JoiHUD AI gives the characters in your Virt-A-Mate scenes a personality and a voice. Add it once as a session plugin, point it at any look, and she/he talks back.
Two ways to run it:
- FREE — Local mode. Your own LLM on your own PC. No account, no credits, no card, nothing sent to iamjoi. Ever.
- PREMIUM — the hosted service: Start Scene, saved Contexts, scene actions, up to three characters (plus yourself) and the web Playground you can drive from your phone.
Same plugin. Same HUD. Start free, upgrade only if you want the big stuff.
Btw: this offers all the same features as my previous and free JoiHUD plugin (JoiHUD-.cs), but with extra AI features.
Clear up any doubts in the iamjoi Discord: https://discord.gg/xbZyeadhAh
WATCH THE TRAILER
The looks shown in the video were chosen randomly: Beryl by Xspada, and Ruby by Best-VR-Babes.
Special thanks to the deep ambient techno track "Citadel" by lowkey ambient.
Search YouTube for "this is what deep ambient techno is supposed to feel like" (youtube.com/watch?v=eeD0bjlVggo).
You can buy the whole album via the link in his video description.
═══════════════════════════════════════════════
LOCAL MODE. FREE. OFFLINE. YOURS.
═══════════════════════════════════════════════
You can now run JoiHUD AI end to end on your own machine. Your LLM. Your TTS. Your STT (Speech-To-Text). No iamjoi account, no Premium, no credits. Install iamjoi Link, stay signed out, and the whole loop — you speak, she thinks, she answers out loud, her lips move, she takes her top off if you ask — happens on your PC.
Nothing is sent to iamjoi in local mode. Not your chat, not your looks, not your keys. Once the model is downloaded you can pull the network cable and keep going.
Download iamjoi Link (version 4.0.0 or higher) PC app for free, here: https://www.patreon.com/iamjoi/posts/iamjoi-link-for-154276984
What you get for free:
- Your own LLM. Anything OpenAI-compatible. Ollama is the one-click path (Dolphin 3 recommended, ~5 GB, uncensored and stays in character; Llama 3.1 8B and Qwen 2.5 7B are also in the list and download from inside Link). LM Studio and friends work too — just point Link at the endpoint.
- Invented personalities. Press Start Character (in the HUD) on any look and the model writes her a personality on the spot. Don't like it? Edit it in Link (the complementary PC app), or write your own and it sticks.
- Local voice with lip sync. Kokoro runs inside Link. Ten catalog voices, and Change Voice button in the HUD rolls a new one mid-session. Costs nothing.
- Speak to her with your voice. The HUD mic has two modes: local Whisper, which never leaves your PC, or Chrome speech, which is noticeably sharper but sends the audio to Google's speech model (never to iamjoi). Your call, one dropdown.
- Bring your own ElevenLabs key if you want the high quality TTS. Paste it into Link and local mode speaks in ElevenLabs voices, billed to your ElevenLabs account at your rate. The key is stored only on your PC — iamjoi never sees it, never proxies it. Pick a catalog name or paste any voice id from the ElevenLabs voice library.
- Or your own speech server. Any endpoint exposing POST /audio/speech returning WAV. Or turn voice off entirely and read subtitles.
- Clothing that responds to conversation. Ask her to take her top off and she does — top, bottom, undertop, underbottom, accessories. Optionally hand the outfit decisions to a second, sharper model so your main chat model never gets distracted by wardrobe logic.
- A Try it box in Link so you can talk to her before VaM is even running.
What local mode does NOT do (this is Premium):
Start Scene, saved Contexts, scene and pose actions, vision / look stills, more than one AI character, wardrobe search and acquire, and the web Playground.
Sign into Link and HUD chat switches to the hosted AI instead of your local model.
Local mode is restricted to one character, HUD Start Character and HUD chat bar.
LOCAL MODE — SETUP (about ten minutes, most of it downloading)
1. Install iamjoi Link (Windows). Do not sign in — or log out if you already did.
2. Open the Local tab and turn Local mode on.
3. Set your Root VaM Directory (needed for pairing and audio).
4. Run a local LLM. Easiest: install Ollama, leave the endpoint at http://127.0.0.1:11434/v1, and hit Download on Dolphin 3 inside Link (~5 GB).
5. Optional: write a one-line scene description. Personality is invented on Start Character; you can edit it afterwards.
6. In VaM, add JoiHUD as a Session plugin, click ALLOW ALWAYS if asked, select a person and press Start Character (not Start Scene).
