Cloud AI is convenient, but every request leaves your home. Ollama local AI takes a different path: it runs open models on your own computer, so prompts and answers never go to an outside company. This guide explains what Ollama is, what hardware you realistically need, and how to connect it to Cinematic Recorder on your phone.
What is Ollama?
Ollama is a free program for Windows, macOS and Linux that downloads and runs AI models locally. You install it, pull a model with one command, and it serves that model on your computer. Other apps, including apps on your phone, can then send requests to it over your home network.
The models are "open-weight" models, which means their files can be downloaded and run by anyone. They range from small ones that run on an ordinary laptop to large ones that need a powerful graphics card.
Why choose local AI?
- Privacy. Your prompts stay on hardware you control. This is the main reason most people choose it.
- No per-request bill. After setup, you only pay for electricity.
- No account or card needed. Useful if international payments are hard for you.
- Works without internet once the model is downloaded, as long as your phone and computer share a local network.
The honest downsides
- Local models are usually smaller than top cloud models, so answers can be less polished, especially in Bangla.
- Speed depends on your computer. A slow machine means slow replies.
- Your computer must be on and reachable whenever you want AI on your phone.
- Setup takes a little more effort than pasting a key.
What hardware do you need?
There's no single answer, because it depends on the model size. As a rough guide:
| Computer | What to expect |
|---|---|
| Laptop with 8 GB RAM, no graphics card | Small models (a few billion parameters) work for short text tasks, slowly |
| 16 GB RAM, or an Apple Silicon Mac | Medium models run comfortably for titles, summaries and drafts |
| Desktop with a recent graphics card with plenty of video memory | Larger models and much faster replies |
Free disk space matters too: each model can take several gigabytes. Start with a small model, see how it feels, and move up only if you need to. Our guide on choosing an AI model explains why bigger isn't always better.
Setting up Ollama on your computer
- Download Ollama from the official website and install it.
- Open a terminal (Command Prompt on Windows, Terminal on Mac) and pull a model, for example with
ollama pullfollowed by the model name listed on Ollama's model library. - Test it with
ollama runand the model name. Type a question; if it answers, the model works. - By default Ollama only listens to the computer itself. To let your phone reach it, set the
OLLAMA_HOSTenvironment variable to0.0.0.0and restart Ollama. Ollama's documentation explains this for each operating system. - Allow Ollama through your computer's firewall for your private (home) network only.
- Find your computer's local IP address (it usually looks like 192.168.x.x). Ollama normally uses port 11434.
Connecting Ollama to Cinematic Recorder (once it is supported)
- Connect your phone to the same Wi-Fi as the computer.
- Open Cinematic Recorder and go to the AI or "bring your own key" settings.
- Choose Ollama as the provider.
- Enter your computer's address, for example
http://192.168.1.20:11434(use your own IP). - Choose the model you pulled, then save.
If it doesn't connect, check that both devices are on the same network, that Ollama is running, and that the firewall allows it. Some routers isolate devices on guest networks, so use your main Wi-Fi.
Read more about provider options on the features page. Keep in mind that captions from speech currently run on our server with credits, or with your own OpenAI or Gemini key. AI editing in plain words and AI translation are still coming soon; setting up Ollama now means they can use your local model when they arrive.
When local AI makes the most sense
Work content you can't share
If you record internal tools, client dashboards or school records, local AI keeps that text on your own machine. Pair it with careful recording habits from our guide to privacy before you record.
Lots of small, repeated jobs
If you generate many titles or summaries each week, local AI avoids a growing bill. It's one of the best ways to keep AI costs low.
Learning and experimenting
Trying different open models is a good way to understand what AI can and can't do, without worrying about cost.
FAQ
Can Ollama run directly on my Android phone?
Cinematic Recorder will connect to Ollama running on a computer on your network. Running large models on a phone itself is slow and drains the battery, so a computer is the practical choice.
Is Ollama completely private?
The model runs on your computer, so requests don't go to an AI company. Your privacy then depends on keeping that computer and network secure.
Is Ollama free?
Ollama and many open models are free to download. Always check each model's licence if you plan to use it for business.
Why are answers slow?
The model may be too big for your hardware. Try a smaller model, close heavy programs, or use a computer with a graphics card.
Local AI with Ollama trades a bit of convenience for real control. If privacy matters to you, it's well worth an afternoon of setup — and the editor keeps working fully even when your computer is off.
Related features