HomeRig Help & Privacy

What HomeRig is

HomeRig doesn't create anything by itself. It is a remote control for programs you install on your own computer (the "rig"). The heavy work — running AI models — happens on that computer; the app only sends your requests and shows the results.

▶ Watch the how-to video — setting up, connecting and using HomeRig, step by step.

No computer yet? Turn on Demo mode to try every tab with sample results made on your iPhone or iPad.

What you need

• A computer (this guide shows Windows) that stays on, awake and online while you use the app. A graphics card from NVIDIA makes image, video and voice work much faster.
• The program(s) you want to control: ComfyUI, LM Studio and/or Applio. You only need the ones you use — each tab works on its own.
• Tailscale (free for personal use) on the computer and on your iPhone or iPad — or, at home, just the same Wi-Fi for both (details).
• iPhone or iPad with a recent version of iOS / iPadOS.

Quick start (10 minutes if the programs are already installed)

1. Install Tailscale on the computer and on your iPhone/iPad and sign in to the same account. Details
2. Start your program (ComfyUI, LM Studio server and/or Applio) and check it works on the computer. Details
3. Publish it to your tailnet with one command per program. Details
4. Enter the address in the app (⚙︎ Settings in each tab) and tap Test Connection. Details

The Host is the same for all three — your computer's Tailscale name, for example desktop-abcd.tail1234.ts.net (without https://).

Set up Tailscale

Tailscale creates a private, encrypted network between your own devices, so the app can reach your computer from anywhere without opening ports on your router or exposing anything to the public internet.

1. Install it on the computer. Download from tailscale.com/download, run the installer and sign in (Google, Microsoft, Apple or GitHub). Remember which account you used.
2. Install it on your iPhone/iPad. App Store → search Tailscale → install → open → sign in with the same account → switch it on until it says Connected.
3. Turn on names and HTTPS. In the DNS page of the admin console make sure MagicDNS is on, then click Enable HTTPS. This gives your computer a name like desktop-abcd.tail1234.ts.net.
4. Check. On the computer open Command Prompt (Windows key, type cmd, Enter) and run:

   tailscale status

   You should see both the computer and your iPhone/iPad in the list.

Install the programs on the computer

Install only what you want to use. Test each one on the computer first; only then connect the app. Steps can differ slightly between versions — if something doesn't match, follow the program's own instructions. HomeRig only needs each program to be running.

ComfyUI — images and videos

1. Download ComfyUI (Desktop installer or portable version) from comfy.org/download and run the installer. Choose your graphics card (NVIDIA if you have one). The first install downloads a lot.
2. Open ComfyUI, load one of the built-in templates, download the model it asks for and press Run. If a picture appears, it works.
3. Note the port: open ComfyUI in a browser on the computer and look at the address bar — 127.0.0.1: followed by a number (often 8000 for the Desktop version, 8188 for the portable one). You need that number in the next step.
4. Save workflows in the right folder and format for the app — see Workflow files.

LM Studio — chat

1. Download and install LM Studio from lmstudio.ai.
2. Open the Discover tab, search for a small model (for example a Gemma or Qwen model that fits your graphics card memory) and download it.
3. Open Chat, load the model and say hello to make sure it works.
4. Open the Developer tab and switch the server on (it should say Running, default port 1234). In Server Settings turn on Just-in-Time model loading so the app can start models by itself.

Applio — voice conversion and text-to-speech

1. Download the latest Windows package from the Applio project page (or follow the install instructions there).
2. Unzip it into a short folder path such as C:\Applio.
3. Double-click run-install.bat and wait until it finishes.
4. Double-click run-applio.bat. A black window stays open — leave it open — and Applio opens in your browser at http://127.0.0.1:6969.
5. In Applio use the Download tab to get a voice model (or train your own). It should then appear on the Inference tab.
6. Try a conversion in Applio itself. If it works there, it will work from the app.

Open the doors with tailscale serve

These commands let your iPhone/iPad reach each program safely through Tailscale. Run only the ones you use, in Command Prompt on the computer, after the program is running. You do this once; it keeps working after restarts. Tailscale can offer HTTPS only on ports 443, 8443 and 10000, which is why each program gets a different one.

tailscale serve --bg 8188
tailscale serve --bg --https=8443 1234
tailscale serve --bg --https=10000 6969

• The first line is ComfyUI (replace 8188 with your ComfyUI port if it's different).
• The second is LM Studio. The third is Applio.