7. Open the chat bar (gear icon → Chat Bar) and talk. Change Voice is on the extra HUD page.
Local mode requirements:
- Windows (iamjoi Link is a Windows app).
- A PC that already runs VaM. For the LLM itself, a 7–8B model wants ~8 GB RAM and ~5 GB disk. A GPU makes her answer faster; it is not required.
- This package — the full JoiHUD AI (the HUD title bar should read "JoiHUD AI"). JoiHUD-.cs (minus) has no bridge and cannot do this.
- Internet once, to fetch Link, Ollama and the model. After that, offline is fine unless you are using the chrome STT option.
═══════════════════════════════════════════════
IT WORKS WITH THE SCENES YOU ALREADY HAVE
═══════════════════════════════════════════════
Not scenes authored for iamjoi. The scenes you downloaded years ago, from whichever creators you follow. Add JoiHUD as a session plugin, load a scene, and the characters in it can talk after a couple of clicks.
Scenes with no pre-made Context
Hit Describe, type one sentence about what is going on and who everyone is, and the AI writes the scene for every character in it — including the private details only she knows.
Scenes with a pre-made Context
If a scene does have a Context, setup is one click, and the characters also get scene actions they can choose to take on their own: change position, move to another location, whatever that Context defines.
═══════════════════════════════════════════════
PREMIUM — WHAT THE HOSTED SIDE ADDS
═══════════════════════════════════════════════
No LLM to host, no model to download, no server in your tray, no GPU budget. Way smarter, uncensored LLMs at a slightly higher latency cost.
Link signs into your account and connects VaM to AI running on AWS.
Download iamjoi Link (version 4.0.0 or higher) PC app for free, here: https://www.patreon.com/iamjoi/posts/iamjoi-link-for-154276984
1. Drop the .var into AddonPackages/.
2. Add JoiHUD as a Session plugin (not on a Person atom).
3. Run iamjoi Link and sign in.
4. Open the Playground in any browser — your phone works fine — and press Start Scene in the HUD (it's better if you click Add Plugins right before this).
You manage Contexts from the Playground on your phone or PC, and tweak everything else from the HUD in-game.
- Personality from a look. Point it at any VaM appearance and iamjoi writes the character from what she looks like. Save a personality and it comes back on that look every time; it gets stored in your root VaM directory, in a new special folder (iamjoiLinkData).
- Voice. Spoken dialogue with lip sync, plus in-scene subtitles.
- STT. Speak to the characters in your own voice, in your preferred language.
- Saved Contexts. Build a scene once — the setting, each character's role, the actions they're allowed to take — and reload it exactly as you left it.
- Scene awareness, both directions. She reacts to what happens in VaM through the trigger system, and what she says can fire animations, expressions, clothing changes and scene events back.
- Up to three AI characters in one scene.
- Describe. One sentence in, a full scene brief out: shared context for everyone, plus private notes each character keeps to herself.
- Add Plugins button. Adds a configurable list of plugins to every male and female character in the scene. Two come pre-filled — Embody by AcidBubbles and BodyLanguage by CheesyFX (in my opinion the two best plugins in all of VaM) — and you can swap them for whatever you prefer in the plugin's Custom UI. It only attaches plugins already sitting in your AddonPackages; it does not download anything for you.
BEING STRAIGHT WITH YOU ABOUT THE TRADE-OFFS
Local mode (free)
- Nothing reaches iamjoi. No chat, no look images, no API keys.
- Two exceptions, both your own choice and both clearly labelled in Link: if you pick Chrome speech for the mic, your spoken audio goes to Google's speech
model; if you paste an ElevenLabs key, the reply text goes to ElevenLabs and is billed to you. Pick local Whisper and Kokoro and neither happens.
- Quality is your hardware's quality. A 7B model on a laptop is not Grok. Chat history lives in memory and dies when you close Link. Scene text and personality are saved to this PC.
Premium (hosted)
- It is a hosted AI service. Virt-A-Mate never opens a connection to iamjoi servers: JoiHUD AI talks only to iamjoi Link on your PC. Most chat is typed on the iamjoi website and goes from your browser to our servers over HTTPS. Link relays VaM-side data to the same servers — look stills for personality generation, scene and session info, clothing-search hits if you use that, and HUD chat if you type in-scene. We can read that session data while it exists. Signing in is the control you have; there is no per-field filter of what is sent.
