT LOCALWHISPER URL Check Local whisper endpoint. Leave as is if you installed localwhisper through the Distro. [help/doc]
STT LOCALWHISPER FORMFIELDaudio_filefile Form field name for audio file. Sometimes needed to change to file to use another shiper implementations
STT DEEPGRAM API_KEY UnhideDeepgram API key.
STT DEEPGRAM LANG Language
STT DEEPGRAM MODELnova-3nova-2whisper-mediumenhancednovabase Model to use
STT GEMINI API_KEY UnhideGoogle AI API key (from aistudio.google.com).
STT GEMINI LANG Language (e.g., en, es, ja)
STT GEMINI MODELgemini-2.5-flashgemini-2.5-flash-litegemini-2.5-pro Gemini model. 2.5 Flash recommended for best speed and quality.
STT PARAKEET LANG Language to detect for STT.
STT INWORLD API_KEY UnhideInworld API key. Can also be managed from API Badge under Inworld.
STT INWORLD MODEL_ID Model identifier in provider/model format, e.g. groq/whisper-large-v3.
STT INWORLD LANGUAGE Language code in BCP-47 format, e.g. en-US. Leave blank for auto-detect.
ITTFUNCTIONopenroutercustomopenaigoogle_openaillamacpp Image recognition aka Soulgaze spell. OpenAI also works as a connector to OpenRouter!Must be configured in default profile!
ITT custom url Check Endpoint for a custom OpenAI-compatible vision/chat completions service.
ITT custom model Model to use for image analysis.
ITT custom max_tokens Integer: Maximum tokens to generate
ITT custom detaillowhigh Low or high fidelity image understanding
ITT custom API_KEY UnhideOptional API key for the custom service
ITT custom AI_VISION_PROMPTLet's roleplay in the world of Skyrim. Describe this Skyrim image as if it is real life. Describe the environment, objects, and people you see at a fifth grade reading level. Ignore video game HUD and UI elements in your description.Prompt to send to the custom vision model.
ITT custom AI_PROMPT#HERIKA_NPC1# describes what they are seeingPrompt for the AI NPC to follow when describing the scene.
ITT openai url Check OpenAI API or OpenRouter endpoint. Use this for OpenRouter (https://openrouter.ai/api/v1/chat/completions)
ITT openai model Model to use [help/doc]
ITT openai max_tokens Integer: Maximum tokens to generate
ITT openai detaillowhigh Low or high fidelity image understanding [help/doc]
ITT openai API_KEY UnhideOpenAI API key [help/doc]
ITT openai AI_VISION_PROMPTLet's roleplay in the world of Skyrim. Describe this Skyrim image as if it is real life. Describe the environment, objects, and people you see at a fifth grade reading level. Ignore video game HUD and UI elements in your description.Prompt to send to the OpenAI vision model.
ITT openai AI_PROMPT#HERIKA_NPC1# describes what they are seeingPrompt for the AI NPC to follow when describing the scene.
ITT google_openai url Check Google OpenAI API.
ITT google_openai model Model to use [help/doc]
ITT google_openai max_tokens Integer: Maximum tokens to generate
ITT google_openai detaillowhigh Low or high fidelity image understanding [help/doc]
ITT google_openai API_KEY UnhideOpenAI API key [help/doc]
ITT google_openai AI_VISION_PROMPTLet's roleplay in the world of Skyrim. Describe this Skyrim image as if it is real life. Describe the environment, objects, and people you see at a fifth grade reading level. Ignore video game HUD and UI elements in your description.Prompt to send to the OpenAI vision model.
ITT google_openai AI_PROMPT#HERIKA_NPC1# describes what they are seeingPrompt for the AI NPC to follow when describing the scene.
ITT llamacpp URL Check URL of the llama.cpp server [help/doc]
ITT llamacpp AI_VISION_PROMPTUSER:Context, roleplay In Skyrim universe, #HERIKA_NPC1# watchs this scene:[img-1]. Describe the vision while keeping roleplay. Describe COLORS and SHAPESPrompt to send to the llama vision model.
ITT llamacpp AI_PROMPTPrompt for the AI NPC to follow when describing the scene.
ITT openrouter url Check OpenRouter API endpoint
ITT openrouter model Model to use
ITT openrouter max_tokens Integer: Maximum tokens to generate
ITT openrouter detaillowhigh Low or high fidelity image understanding
ITT openrouter API_KEY UnhideOpenRouter API key
ITT openrouter AI_VISION_PROMPTLet's roleplay in the world of Skyrim. Describe this Skyrim image as if it is real life. Describe the environment, objects, and people you see at a fifth grade reading level. Ignore video game HUD and UI elements in your description.Prompt to send to the vision model.
ITT openrouter AI_PROMPT#HERIKA_NPC1# describes what they are seeingPrompt for the AI NPC to follow when describing the scene.
FEATURES MEMORY_EMBEDDING ENABLED True False Enable long-term memory recall. The system injects relevant memory summaries into AI context for responses.
FEATURES MEMORY_EMBEDDING TXTAI_URL Check MiniMe/TXT2VEC service base URL. Use http://127.0.0.1:8082 for the local DwemerDistro service, or set the reachable URL of a remote service.
FEATURES MEMORY_EMBEDDING USE_TEXT2VEC True False Use TXT2VEC to create embeddings for memories. These are more accurate than keywords.
FEATURES MEMORY_EMBEDDING MEMORY_TIME_DELAY Integer: Time in minutes to delay before using a memory in a prompt. Used to avoid pushing recent dialogues as memories.
FEATURES MEMORY_EMBEDDING MEMORY_CONTEXT_SIZE Integer: The amount of the most relevant memory records that will be injected into the prompt.
FEATURES MEMORY_EMBEDDING AUTO_CREATE_SUMMARYS True False Will combine individual memory logs into larger ones. Is more accurate for memory recollection but will use up more tokens. If using koboldcpp, use the multiuser mode to avoid locking.
FEATURES MEMORY_EMBEDDING AUTO_CREATE_SUMMARY_INTERVAL Integer: Time frame used to pack summary data. Each point is about 0.24 in-game hours. Examples: 5 = 1.2 in-game hours | 10 = 2.4 in-game hours | 50 = 12 in-game hours
FEATURES MEMORY_EMBEDDING AUTO_CREATE_SUMMARY_MIN_EVENTS Integer: Minimum number of events required in a packed time bucket before creating a global memory summary. Default: 5
FEATURES MEMORY_EMBEDDING INDIVIDUAL_MEMORY_SUMMARY_THRESHOLD Integer: How many global memory summaries involving an NPC are required before creating one NPC-scoped individual memory summary. Default: 3
FEATURES MEMORY_EMBEDDING MEMORY_BIAS_A Decimal: From 0 (never) to 100 (always). Minimal distance to offer memory.
FEATURES MEMORY_EMBEDDING MEMORY_BIAS_B Decimal: From 0 (never) to 100 (always). Minimal distance to offer and endorse memory.
FEATURES MISC ADD_TIME_MARKS True False Add timestamps to the context logs. Helps with memory recollection.
FEATURES MISC ITT_QUALITY Integer: Only for Soulgaze HD. Compression quality can be set from 0 (lower and unusable) to 100 (near no compression). More quality means higher file size, ergo more tokens.
FEATURES MISC TTS_RANDOM_PITCH True False WIP DO NOT USE! Adjusting the pitch when generating the voice for this actor will add variation, so actors using the same voice sound slightly distinct.
FEATURES MISC LIFE_LINK_PLUGIN True False WIP. Is disabled currently, do not enable is a work in progress.