To see the address that was created, and to check what is published:

tailscale serve status

It prints your computer's web address, for example https://desktop-abcd.tail1234.ts.net. The part after https:// and before any : is your Host.

To stop publishing one of them, add off — for example tailscale serve --https=8443 off.

Alternative: connect to the Tailscale IP (no HTTPS)

Run tailscale ip -4 to get an address like 100.101.102.103. In the app enter that as the Host, the program's own port (8188, 1234 or 6969), and leave Use SSL on Auto (it uses plain HTTP on those ports, which is fine inside Tailscale's encrypted network). For this to work the program must accept outside connections: LM Studio needs Serve on Local Network; ComfyUI needs to be started with --listen 0.0.0.0; Applio needs Gradio's server name set to 0.0.0.0 (for example set GRADIO_SERVER_NAME=0.0.0.0 before run-applio.bat). The tailscale serve method above needs none of that.

Connect on the same Wi-Fi (no Tailscale)

When the computer that runs ComfyUI, LM Studio or Applio and your iPhone/iPad are connected to the same home router, HomeRig can talk to the computer directly using its local address (for example 192.168.1.20). You don't need a Tailscale account, and the traffic never leaves your own network.

What you'll do

1. Find the computer's local IP address (step A).
2. Let each program accept connections from the network — by default they only listen to the computer itself (step B).
3. Allow the programs through the firewall (step C).
4. Check it from the phone's browser, then enter the address in HomeRig (steps D and E).

A · Find the computer's local IP address

A home address normally starts with 192.168., 10. or 172.16.–172.31.. If you see 169.254.… the computer isn't connected to the router properly.

B · Let each program accept connections from the network

Out of the box, ComfyUI, LM Studio and Applio only answer the computer they run on (127.0.0.1). Change this once for each program you use:

ComfyUI (port 8188, Desktop app: 8000)

• Windows portable version: right-click the .bat file you start ComfyUI with (usually run_nvidia_gpu.bat) → Edit (or open it in Notepad). At the end of the line that contains main.py, add --listen 0.0.0.0, save, and start ComfyUI again with that file. The line then looks like:

  .\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build --listen 0.0.0.0

• ComfyUI Desktop app: open its Settings, go to the server settings (Server-Config), set the listen address (Host) to 0.0.0.0, then fully quit and reopen the app. The Desktop app normally uses port 8000 — check the address bar of its window (127.0.0.1:8000). The exact menu names can differ a little between versions.
• Manual install (Windows / macOS / Linux): start it with

  python main.py --listen 0.0.0.0 --port 8188

When it works, the ComfyUI console shows To see the GUI go to: http://0.0.0.0:8188 instead of 127.0.0.1.

LM Studio (port 1234)

1. Load a model (or turn on Just-in-Time model loading).
2. Open the Developer tab (the >_ icon on the left) and switch the server Status to Running.
3. Open the server Settings and turn on Serve on Local Network.
4. The page now shows Reachable at: http://192.168.x.x:1234 — that address is the one to use. If it still says 127.0.0.1, the switch isn't on.

Applio (port 6969)

Applio is built on Gradio, which listens on the network when GRADIO_SERVER_NAME is set to 0.0.0.0:

• Windows, each time — open PowerShell in the Applio folder (Shift + right-click inside the folder → Open PowerShell window here) and run:

  $env:GRADIO_SERVER_NAME="0.0.0.0"
  .\run-applio.bat

  In Command Prompt instead: set GRADIO_SERVER_NAME=0.0.0.0 then run-applio.bat.
• Windows, permanently — run this once in Command Prompt, close the window, and from then on just double-click run-applio.bat as usual:

  setx GRADIO_SERVER_NAME 0.0.0.0

• macOS / Linux — in a terminal in the Applio folder:

  GRADIO_SERVER_NAME=0.0.0.0 ./run-applio.sh

Applio's window then shows Running on local URL: http://0.0.0.0:6969.

