← All guides
Getting Started

Setting up AI

2026-10-27 · 4 min read
On this page

    What's it for

    Well, one translator powers everything that translates, such as: Live Audio, OCR Capture's translate mode, Quick Translate, Song Lyrics, Dictionary and the browser extension just to name most things. You can just set it up once, and there's generally insight in how to set them up (unless it's local because you should have your own set up, you just have all the parameters handed to you that you'd need)

    Where to find it

    • Open the pyontan app Settings > Providers. Whatever is currently being used is marked as "In Use".
    • For pyontan Hop, it defaults to the same as the app but if you want to use a different provider, there's an engine option exclusively for this in Settings > Web Extension

    Before you start

    • pyontan AI is automatically set up and should default to being "In Use"
    • Otherwise if you want to use pyontan AI, from the tray menu or main menu > Settings > Providers, select pyontan AI and click "Apply" / "Use In Translation"
    • Save and Close

    Note: Daily allowance for Free-tier users is 50 so this may run out very quickly, for Premium it's 2,500 and for Super Premium it's 6,000. Also pyontan is AI is tuned to handle and process slang (including chat slang), speech and manga especially

    pyontan AI options

    • Text type: what the on-screen text usually is so the AI can tailor to be more accurate, e.g. whether it's a live chat/stream/game/slang scenario, etc. Or manga/games, general for standard writing, etc.
    • Explain level: For the "Explain grammar" option in Dictionary, it will tailor the explanation according to what level you are in the language

    Using your own AI

    GroupProvidersNeeds
    pyontanpyontan AIAn account
    Translation ServicesDeepL, Google TranslateAPI key
    AI ModelsOpen AI, Claude, Google Gemini, DeepSeek, OpenRouterModel + API key
    Local & CustomOllama, LM Studio, OpenAI-compatible, OfflineServer URL + model or download

    AI models are a slightly more advanced area of setting up AI models and Local & Custom expect you to know how to set them up yourself as they are an advanced level of AI setup.

    Steps for an AI model (OpenAI, Claude, Gemini, DeepSeek, OpenRouter)

    1. Click it on the list, its little description blurb has a "Get a key" link to that company's key page
    2. Get your API key and make sure you have an account in order to get an API key and paste it into "API key" (you can hide/show it with the eye icon)
    3. Press "Refresh" to list the models that key can use, then pick one in "Model" or you can manually type a model name instead
    4. Optional: "Temperature" (lower = more literal and consistent; some models only accept their default) and "Advanced (JSON)" are extra options sent with every request; most people likely don't need to modify this though, only do if you know what you are doing
    5. Press "Test", it will either say "Works" or what went wrong
    6. Click "Use for translation", "Apply" or "Save & Close", all of these will save your option

    Translation Services

    DeepL:

    • Their free plan covers 500,000 characters a month, this is a lot and generally hard to use up
    • Sign up for the DeepL API Free plan, copy the key from the API Keys page and paste it, the "Free plan key" will be automatically ticked if the key is a free key (indicated by ending with :fx)

    Google Translate:

    • Separate free quota from DeepL, it is partially free (500,000 characters free, then it becomes pay as you go per month although it's fairly hard to use up)
    • This requires a Google Cloud project with billing turned on (which requires a card set up), the individual steps are found in the explanation blurb in Settings > Provider and select Google Translate

    Things worth noting

    • The list on the left hand-side marks the one "In use". Hovering a provider shows "Set up" or "Not set up yet"
    • Keys are stored encrypted on your PC
    • Settings > Web Extension > Translation Engine allows selecting a separate provider from the app for the browser extension (Live Chats/VODs such as YouTube and Twitch)
    • Getting Started ticks off the translator when one is actually set up that works, not just when an engine is set
    • pyontan's daily allowance applies only to pyontan AI, your won AI key is billed by that company, not us
    • Google Translate requires billing enabled, even for the free monthly quota

    Common Issues

    These mostly address errors people are likely to see, you'll get insightful error messages but in case people come across them anyway.

    • "<name> isn't set up yet. ... (Settings > Providers)": the engine in use is missing its key or model
    • "<name> rejected this API key.": wrong API key provided or a key copied with a missing character or space (malformed)
    • "Couldn't reach <name>. Check your internet connection": as it says, check your internet connection, or the service could be down and you may need to try again later (if it's pyontan AI then you'd know via our Status page)
    • "Couldn't reach <name>. Check it's running and the server URL is right.": for Ollama/LM Studio/Custom, this means the local server isn't started, or the URL is wrong.
    • "No models found." after Refresh: the key can't use any models yet (some companies need billing set up first), or for a local server no model is downloaded/loaded.
    Related guides
    Ready to try pyontan?
    Free covers the basics. No card needed.
    Download