Docs

Local AI Chat docs for offline AI, model imports, images, and voice.

Use this guide to set up Local AI Chat on iPhone or iPad, download local models, import compatible GGUF files, keep work private, ask questions about images, and listen with text-to-speech.

Updated May 23, 2026Setup guideiPhone and iPad
Local AI Chat running a private offline AI workflow on iPhone.

Quick answer: Install Local AI Chat, download a supported model once, then use supported local AI features without an account, cloud API key, or constant internet connection. For custom models, import a compatible GGUF file from a trusted direct download URL.

Start here

New to offline AIStart with a built-in model, send a short prompt, then test airplane mode after the model is available on device.
Using your own modelChoose a compatible GGUF file that fits your device, copy its direct download URL, and import it in the app.
Privacy-sensitive workUse local models for notes, drafts, screenshots, study material, and other prompts you do not want to send to a cloud chatbot.

Quick setup

  1. Install Local AI Chat. Download the app from the App Store on a supported iPhone or iPad.
  2. Choose a local model. Start with a smaller built-in model if you want faster first-run testing.
  3. Download the model on Wi-Fi. Model files can be large, so keep the app open until the download finishes.
  4. Send a practical prompt. Try summarizing a note, rewriting a message, explaining a screenshot, or drafting a reply.
  5. Test offline behavior. Turn on airplane mode after setup and confirm that the supported local model still responds.

Use offline AI chat

Offline chat depends on having the model file available on your device. Once a supported local model is installed, Local AI Chat can generate responses on device for supported workflows without Wi-Fi or cellular data.

For best results, keep one smaller model installed for fast everyday work and a larger model only when you need better reasoning or writing quality. Larger models can improve output quality, but they also use more storage, memory, and battery.

Import compatible GGUF models

Local AI Chat supports importing compatible GGUF models by URL. GGUF is a common local LLM file format used by many open model communities, but not every file is a good fit for every mobile device.

  1. Pick a trusted source. Use reputable model publishers and read the model card before downloading.
  2. Choose a mobile-friendly file. Smaller quantized models are usually better for phones than very large desktop-oriented files.
  3. Copy the direct file URL. The import flow needs a downloadable model file link, not a general project page.
  4. Paste the URL in Local AI Chat. Start the import and wait for the file to finish downloading.
  5. Test with your real task. Compare speed, answer quality, and battery impact before making it your default model.

Model import checklist: compatible GGUF file, direct HTTPS download link, enough free device storage, trusted source, and a model size your device can run comfortably.

Model downloads and storage

Local models are real files stored on the device. If a download fails, check available storage, network stability, and whether the source URL points directly to the model file. If the app feels slow, try a smaller model before assuming the device is unsupported.

Delete models you no longer use. Keeping several large files can quickly consume device storage, especially when testing Llama, Gemma, Mistral, Phi, Qwen, SmolLM, or other model families.

Privacy basics

The App Store privacy label for Local AI Chat states that data is not collected. Supported local AI features are designed so prompts can be processed on device instead of being sent to a cloud AI provider.

Model downloads are different from chat prompts. If you import a model from an external URL, your device still has to contact that host to download the file. Use trusted sources and avoid private or suspicious model links.

Ask questions about images

Local AI Chat supports image understanding for photos, screenshots, documents, handwritten notes, charts, and diagrams. Good image prompts are specific: ask what you want extracted, checked, summarized, or explained.

ScreenshotAsk the model to summarize settings, explain an error, or extract visible text.
Document photoAsk for a short summary, action items, or a clearer rewrite of visible notes.
Chart or diagramAsk what the chart shows, what changed, and what conclusion is reasonable.

Listen with text-to-speech

Text-to-speech helps when you want to listen to AI responses while studying, walking, traveling, or reviewing drafts hands-free. Generate a response, then use the app's speech controls to hear it aloud.

Available voices and languages can depend on iOS voice settings and installed system voices. If speech is silent, check device volume, silent mode, audio output, and whether the selected voice is available on the device.

Troubleshooting

Model will not importConfirm the link is a direct GGUF file URL, the file is compatible, and the device has enough free storage.
Responses are slowTry a smaller quantized model, close other heavy apps, or use a built-in model for quick work.
Offline test failsMake sure the model finished downloading before enabling airplane mode.
Image answer is weakUse a sharper image and ask a more specific question about the visible content.
Voice is not playingCheck volume, silent mode, Bluetooth output, and installed iOS voices.

Related pages

Offline AI chatHow private local chat works without constant internet. Local LLM for iPhoneModel formats, imports, and on-device inference basics. Private AI chatWhy local processing changes the privacy model. Image visionUse photos, screenshots, and documents in local AI workflows. Offline text-to-speechListen to generated AI responses on iPhone and iPad. FAQShort answers about devices, models, privacy, and app support.