C · Allow the programs through the firewall

Windows

1. Mark your Wi-Fi as a private network: Settings → Network & internet → Wi-Fi → click your network (or Ethernet for a cable) → Network profile type: Private network. On a "Public" network Windows blocks everything below.
2. The first time each program starts listening, Windows shows "Windows Defender Firewall has blocked some features…" — tick Private networks and click Allow access. (ComfyUI and Applio appear as Python.)
3. Missed that window? Open Windows Security → Firewall & network protection → Allow an app through firewall → Change settings, find Python / LM Studio and tick Private. Or open PowerShell as administrator and allow the ports directly:

   New-NetFirewallRule -DisplayName "HomeRig" -Direction Inbound -Protocol TCP -LocalPort 8188,8000,1234,6969 -Action Allow -Profile Private

macOS

If System Settings → Network → Firewall is on, macOS asks whether to accept incoming connections the first time — click Allow. You can also add Python / LM Studio under Firewall → Options.

Linux

If you use ufw: sudo ufw allow from 192.168.0.0/16 to any port 8188,1234,6969 proto tcp (adjust to your network).

D · Check it from the phone's browser

Before opening HomeRig, open Safari on the iPhone/iPad (on the same Wi-Fi) and type the address — using your computer's IP instead of 192.168.1.20:

If Safari can open it, HomeRig will work too. If it keeps loading or says it can't connect, the problem is on the computer (steps B and C) or the network — see If it doesn't connect. On Windows you can also check with netstat -an | findstr 8188: 0.0.0.0:8188 … LISTENING is right; 127.0.0.1:8188 means step B isn't done yet.

E · Enter it in HomeRig

1. In the tab you want (ComfyUI, LM Studio or Applio) tap the ⚙︎ gear.
2. Tap Find on this Wi-Fi. The first time, iOS asks "HomeRig would like to find and connect to devices on your local network" — tap Allow. If nothing is listed on the first try, tap Find on this Wi-Fi again.
3. Tap the computer in the list — Host, Port and Use SSL are filled in for you. Then tap Test Connection.
4. Or type it yourself:

   Field	ComfyUI	LM Studio	Applio
   Host	The computer's IP, e.g. 192.168.1.20 (no http://)
   Port	8188 (Desktop 8000)	1234	6969
   Use SSL	Auto — it uses plain HTTP on these ports, which iOS allows inside your own network

5. In the LM Studio and Applio tabs, Use the ComfyUI host (same PC) copies the IP from the ComfyUI tab so you only change the port.

If it doesn't connect

Demo mode

Demo mode lets you try every tab without a computer. Tap Try the Demo on any tab (or switch on Demo mode in Settings). The app then talks to a pretend ComfyUI, LM Studio and Applio built into the app itself: two sample workflows that draw a picture or a short video of your prompt, a chat that answers with a canned reply (and can hand a “draw…” request to the sample workflow), and voice conversion that returns your recording, with text-to-speech read by the iOS voice.

Nothing is sent anywhere in Demo mode — all sample results are made on your device. A banner at the top shows while it's on; tap Exit (or turn it off in Settings) and your own connection settings come back.

Connect the app

In each tab tap the ⚙︎ gear icon and fill in the connection. The Host is the same every time; only the Port changes.

Then tap Test Connection. A green check means you're connected (it also tells you how many workflows, models or voices it found).

ComfyUI tab — images and videos

Run a workflow

1. The first screen lists the workflows found on your computer. Pull down to refresh; use the search bar to filter. Tap the ☆ star on a workflow (or long-press it → Add to Favorites) to keep it at the top of the list under Favorites.
2. Tap a workflow. The app turns it into a form: the prompt fields, input image / video / audio pickers, numbers (seed, steps, size…), dropdowns, and an Other section for anything it doesn't recognise. The seed field has a dice button that picks a new random seed.
3. Fill in what you want and tap Generate. A progress bar and live status show what your computer is doing. You can leave the app; a notification tells you when it's done.
4. Tap a result to view it full screen. From there you can save to Photos or share/export it.

Input images, videos and audio

• Images: pick from your photo library, or choose an earlier result from the Gallery.
• Videos: pick from the photo library.
• Audio: record with the microphone or choose a file from the Files app.

