This tutorial is available in English · 中文 · 日本語
Customize Your Companion: Four Steps to Make Them Yours
A fresh install puts a default avatar on your desktop. It can sit there, get dragged around, and react when you pet it — but it is not your character yet. This page walks the whole route: swap the look for a character you actually like, write down who they are and how they treat you, then make them understand what you say and answer in a voice of their own. Four steps, one for each piece: look, character card, brain, voice.
Two things worth knowing before you start. You don't have to finish in one sitting — do the first two steps and your character is on your desktop today. And importing your own character is a full-version feature; the free version lets you run the same flow with the built-in characters, same screens, same clicks.
What each of the four steps changes
People rarely get stuck on a step because it's hard. They get stuck because nobody told them a "custom companion" is assembled from four separate parts, and each one adds a different ability:
| Step | What it decides | Where you do it | Section |
|---|---|---|---|
| 1. Look | What they look like — your own VRM model replacing the default avatar | Workshop → Add character | Jump to step 1 |
| 2. Character card | Who they are — name, personality, way of speaking, your relationship | Workshop → character card | Jump to step 2 |
| 3. Brain | Whether they can follow you and answer — LLM plus speech recognition | Settings → AI Services | Jump to step 3 |
| 4. Voice | What they sound like — the TTS voice | Settings → AI Services | Jump to step 4 |
The first two build the character itself, and both live in the Workshop. The last two are the services behind the character, and both live in Settings. That split matters: swapping a look or editing a card is a local change you can undo anytime, while connecting an AI service means getting an API key from a provider — the only step that needs anything from outside the app.
Step 1: The look — how do you get your own VRM character onto the desktop?
The look is their face. AniMate uses VRM models, the format most anime-style characters are published in — VRoid, BOOTH, and the overwhelming majority of author pages all hand out .vrm files. Your job is to get one into the app and replace the default avatar with it.
The route: right-click the running avatar on your desktop and open the Workshop. Click Add character at the bottom left, pick the .vrm file you downloaded, fill in a name, thumbnail, and a tag or two, then confirm the import. Select the new card and click Apply — your character is now the one standing on your desktop.
Three things that trip people up. The file has to be .vrm; .pmx and .fbx need converting first. Importing your own character requires the full version, so the free version will show an activation prompt at that button — everything else in the flow looks the same. And check the model's license before you record, stream, or post screenshots with it. For the click-by-click version, see how to import a VRM model; if you don't have a model yet, start with where to find free VRM models.
Step 2: The character card — where do their name, personality, and backstory live?
The character card is what turns a model into a person. The same model with two different cards behaves like two completely different characters, which is the part most people miss on their first try — and the part that is actually creative work. Open the card in the Workshop and you'll find a handful of fields:
- Identity: their name, what you call them, and what kind of being they are.
- Personality and speech style: how they talk, whether they're warm or reserved, any verbal tics.
- Your relationship: partner, roommate, co-worker, familiar — this is what decides how they address you.
- Service binding: which AI services this card uses, either the global defaults or an override.
One rule of thumb: the more specific the card, the steadier the character. "Gentle" gets you nothing you can rely on; "answers half a beat slow, explains things with food metaphors, bad at saying she cares" keeps them recognisable fifty messages later instead of only for the first line. If you'd like a reference for how the fields read, open the built-in Faelmu card — it's the read-only product assistant that uses the global defaults, and it's a decent template to copy phrasing from.
Step 3: The brain — what do the LLM and speech recognition each do?
The "brain" isn't one thing. It's two services that do completely different jobs, and you can add them separately:
| Capability | What it does | What happens without it |
|---|---|---|
| LLM | Conversation and reasoning: how they read your message and compose a reply | The panel tells you no service is configured, and sent messages get no answer |
| Speech recognition (ASR) | Turns your speech into text — voice input and real-time voice rely on it | Typing still works, but they can't hear you talk |
Both are added under Settings → AI Services, and both want an API key from a provider. This is the only step in the whole route that needs something from outside the app, and it's where most people stall: it isn't a couple of clicks, it's signing up somewhere, copying a key, and pasting it back. So we wrote it up separately — see how to get free LLM credits, and the same idea for speech recognition.
Once it's set, go back to the desktop, right-click, and open the conversation panel. Type something. If you get an answer, your character has a brain.
Step 4: The voice — how do you pick a TTS voice?
The voice is the last piece and the one that changes the most about how they feel. TTS reads their replies out loud, and the voice you choose is the difference between bubbly, aloof, and clingy.
The entry point is TTS in Settings → AI Services. Configuration works like the previous step — it needs a key — but TTS providers usually let you choose a voice, often dozens of them. Pick one that matches the personality you wrote into the character card, and they'll keep using it. For getting started without paying, there's a guide to free TTS credits.
The voice is optional. Skip TTS and conversations still work as text; they just don't speak. Some people prefer it quiet, and that's a perfectly good way to run things.
You don't have to do it all at once
Doing all four steps is the goal, but don't let it look like one big project. In practice it comes in two halves. With the look and the character card done, you have a desktop companion you can see and interact with: it sits with you, follows your cursor, reacts to head pats, dances to your music. Add the brain and the voice and it stops just being with you and starts talking to you.
So the order is simply top to bottom: get the look in first, since it works the same day; write the card when you feel like it, because the personality tends to clarify itself while you write; and come back for the LLM, ASR, and TTS when you actually want them to talk. Nothing here leaves you in a broken half-state, and you can stop wherever you like.
Common questions
Do I need to know 3D modelling or coding to customize a companion?
No. All four steps happen inside the app: import a model, fill in a card, paste a key, choose a voice. The only genuinely creative part is the writing on the character card, and that's writing, not technical work. You don't need to build a model either — using an existing VRM is the normal path.
Do I have to pay to import my own VRM model?
Importing your own character is a full-version feature. The free version lets you experience the whole interface and every interaction with the built-in characters. Our honest suggestion: run the four steps with a built-in character first, and only activate once you're sure you'll use it — that way you're not paying because of one specific model and then discovering the workflow doesn't suit you.
Do I need an LLM, ASR, and TTS all at once? Can I chat with only an LLM?
They're independent. An LLM alone gets you text chat, which is the most common minimal setup. Add ASR when you want to talk instead of type, and TTS when you want them to answer out loud. Each has its own free-credits guide, so you can start with just the LLM one.
Can I see what chatting feels like before I have any API key?
Yes — during first-run setup the character sends you a short tutorial message and answers one question, and none of that needs an AI service. It's there precisely so you can see what a conversation looks like before configuring anything. Come back once you've added a provider and keep going.
If I change the look or the character card, do I lose the conversation history?
Conversation history is kept per character card, so switching back to a card brings its history with it. That means you can keep one card for testing a personality and another for a different character without losing either. Exact behaviour depends on the current version.
Next: get the look in place
Of the four steps, importing a model is the most visible and the fastest to pay off. Follow the step-by-step tutorial and your character will be on your desktop today.