Phone Calls
Random sims call you in character — your kid's friend, your boss, your ex. Each voice shaped by who they are.
Your sims have things to say. Hand them a phone.
… downloads on CurseForge
… downloads on GitHub
… downloads on Nexus Mods
Every call and text reads your save state before a word gets written. Traits, career, aspirations, family role, who lives with them. Live weather where each sim is standing. Recent milestones, parties you both attended, the in-person chat you had yesterday, group texts you were both in, freeform notes you've written about the sim. All of it lands in the prompt.
So your dad opens with "hey kiddo, how's the hospital?" because he knows you're on the doctor career. Your ex mentions the funeral you both went to last week. Your best friend brings up the storm you're stuck in. And when you tell someone you need space, the mod picks that up — the next reply from them is a boundary conversation, not another cheerful check-in.
A flirty teen texts in lowercase with dramatic slang. An elder writes long, warm paragraphs. A grandparent dotes; an enemy sneers. A parent never opens with "hey man"; a Goofball cracks jokes mid-call. The voice fits.
Random sims call you in character — your kid's friend, your boss, your ex. Each voice shaped by who they are.
Check-ins, group drama, gossip. Hit Reply and they keep the thread going — full back-and-forth, in character.
Dating app for your sim. Opt in for inbound matches, or browse Hinge-style profiles and message someone yourself.
Social media for your sims. Post to friends or the world, and real sims from your save comment, like, and follow. Friends post too, and they bring it up when they call.
Writes back where it matters. Emotional texts apply moodlets and nudge friendship or romance up and down. New sims land in your relationships panel when you engage.
Same code, totally different writing — the AI adapts to each sim's age, traits, and how they feel about you. Here are real texts from the same save:
Your sims are online now. Post to your friends or to the whole world, and the sims in your save react: family, friends, exes, and strangers who've never met you. Sometimes a few people comment, sometimes nobody does. Your friends post about their own lives too, and the people you text and call have seen it all.
Flirty and warm texts now nudge the game's own friendship and romance bars. Plus player-authored context that shapes every prompt, hired service NPCs that write in the right register, and a fifth AI provider option.
Llamafone got its own tile on the phone. And inside: Llamadate. Opt in, browse profiles, message someone, or wait for a match to slide into your DMs.
Alice broke up with Chad. She wants him to stop calling her every hour. She doesn't want him gone from the game — just from her phone. Bob (her brother) is still friends with Chad and should keep getting his texts. This is what per-relationship preferences are for: mute, "asked for space," or favorite — scoped to a specific household member's view of a specific contact.
The paused ("asked for space") state is the interesting one. Instead of blocking the contact, it drops their auto-event rate to 20% and tells the AI you asked for space — so their response is driven by who they are.
Send one message to 2–4 sims at once. Each recipient replies in their own voice, staggered like real texts arriving over a few minutes. Every reply sees what the others just said — so responses stay distinct instead of a chorus of "same." The dialog title and the reply prompt both list the full group roster, so you always know who you're talking to.
A briefing call at group creation (using the default model) synthesizes each participant's voice and their web of relationships, then gets cached for every subsequent reply — so per-turn cost stays sane even over long threads. Persisted per-save in GroupTexts.json so quitting mid-thread doesn't lose it.
No cheat console required for everyday use. Call, text, and tweak settings from the in-game phone the way your sim would.
llamafone.cfg, keeping your comments.llamafone.cfg.Every call and text is filtered through what's true in your save right now — not generic AI fill.
If a sim lives in your world, they'll suggest grabbing coffee. If they live in San Myshuno and you're in Willow Creek, they keep it long-distance — no "come over" from across the map.
Live read of the weather where each sim actually is — light rain, blizzard, heatwave, thunderstorm. Dramatic weather comes up in conversation; routine weather stays background. Off-world callers know THEIR climate too, so Sulani friends won't pretend it's snowing during a globally-Winter season.
Sim quit a job, got promoted, had a baby, got married, aged up? The mod catches it automatically and your friends bring it up — "heard you quit, you good?" — within days of it happening.
Text your sim's dad while he's on shift at the hospital and he might write back "sorry, at work, catch you later" — not generic small talk. Reads live career state — currently at work, or should be but isn't — plus school schedules for teens and kids, so people who should be busy sound busy.