Files you pick are uploaded to your own computer (ComfyUI's input folder) when you tap Generate.

Toolbar

• Top left — Gallery: every result is saved automatically on the device. Select several to delete; long-press to delete one.
• ⋯ menu: Select Workflows to hide some from the main list, and Hidden Workflows to use or unhide them (swipe left or long-press a row to unhide).
• Quote icon — Prompt Library: save prompts you use often, copy them, or import/export them as a file.
• ⚙︎ Settings and ⓘ Setup Guide (this guide, inside the app).

If ComfyUI rejects a value

When a dropdown value doesn't exist on your computer (for example a model file name), the app turns that field into a picker with the valid options. Choose one and tap Generate again.

Where to save workflow files

The app reads workflows straight from your ComfyUI server — nothing is uploaded from the iPad. They must be saved on the computer that runs ComfyUI, inside ComfyUI's user folder:

ComfyUI/
└── user/
    └── default/
        ├── workflows_api/  ← API workflows (required)
        └── workflows/      ← normal workflows (optional)

Portable Windows build: ComfyUI_windows_portable\ComfyUI\user\default\. ComfyUI Desktop: the folder you chose at install (often Documents/ComfyUI), then user/default/. If ComfyUI runs with --multi-user, use user/<your name>/ instead of user/default/.

1 · API workflow (required)

1. Build and test the workflow in ComfyUI first.
2. Export it in API format: Workflow → Export (API) (older versions: turn on Dev mode options in ComfyUI settings, then use Save (API Format)). A normal Save makes a different format the app can't run.
3. Copy the .json file into ComfyUI/user/default/workflows_api/ (create the folder if needed). Only .json files directly inside that folder are listed, not sub-folders. The file name becomes the workflow name in the app.
4. Refresh the list in the app (pull down).

2 · Normal (UI) workflow (optional)

Some custom nodes (for example KJNodes WidgetToString) need the full graph that an API export doesn't contain. For those, also save the normal version with Ctrl+S in ComfyUI using exactly the same file name. It is stored in workflows/ and the app sends it along automatically when you run the workflow.

Different folders

Folder names are relative to user/default/. If you keep your files elsewhere, change Workflow API folder / Workflow UI folder in the ComfyUI tab's Settings.

LM Studio tab — chat

1. Connect: ⚙︎ → enter Host and Port (8443 with the serve command above) → Test Connection.
2. Choose a model with the chip icon at the top right. The list comes from LM Studio; if you turned on Just-in-Time loading, choosing a model that isn't loaded yet loads it on first use (the first reply can take longer).
3. Chat. Replies stream in as they are written. Models that "think" show a collapsible Reasoning section. Tap stop (■) to interrupt a reply.

• Top left — History: conversations are saved on the device. Swipe to delete.
• New chat (pencil icon) starts a fresh conversation.
• Long-press a message: copy it, Save to Prompt Library (so you can reuse a prompt in the ComfyUI tab), or turn it into an image/video.
• ⚙︎ Settings: connection, optional token, and a system prompt sent with every chat (for example "You write detailed prompts for image and video models").

Create images and videos from the chat

Language models only write text. So the chat borrows your ComfyUI workflows to make pictures and videos, and shows the result right in the conversation. ComfyUI must be set up in the ComfyUI tab first.

1 · Choose the workflows

Tap the + button next to the message field → Choose workflows… and switch on the workflows the chat may use. Workflows that run from a text prompt, or from a picture or video you attach (image-to-image, image-to-video), are supported. Workflows that need audio are turned down with a reason.

Attach pictures, videos and documents

Tap + → Photos, Video or Files to attach to your message (up to 6).

• Pictures are sent to the model to look at — the model must support vision (for example a Gemma 3 or Qwen-VL model in LM Studio).
• Videos are shown to the model as 4 still frames, because LM Studio cannot take video directly.
• Text files and PDFs are read and included with your message.
• As workflow input: attach a picture, then choose Create image with an image-to-image workflow (or Create video with an image-to-video one). The picture is uploaded to your computer and used as the workflow's image input; an attached video goes into the video input. The most recent attachment is used, and Create again reuses it.