- Replies use the cloud LLM you pick in the Playground (OpenRouter, Nano or xAI, depending on the model). Some helper steps — Describe expand, deciding who
speaks next — currently use xAI even if your chat model is not Grok.
- Voice: Kokoro can stay on your PC via Link and cost no credits. Choose cloud TTS and the reply text is sent to ElevenLabs or xAI.
- Link can write a small set of files under your VaM folder at iamjoiLinkData/ (saved personalities and appearance notes), when the Playground asks it to. Not arbitrary paths on your disk.
- No long-term memory yet. She follows the conversation within a session. Saved personalities and Contexts come back; chat history does not carry across sessions yet.
- Light on configuration. Few knobs, sensible defaults.
Full detail: https://iamjoi.ai/redTape/privacyPolicy
WHAT YOU NEED
Local mode: Windows, VaM, iamjoi Link, a local LLM. That's it. Free.
Premium: an internet connection and an iamjoi Premium membership, which comes with 500 AI credits. Extra credits are $1 = 100 credits. Credits are usage-based (LLM, plus TTS if you use cloud voice — local voice is free either way).
Get access on Patreon → https://www.patreon.com/iamjoi/membership
This is a similar but different plugin to JoiHUD-.cs (aka JoiHUD minus), which is fully free and has no external-service option at all; you can find it here: https://hub.virtamate.com/resources/joihud-the-ultimate-scene-control-hud.67247/
Looks, scenes and other third-party content shown in the videos and screenshots are for demonstration and are not part of this package.
CREDITS (CODE IN THIS PACKAGE)
JoiHUD includes code adapted from "Utils" by MacGruber (CC BY 4.0). Specifically the custom-trigger UI: SimpleTriggerHandler and EventTrigger, derived from MacGruber's CustomTrigger / EventTrigger. The MacGruber_Utils.cs file is not included in this .var.
Embody by AcidBubbles and BodyLanguage by CheesyFX are optional Hub plugins the Add Plugins button can attach. They are not bundled with this package.
HOW THE HUD LOOKS
HOW IAMJOI LINK LOOKS
Please click "Allow Always" for the HUD to work properly
A review from a user
Questions, bugs and feature requests are welcome in the discussion thread!
Two ways to run it:
- FREE — Local mode. Your own LLM on your own PC. No account, no credits, no card, nothing sent to iamjoi. Ever.
- PREMIUM — the hosted service: Start Scene, saved Contexts, scene actions, up to three characters (plus yourself) and the web Playground you can drive from your phone.
Same plugin. Same HUD. Start free, upgrade only if you want the big stuff.
Btw: this offers all the same features as my previous and free JoiHUD plugin (JoiHUD-.cs), but with extra AI features.
Clear up any doubts in the iamjoi Discord: https://discord.gg/xbZyeadhAh
WATCH THE TRAILER
The looks shown in the video were chosen randomly: Beryl by Xspada, and Ruby by Best-VR-Babes.
Special thanks to the deep ambient techno track "Citadel" by lowkey ambient.
Search YouTube for "this is what deep ambient techno is supposed to feel like" (youtube.com/watch?v=eeD0bjlVggo).
You can buy the whole album via the link in his video description.
═══════════════════════════════════════════════
LOCAL MODE. FREE. OFFLINE. YOURS.
═══════════════════════════════════════════════
You can now run JoiHUD AI end to end on your own machine. Your LLM. Your TTS. Your STT (Speech-To-Text). No iamjoi account, no Premium, no credits. Install iamjoi Link, stay signed out, and the whole loop — you speak, she thinks, she answers out loud, her lips move, she takes her top off if you ask — happens on your PC.
Nothing is sent to iamjoi in local mode. Not your chat, not your looks, not your keys. Once the model is downloaded you can pull the network cable and keep going.
Download iamjoi Link (version 4.0.0 or higher) PC app for free, here: https://www.patreon.com/iamjoi/posts/iamjoi-link-for-154276984
What you get for free:
- Your own LLM. Anything OpenAI-compatible. Ollama is the one-click path (Dolphin 3 recommended, ~5 GB, uncensored and stays in character; Llama 3.1 8B and Qwen 2.5 7B are also in the list and download from inside Link). LM Studio and friends work too — just point Link at the endpoint.