The butler, maid, babysitter, gardener, and repair tech write in the right register. Butlers stay professional and deferential; maids talk supplies and schedules; babysitters check in on the kids; gardeners talk plants and season; repair techs quote parts and timing. The mod reads the household's live service records so a hired NPC calling about tomorrow's schedule lands like they should, not like a random friend.
Sims 4 drops calendar events from the UI the moment they end. Llamafone snapshots them. Text a guest the morning after a house party — "great party last night." Text your ex the day of a funeral you both attended. Love Day and Winterfest come up on the day.
When your sim actually chats, flirts, argues, or kisses another sim in-game, the mod logs it. Next time you text or call, the AI knows you saw each other in person — so the opener is "great catching up earlier," not generic small talk. Uses the same Sims 4 relationship bits (Just Chatted, Just Kissed, Social Context) the game itself tracks.
Your Outgoing Adventurous dad doesn't text you like a frat bro — he texts like your dad. Parents sound parental, siblings tease, spouses get intimate, grandparents ramble warmly. Traits are interests, not voice.
Friends are warm. Acquaintances are polite. Enemies are openly hostile — no "rooting for you 💕" from someone who hates your guts. The current relationship beats anything from old chat history.
Write a freeform note on any contact — "kid's teacher," "asked for space after breakup," "boss's boss," "childhood best friend." Surfaced in every prompt for that pair, weighted heavily, so the AI stays in character on nuance the game itself can't label. Per household member — Alice's notes stay Alice's.
Mute a sim or ask them for space in the Contacts UI, separately for each household member. Every subsequent AI prompt for that pair leads with the boundary, not with warmth. A love-heavy ex ignores it and keeps calling; a decent friend apologizes; a mean one guilt-trips. The drama plays out based on who they are.
When Alice later texts Sarah 1:1, the AI knows they were both just in a group text with Bob and Kate. The last few turns from that group thread land in the prompt — so 1:1 conversations reference shared context like real friend groups do. Runs both directions, capped to recent threads only.
Post on Llamagram and your friends saw it. Next time one of them texts or calls, they can bring it up — "saw your post, are you okay?" — and they remember what the two of you said in the comments. Works both ways: you can ask about their latest post too. Only the last few in-game days count, and only posts they could actually see, so a friends-only post never reaches a stranger.
Every prompt now carries the current in-game clock — 7:15 PM (evening), 9:03 AM (morning), etc. No more "dinner ready by six" from your butler when it's already 7. Times of day sound like real times of day, and future-tense references (tonight, tomorrow, next week) land in the right window.
Match with someone on Llamadate today, text them three in-game weeks later, and the AI still knows the origin — that's how they'll bring it up. Mutual-friend intros remember the mutual's name too. Stored on the pair alongside contact preferences, ready for future asymmetric events (who kissed who, who proposed to who).
Open the cheat console with Ctrl+Shift+C, type a command, press Enter.
llama.call — incoming callllama.text — incoming textllama.sendtext Bella Goth hey!llama.sendcall Bella Goth I have newsllama.reply <message>llama.post hello world — post to friendsllama.post public hello worldllama.npcpost — a friend posts nowllama.inbox — notificationsllama.feed — followers & recent postsllama.status — setup & all commandsllama.auto_events on — timed calls & textsllama.chat <message> — freeformllama.journal — recent historyllama.reload — after editing configFor storytelling. The base game's phone calls and texts are limited dropdown options — useful for gameplay, repetitive for narrative. This mod replaces them with text written specifically for your household by the AI provider you configure (Claude, OpenAI, Gemini, or a local model through Ollama or LM Studio).
If you're a storytelling-focused player who's tired of the same scripted lines and wants your sims' relationships to feel alive, this is for you.
The mod never edits your save files. It reads game data (who lives in the household, their traits, careers, and relationships) to write its calls and texts. A few things it does change in-game, which the game then keeps when you save: messages nudge friendship and romance (turn it off under Llamafone Settings), a charged message can give a sim a moodlet, and with Get Famous a Llamagram post that truly goes viral can raise fame.
Llamafone's own data (journal, sim bios, save notes, Llamagram posts, and so on) lives in its own folder per save: Documents/Electronic Arts/The Sims 4/saves/Llamafone/Slot_…. Those are plain JSON files you can open and edit. To uninstall, delete Llamafone.ts4script, Llamafone.package, and llamafone.cfg from your Mods folder, and the saves/Llamafone folder if you don't want to keep the history.
Just what's needed for the current command — names, ages, traits, careers, aspirations, household members, and relationship history. Nothing personal about you, no real names beyond what your sims are named, no save file contents.