2 · Create image / Create video

1. Tap + and choose Create image or Create video. A chip appears above the field (with a menu to pick the workflow if you have several of that kind). The one you pick becomes the default for next time; you can also set the default image and video workflow in + → Choose workflows….
2. Type what you want and send. The text goes straight to your ComfyUI workflow — the chat model never has to guess what you meant.
3. Shape and length: the chip also shows a Square / Portrait / Landscape menu, and for videos a Duration menu in seconds. The app reads them from your workflow: the shape changes the width and height of the workflow's size node (a portrait size is simply turned sideways; from square it keeps about the same number of pixels), and the duration is turned into frames using the workflow's own frame rate. These menus only appear when the workflow has those settings. You can also just ask the model for "a vertical image" or "a 5-second video".
4. Keep Let the AI write the prompt on to have the language model turn your words (in any language) into a detailed English prompt first — helpful because most image models understand English best. Turn it off to send your text exactly as typed.
5. A progress bar shows while your computer works, then the picture or video appears. Tap it for full screen; use Create again for a new variation (new random seed).

Other ways

• Under any reply, Create image / video… turns that text into a picture (you can edit the prompt and pick a workflow first).
• Let the model create by itself (in Choose workflows…): models that support tool use may start a creation on their own when you ask them to draw something. Not every model does this well — if yours doesn't, use the + menu.

Everything created is also kept in the ComfyUI tab's Gallery on your device.

Applio tab — voice conversion and text-to-speech

Applio must be running on your computer with at least one voice model. Connect in ⚙︎ (Port 10000 with the serve command above) and tap Test Connection — it reports how many voice models it found.

Convert a voice

1. Give it a voice. Tap Record (microphone) or Choose file (WAV, MP3, M4A, FLAC, AIFF… from the Files app). You can play it back before converting. Applio only accepts WAV, so recordings are saved as WAV (16-bit, mono) and any other format is converted to WAV on your device before it is sent — you don't need to do anything.
2. Pick a voice model and, if you like, an index file (Automatic picks the one that belongs to the model, None uses no index). Refresh models re-reads Applio's list so newly added models appear without restarting Applio.
3. Adjust conversion settings if needed: pitch (semitones), search feature ratio, volume envelope, protect voiceless consonants, pitch-extraction method, output format, split audio, autotune, clean audio. The defaults are a good start; Reset to defaults restores them.
4. Agree to Applio's Terms of Use. Applio requires this before every conversion. The app switches it on for Applio only after you turn on the switch yourself, and links to the terms so you can read them. Only convert voices that are yours or that you have permission to use.
5. Tap Convert voice. The app uploads the recording to your computer, Applio converts it, and the result downloads back to the device. Long clips take longer; you can cancel.

Text to speech

If your Applio has a text-to-speech tab, a Voice / Text to Speech switch appears at the top. Type the text, choose a speaking voice (searchable, for example th-TH or en-US) and a speed, choose your voice model and tap Speak with this voice. Applio first speaks the text with a Microsoft Edge voice and then turns it into your voice model, so the computer needs internet access for this.

Train your own voice model

1. At the bottom of the Applio tab tap Train a new voice model.
2. Type a model name (English letters, numbers, - or _) and add training audio: Record with the microphone or Choose files from Files on your iPhone/iPad (WAV, MP3, M4A, FLAC…). Use 10–30 minutes of clean speech from one voice — no music, echo or background noise — and at least 1 minute.
3. Pick the settings: sampling rate (40000 is a good default), epochs (100–300 is typical; more takes longer), batch size (4 for 6–8 GB of GPU memory, 8 for 12 GB+), save every N epochs and the vocoder. Agree to Applio's Terms of Use and tap Start training.
4. HomeRig uploads the clips into a dataset on the computer (assets/datasets/<name>), then Applio preprocesses them, extracts features and trains. Keep HomeRig open until training starts (usually a few minutes). Training itself takes minutes to hours — you can close HomeRig; the computer keeps training, and HomeRig reads the progress from the checkpoints Applio saves every N epochs.
5. When it's finished, tap Use this model to select it in Voice mode. Stop ends training early.

Voice Library

Every result is saved on the device (tray icon, top left). Tap Play or Export (Save to Files, AirDrop, Messages…). Press and hold a row for more: play or export the original recording, convert the original again, copy the text, or delete. Swipe to delete.