- Invented personalities. Press Start Character (in the HUD) on any look and the model writes her a personality on the spot. Don't like it? Edit it in Link (the complementary PC app), or write your own and it sticks.
- Local voice with lip sync. Kokoro runs inside Link. Ten catalog voices, and Change Voice button in the HUD rolls a new one mid-session. Costs nothing.
- Speak to her with your voice. The HUD mic has two modes: local Whisper, which never leaves your PC, or Chrome speech, which is noticeably sharper but sends the audio to Google's speech model (never to iamjoi). Your call, one dropdown.
- Bring your own ElevenLabs key if you want the high quality TTS. Paste it into Link and local mode speaks in ElevenLabs voices, billed to your ElevenLabs account at your rate. The key is stored only on your PC — iamjoi never sees it, never proxies it. Pick a catalog name or paste any voice id from the ElevenLabs voice library.
- Or your own speech server. Any endpoint exposing POST /audio/speech returning WAV. Or turn voice off entirely and read subtitles.
- Clothing that responds to conversation. Ask her to take her top off and she does — top, bottom, undertop, underbottom, accessories. Optionally hand the outfit decisions to a second, sharper model so your main chat model never gets distracted by wardrobe logic.
- A Try it box in Link so you can talk to her before VaM is even running.
What local mode does NOT do (this is Premium):
Start Scene, saved Contexts, scene and pose actions, vision / look stills, more than one AI character, wardrobe search and acquire, and the web Playground.
Sign into Link and HUD chat switches to the hosted AI instead of your local model.
Local mode is restricted to one character, HUD Start Character and HUD chat bar.
LOCAL MODE — SETUP (about ten minutes, most of it downloading)
1. Install iamjoi Link (Windows). Do not sign in — or log out if you already did.
2. Open the Local tab and turn Local mode on.
3. Set your Root VaM Directory (needed for pairing and audio).
4. Run a local LLM. Easiest: install Ollama, leave the endpoint at http://127.0.0.1:11434/v1, and hit Download on Dolphin 3 inside Link (~5 GB).
5. Optional: write a one-line scene description. Personality is invented on Start Character; you can edit it afterwards.
6. In VaM, add JoiHUD as a Session plugin, click ALLOW ALWAYS if asked, select a person and press Start Character (not Start Scene).
7. Open the chat bar (gear icon → Chat Bar) and talk. Change Voice is on the extra HUD page.
Local mode requirements:
- Windows (iamjoi Link is a Windows app).
- A PC that already runs VaM. For the LLM itself, a 7–8B model wants ~8 GB RAM and ~5 GB disk. A GPU makes her answer faster; it is not required.
- This package — the full JoiHUD AI (the HUD title bar should read "JoiHUD AI"). JoiHUD-.cs (minus) has no bridge and cannot do this.
- Internet once, to fetch Link, Ollama and the model. After that, offline is fine unless you are using the chrome STT option.
═══════════════════════════════════════════════
IT WORKS WITH THE SCENES YOU ALREADY HAVE
═══════════════════════════════════════════════
Not scenes authored for iamjoi. The scenes you downloaded years ago, from whichever creators you follow. Add JoiHUD as a session plugin, load a scene, and the characters in it can talk after a couple of clicks.
Scenes with no pre-made Context
Hit Describe, type one sentence about what is going on and who everyone is, and the AI writes the scene for every character in it — including the private details only she knows.
Scenes with a pre-made Context
If a scene does have a Context, setup is one click, and the characters also get scene actions they can choose to take on their own: change position, move to another location, whatever that Context defines.
═══════════════════════════════════════════════
PREMIUM — WHAT THE HOSTED SIDE ADDS
═══════════════════════════════════════════════
No LLM to host, no model to download, no server in your tray, no GPU budget. Way smarter, uncensored LLMs at a slightly higher latency cost.
Link signs into your account and connects VaM to AI running on AWS.
Download iamjoi Link (version 4.0.0 or higher) PC app for free, here: https://www.patreon.com/iamjoi/posts/iamjoi-link-for-154276984
1. Drop the .var into AddonPackages/.
2. Add JoiHUD as a Session plugin (not on a Person atom).
3. Run iamjoi Link and sign in.
4. Open the Playground in any browser — your phone works fine — and press Start Scene in the HUD (it's better if you click Add Plugins right before this).
You manage Contexts from the Playground on your phone or PC, and tweak everything else from the HUD in-game.