Your prompts go from your computer directly to whichever AI provider you configured — there's no middleman server. Each provider has its own data-retention policy: Anthropic, OpenAI, Google, and OpenRouter all state that paid API traffic is not used to train models. If you really don't want any data to leave your machine, use Ollama — it runs entirely locally.
Yes. The key lives only in llamafone.cfg on your computer. It's never uploaded anywhere except to make the API call to whichever provider you chose. The mod doesn't have a server, doesn't phone home, and isn't open to anyone else.
That said, treat the key like a password — don't paste it in screenshots, Discord, or screen-shares. If you ever suspect it leaked, revoke it on your provider's dashboard and make a new one. (Or just switch to Ollama or LM Studio — no key, no exposure.)
Yes. Everything is opt-in. Auto-events (random calls/texts/stories while you play) can be toggled with llama.auto_events off in the cheat console. You can also edit llamafone.cfg to disable specific auto-event types, change interval/chance, or restrict which sims can call you.
Auto-events are off until you turn them on. Birth announcements are on by default; turn them off under Llamafone Settings (or birth_announcements = false in the cfg). Llamagram can be turned off, kept with its pop-ups silenced, or have friends post less (or never) under Llamafone Settings > Llamagram.
Yes — drop the .ts4script into ~/Documents/Electronic Arts/The Sims 4/Mods/ and it works identically to Windows. Same commands, same behavior.
You pay your AI provider directly for what the mod actually uses — there's no subscription to me or anyone else. Cost depends on which provider you pick:
Prices and model names change. These are rough estimates from October 2026; check your provider's pricing page for current rates.
| Provider | Model | Per call |
|---|---|---|
| Anthropic Claude | claude-haiku-5-5 (recommended) | ~$0.001 |
| Anthropic Claude | claude-opus-5-5 | ~$0.05 |
| OpenAI | gpt-4o-mini | ~$0.003 |
| OpenAI | gpt-4o | ~$0.03 |
| Google Gemini | a current Flash model | free (rate limited) |
| Ollama / LM Studio (local) | any model | free (uses your GPU) |
A typical session of ~30 calls costs a few cents on Claude Haiku 5.5 and under a dime on GPT-4o-mini. Heavier sessions on the premium models (like Claude Opus) can run a dollar or two. Gemini's free tier covers most casual play. Ollama and LM Studio are fully free if you've got the hardware.
Llamagram adds a little. Each of your posts is one larger call (it writes all the comments at once), and each reply you post to a comment is one more. With auto-events on, friends posting on their own is one call per post, about one in five auto-events at the default setting. Set social_npc_post_weight = 0 to turn friends' posts off.
No. All API calls happen on background threads, so the game keeps running while text is generated. You'll see a notification popup when the response arrives, usually within a few seconds.
The mod needs an internet connection to call the AI provider. If you're offline, commands silently fail without interrupting your game — no crashes, no popups. Your sims still play normally; you just won't get AI-generated content during offline sessions.
No. The Sims 4 supports script mods officially through its Mods folder and Custom Content settings. This mod uses the documented script-mod system and doesn't modify any game files. EA's stance on script mods is that they're allowed; they just won't support issues that occur while mods are installed.
Three places to grab it, same files everywhere — CurseForge (recommended for most players), Nexus Mods, or the GitHub releases page. You need three files: Llamafone.ts4script (the script), Llamafone.package (the phone UI tunings), and llamafone.cfg (the settings file).
Llamafone works with six different AI providers — pick whichever fits your budget, preferences, or paranoia. You only need credentials for the one you pick. Ollama and LM Studio are free because they run entirely on your computer.
Model names change. AI companies retire old models and release new ones every few months, and prices change too. The names below were current as of October 2026; check your provider's model list for the latest. A "model not found" error usually means a model was retired. If you get empty, cut-off, or very slow replies, try a model that doesn't "think" (reason) before answering, like gpt-4o-mini or, locally, Llama 3.2.
Replies seem confused? Try a different model. On Claude, turn up Thinking to xhigh (it's already the default on Haiku 5.5).
What the mod was built and tuned against.
claude-haiku-5-5)sk-ant-api03-… key immediately — you only see it onceIn llamafone.cfg: provider = claude. Recommended model: claude-haiku-5-5; for richer writing try claude-sonnet-5-5 or claude-opus-5-5.