App settings and what is stored on your device

App lock

ComfyUI tab → Settings → Lock app requires Face ID, Touch ID or your device passcode every time the app opens, and hides the content in the app switcher. HomeRig never receives your biometric data — iOS only tells the app whether you passed.

Notifications

When a generation finishes while you're in another app, HomeRig shows a local notification (created on your device, not sent from any server). You can turn them off in iOS Settings.

Backup and moving to a new device

Saved results are part of the app's data and follow your device backup. To keep something permanently, use Export / Save to Photos before deleting the app.

Troubleshooting

Still stuck? Contact support and include the exact message you see and which tab it happens in.

Frequently asked questions

Privacy Policy

Effective date: September 27, 2026  ·  App: HomeRig  ·  Contact: prowitinw2@gmail.com

This policy explains what HomeRig ("the app") does with information. HomeRig is a remote control that connects your iPhone or iPad to programs (such as ComfyUI, LM Studio and Applio) that you run on your own computer. The developer operates no servers for the app.

What data the app handles

The developer does not receive any of this. The app does not include analytics, crash-reporting, advertising or other third-party SDKs, does not track you across apps or websites, and does not create or use an account. Because nothing is collected by the developer, HomeRig's App Store privacy label is "Data Not Collected". In Demo mode nothing leaves the device at all.

Permissions the app may ask for

Services you choose to connect

You decide what the app connects to. Those services have their own privacy practices, which are outside the developer's control:

• Your own computer and its programs (ComfyUI, LM Studio, Applio and any custom nodes or models you install). What they do with your content is governed by how you set them up. For example, Applio's text-to-speech feature contacts Microsoft's online speech service, and ComfyUI's optional paid "API nodes" send prompts to the company behind that node — both from your computer, only if you set them up there. HomeRig itself never contacts them.
• Tailscale (or another VPN/tunnel) carries the encrypted connection between your devices. See Tailscale's own privacy policy.
• Web links. Some links in the in-app guide (for example to download a program or to read Applio's Terms of Use) open in your browser. The app itself does not send any data to those sites.

HomeRig connects to these services only using addresses you type in. Traffic to an IP address, a .local name or a Tailscale .ts.net name may use plain HTTP, because it stays inside your private network (Tailscale traffic is encrypted end to end); any other domain name must use HTTPS.

Security

• Tokens and secrets are stored in the iOS Keychain; everything else is kept in the app's private storage and protected by iOS.
• You can add an app lock (Face ID / Touch ID / passcode) in Settings.
• Protect your own computer and accounts: only publish services to your private tailnet, and don't expose these programs to the public internet.

Retention

Data stays on your device until you delete it (inside the app) or delete the app. iOS may keep Keychain items (your saved tokens) after an app is deleted, so clear the token fields in Settings first if you want them gone. Temporary files created while uploading or downloading are kept in the system's temporary storage, which iOS clears automatically.

Children

HomeRig is not directed to children under 13 (or under the applicable age in your country) and does not knowingly collect information from anyone.

Your choices and rights

Because the developer holds no data about you, there is nothing to request, correct or delete on the developer's side. You control everything on your device: delete results, chats, voices and prompts in the app, revoke permissions in iOS Settings → HomeRig, or delete the app. If you have privacy questions or believe your data has been handled incorrectly, write to prowitinw2@gmail.com. Depending on where you live (for example under the EU/UK GDPR or Thailand's PDPA) you may have additional rights, which you can exercise through that address.

Changes to this policy

If the app's data practices change, this page will be updated and the effective date above changed. Significant changes will also be noted in the app's release notes.

Contact

Questions about this policy: prowitinw2@gmail.com.

Contact & support

Questions, bug reports and suggestions are welcome: prowitinw2@gmail.com

To help me help you quickly, please include:

• Your device and iOS version, and the HomeRig version (shown on its App Store page).
• Which tab (ComfyUI, LM Studio or Applio) and what you were doing.
• The exact message shown, or a screenshot. Do not send tokens, secrets or private files.
• How you connect (Tailscale serve or Tailscale IP) and the versions of the program involved.