- Personality from a look. Point it at any VaM appearance and iamjoi writes the character from what she looks like. Save a personality and it comes back on that look every time; it gets stored in your root VaM directory, in a new special folder (iamjoiLinkData).
- Voice. Spoken dialogue with lip sync, plus in-scene subtitles.
- STT. Speak to the characters in your own voice, in your preferred language.
- Saved Contexts. Build a scene once — the setting, each character's role, the actions they're allowed to take — and reload it exactly as you left it.
- Scene awareness, both directions. She reacts to what happens in VaM through the trigger system, and what she says can fire animations, expressions, clothing changes and scene events back.
- Up to three AI characters in one scene.
- Describe. One sentence in, a full scene brief out: shared context for everyone, plus private notes each character keeps to herself.
- Add Plugins button. Adds a configurable list of plugins to every male and female character in the scene. Two come pre-filled — Embody by AcidBubbles and BodyLanguage by CheesyFX (in my opinion the two best plugins in all of VaM) — and you can swap them for whatever you prefer in the plugin's Custom UI. It only attaches plugins already sitting in your AddonPackages; it does not download anything for you.
BEING STRAIGHT WITH YOU ABOUT THE TRADE-OFFS
Local mode (free)
- Nothing reaches iamjoi. No chat, no look images, no API keys.
- Two exceptions, both your own choice and both clearly labelled in Link: if you pick Chrome speech for the mic, your spoken audio goes to Google's speech
model; if you paste an ElevenLabs key, the reply text goes to ElevenLabs and is billed to you. Pick local Whisper and Kokoro and neither happens.
- Quality is your hardware's quality. A 7B model on a laptop is not Grok. Chat history lives in memory and dies when you close Link. Scene text and personality are saved to this PC.
Premium (hosted)
- It is a hosted AI service. Virt-A-Mate never opens a connection to iamjoi servers: JoiHUD AI talks only to iamjoi Link on your PC. Most chat is typed on the iamjoi website and goes from your browser to our servers over HTTPS. Link relays VaM-side data to the same servers — look stills for personality generation, scene and session info, clothing-search hits if you use that, and HUD chat if you type in-scene. We can read that session data while it exists. Signing in is the control you have; there is no per-field filter of what is sent.
- Replies use the cloud LLM you pick in the Playground (OpenRouter, Nano or xAI, depending on the model). Some helper steps — Describe expand, deciding who
speaks next — currently use xAI even if your chat model is not Grok.
- Voice: Kokoro can stay on your PC via Link and cost no credits. Choose cloud TTS and the reply text is sent to ElevenLabs or xAI.
- Link can write a small set of files under your VaM folder at iamjoiLinkData/ (saved personalities and appearance notes), when the Playground asks it to. Not arbitrary paths on your disk.
- No long-term memory yet. She follows the conversation within a session. Saved personalities and Contexts come back; chat history does not carry across sessions yet.
- Light on configuration. Few knobs, sensible defaults.
Full detail: https://iamjoi.ai/redTape/privacyPolicy
WHAT YOU NEED
Local mode: Windows, VaM, iamjoi Link, a local LLM. That's it. Free.
Premium: an internet connection and an iamjoi Premium membership, which comes with 500 AI credits. Extra credits are $1 = 100 credits. Credits are usage-based (LLM, plus TTS if you use cloud voice — local voice is free either way).
Get access on Patreon → https://www.patreon.com/iamjoi/membership
This is a similar but different plugin to JoiHUD-.cs (aka JoiHUD minus), which is fully free and has no external-service option at all; you can find it here: https://hub.virtamate.com/resources/joihud-the-ultimate-scene-control-hud.67247/
Looks, scenes and other third-party content shown in the videos and screenshots are for demonstration and are not part of this package.
CREDITS (CODE IN THIS PACKAGE)
JoiHUD includes code adapted from "Utils" by MacGruber (CC BY 4.0). Specifically the custom-trigger UI: SimpleTriggerHandler and EventTrigger, derived from MacGruber's CustomTrigger / EventTrigger. The MacGruber_Utils.cs file is not included in this .var.
Embody by AcidBubbles and BodyLanguage by CheesyFX are optional Hub plugins the Add Plugins button can attach. They are not bundled with this package.
HOW THE HUD LOOKS
HOW IAMJOI LINK LOOKS
Please click "Allow Always" for the HUD to work properly
A review from a user
Questions, bugs and feature requests are welcome in the discussion thread!