Works with GPT models like gpt-4o-mini and gpt-4o.
sk-… key — only shown onceIn llamafone.cfg: provider = openai + set model to gpt-4o-mini
Has a generous free tier — you can play for a while without a card on file.
AIza… keyIn llamafone.cfg: provider = gemini + model set to a current Flash model from Google's model list (Flash models are on the free tier).
Aggregator that proxies dozens of models (Claude, GPT, Llama, Mistral, DeepSeek, etc.) behind one API key. Useful if you want to A/B-test models without juggling separate accounts.
sk-or-… keyIn llamafone.cfg: provider = openrouter + model set to anthropic/claude-haiku-4.5 (browse the full catalog at openrouter.ai/models)
Runs on your own computer. No API key, no internet, no cost. More setup than the cloud options — you install Ollama, keep it running in the background, and pull a model yourself.
ollama pull llama3.2:3b (small, fast) or a bigger one if your hardware handles it.api_key blank.In llamafone.cfg: provider = ollama + model set to llama3.2:3b (must match ollama list). Run llama.testconnection in the cheat console to verify setup.
Also runs on your own computer, with an app for finding and loading models. No API key, no internet, no cost.
api_key blank.In llamafone.cfg: provider = lmstudio + model set to the model's name in LM Studio. Run llama.testconnection to verify setup — it lists the names LM Studio offers.
Move both downloaded files into:
Documents/Electronic Arts/The Sims 4/Mods/
If you have Sims 4 mods already, you've been here before — same folder.
Open llamafone.cfg with any plain text editor (TextEdit, Notepad, VS Code — anything that saves plain text). Two lines to update near the top:
provider = claude # or: openai, gemini, openrouter, ollama, lmstudio
api_key = YOUR_API_KEY_HERE
Replace YOUR_API_KEY_HERE with the key you copied. Examples per provider:
provider = claude
api_key = sk-ant-api03-Xy7…
# or
provider = openai
api_key = sk-proj-…
# or
provider = gemini
api_key = AIza…
# or
provider = openrouter
api_key = sk-or-…
model = anthropic/claude-haiku-4.5 # vendor/model form -- see openrouter.ai/models
# or (no key needed)
provider = ollama
api_key =
model = llama3.2:3b # must match the model you pulled -- see `ollama list`
# or (no key needed)
provider = lmstudio
api_key =
model = llama-3.2-3b-instruct # the model's name in LM Studio
Save the file. If you're on Ollama, also confirm ollama_endpoint = http://localhost:11434 matches wherever your server runs. Ollama users can run llama.testconnection in the cheat console after loading a save — it verifies Ollama is reachable, the model is downloaded, and end-to-end generation works. LM Studio users: the server runs at http://localhost:1234 by default; if LM Studio shows a different address, set lmstudio_endpoint to it. llama.testconnection checks LM Studio the same way.
In The Sims 4, open Game Options → Other and tick both:
Restart the game. When the main menu loads, you'll see a notification popup confirming the mod is active.
Inside any household, open the cheat console with Ctrl+Shift+C and type llama.status to see all the commands and confirm everything's wired up.
Something not working? See Common errors below.
Got an error? Find the message you saw below. Most are a quick settings fix. First step for anything: type llama.status and llama.testconnection in the cheat console.
Llamafone isn't reading the settings you edited, so it's using its defaults. Type llama.status: it shows the provider, the llamafone.cfg file, and the Mods folder Llamafone is actually using.
provider = line itself. Setting the Ollama address or models isn't enough; it has to say provider = ollama (or lmstudio, openai, and so on).llama.reload in the cheat console (or restart the game).Llamafone needs Windows 10 (version 1803 or newer) or Windows 11. It reaches the AI through curl, a small tool built into those versions of Windows. Windows 8 and older don't include it.
The key in api_key is wrong, or your account with the provider has no credit. Copy the key again from your provider's dashboard (keys are only shown once, so make a new one if you need to), and check that credit is loaded under the provider's billing settings. Then llama.reload.
The model (Model in AI settings, or model in the cfg) has been retired by the provider. AI companies retire old models every few months. Pick a current one from your provider's model list and put it in both lines, then llama.reload. For Claude, claude-haiku-5-5 is a good choice; for Gemini, gemini-3.5-flash-lite. Google retires Gemini 2.5 on October 20, 2026, so if you use a gemini-2.5 model, switch before then.
Fixed in v3.8.3 — update to the latest version. Gemini 3 models (like the free gemini-3.5-flash-lite) then work. Gemini 2.5 is being retired on October 20, 2026, so move off it before then.
