Why you need your own API key, how to get one and how to enter it in Qala.
Qala sends translations directly from your laptop to the AI — not via our server. That gives you three wins: privacy (your content only goes to the provider you chose), cost control (pay-as-you-go, we don't earn anything) and independence (if our server is down, your service still runs).
We strongly recommend the cloud API option — especially Google Gemini:
Local models only make sense if the laptop is very powerful (e.g. M3/M4 Pro/Max) or no internet is available in the room.
Gemini is Google's AI model. Very cheap, often completely free at moderate usage, with excellent translation quality.
Open aistudio.google.com/apikey and sign in with a Google account (your normal Gmail account works fine).
Click the blue button. You can either create a new project or pick an existing one.
You'll get a long code like "AIzaSyD...". Click the copy icon. Important: this key is like a password — never share it.
Open Qala → Admin → Settings → AI provider: Gemini → paste the API key → Save. Done.
Claude from Anthropic delivers outstanding translation quality, especially for nuanced preaching.
Open console.anthropic.com/settings/keys and create an account.
Claude has no free tier — top up $5 or $10 of credit. That usually lasts for many months.
Click "Create Key", give it a name (e. g. "Qala"), copy the key.
Qala → Admin → Settings → AI provider: Claude → paste the API key → Save.
OpenAI offers GPT-4o and GPT-4o-mini. Mature, broad language coverage, good quality.
Open platform.openai.com/api-keys and create an account.
Minimum top-up is $5. That's enough for many months.
Click "Create new secret key", give it a name, copy the key.
Qala → Admin → Settings → AI provider: OpenAI → paste the API key → Save.
Ollama is a free open-source tool that runs AI models directly on your computer. Fully offline, zero cost, maximum privacy.
Open ollama.com/download and download the installer for your operating system.
In your terminal, run ollama pull gemma3:4b — this downloads a translation model (~3 GB). Alternative: ollama pull qwen2.5:7b for higher quality.
Qala → Admin → Settings → Mode: Ollama → pick model → Save. No API key needed.
From now on Qala needs no internet. Perfect for camps, open-air services or churches with strict privacy requirements.
A transparent overview — we earn nothing from API usage, you pay the provider directly.
| Model | 2-hour service | 52 services / year |
|---|---|---|
| Gemini 2.0 Flash | ~ $0.02 | ~ $1 |
| GPT-4o mini | ~ $0.03 | ~ $1.50 |
| Claude Haiku | ~ $0.20 | ~ $10 |
| Claude Sonnet | ~ $0.60 | ~ $30 |
| Ollama (local) | $0.00 | $0.00 |
Prices are estimates based on current API rates (April 2026) and may vary. Many providers offer a monthly free tier that's already sufficient for smaller churches.
Open Qala and paste it into the setup wizard or under Admin → Settings.
Back to install guide →