Usually a model that "thinks" (reasons) before answering, and spends its whole reply on thinking.
gpt-4o-mini or, locally, Llama 3.2.http://localhost:1234.A call or text sends the AI about 8,000 tokens. Reload the model with a larger context length: 16384 (12288 at the very least). In LM Studio, set it when loading the model. In Ollama, set Context length to 16k in its settings.
Your PC is taking too long to write the reply. Without a supported graphics card, a local model can take minutes. Update to v3.8.2 or newer (it waits up to 5 minutes and tells you if it gives up), and try a smaller model, like a 3B one (llama3.2:3b).
Ollama's default context is too small, and it quietly cuts off the start of Llamafone's prompt, which holds most of the instructions. In Ollama's settings, set Context length to 16k.
Run llama.testconnection: it lists the models your Ollama or LM Studio actually has. Set Model to one of those names (Llamafone Settings > AI settings, or model in the cfg, then llama.reload), or install the model you want (ollama pull <name>, or load it in LM Studio).
The game has lost keyboard focus, usually after alt-tabbing to another window. If the cheat console won't take typing either, restart the game. Try to avoid alt-tabbing while a message box is open.
’ or —Fixed in v3.8.2 — update to the latest version.
Still stuck? Report it and include the output of llama.status and your Llamafone_Log.txt (in your Documents folder).
What's changed in each release, newest first.
Fixes from your reports. Drop-in over v3.8.4.
Change almost everything without leaving the game, and sims that know more about each other. Drop-in over v3.8.3.
llamafone.cfg.llamafone.cfg files have a single model line instead of default_model and fast_model, and it's used for everything. Your existing cfg keeps working as is: if it has no model line, Llamafone uses your fast_model (then default_model), the same model your texts and calls already used.Keeps Llamafone working with the newest AI models. Drop-in over v3.8.2.
gemini-3.5-flash-lite) failed with "Request contains an invalid argument." Gemini 2.5 is being retired on October 20, so if you use Gemini, update before then.llamafone.cfg, with a reminder that model names change.Sims remember vacations, a new free local AI option, and a round of bug fixes. Drop-in over v3.8.1.
New:
llama.trips to see what's been recorded.provider = lmstudio to run Llamafone on a model loaded in LM Studio: free, private, no API key. Load the model with a context length of 16k. llama.testconnection checks the server, lists your models, and tells you exactly what's wrong if something's off.Other bug fixes:
Llamafone_Log.txt.Where the mod is heading. Not a contract — priorities shift based on what you all report and ask for.
Calls and texts that fire because something happened in your save, not on a timer. The first one shipped in v3.7: a family member or close friend has a baby → the new parent texts you within seconds. The plumbing behind it (detecting the event, picking who'd reach out, dedup, an off switch) is what the rest build on.
Next in line: pregnancy announcements (a close friend takes the test → "we're expecting", a few hours later). After that: get dumped → your best friend checks in; promotion → mom calls; thunderstorm → dad makes sure you're okay.
v3.7 introduced a model of who knows what: pregnancies and births reach parents in hours, cousins in a day, casual friends in a few, strangers only if told. The same model can gate breakups, divorces, job loss, and deaths — so an acquaintance doesn't open with condolences before your own sister has heard.
Right now sims remember recent messages, and life events and trips for about a week of game time. Core memories would be the moments that define a relationship and never fade: the day Apollo walked Francesca down the aisle, the fight Ingrid and Ben never got past, the vacation in Sulani. A short permanent list for each pair of sims, picked out of big conversations and events, and something you can view, edit, or pin yourself.
Your sim takes a photo with their phone in-game and texts it to someone or posts it to Llamagram, and the AI actually looks at the picture: replies and comments react to who's in it, where they are, and what they're doing. Needs a vision-capable model. Still confirming the mod can reliably grab the exact photo your sim took.
The roadmap reflects what's bubbled up so far. If something would make your save more fun — a feature, a tweak, a weird hypothetical — pitch it. Bug reports and feature requests use the same form.
Submit an ideaThe mod is actively maintained and bug reports get read. Use the form below and you'll be filling out a GitHub issue with structured fields (what happened, log excerpt, screenshots) — no GitHub experience required, just sign in.
Report a bugOr browse existing reports first to check if it's already known.
Prefer Bitcoin? Scan or copy the address:
351FwouuUN7QRNGy6iDXXFwr7rgt4qDnHx