Convert, Polish, and Explain need a configured AI service. The service setting decides where the request goes, who bills the request, and which models you can select. It does not change UnimeType's local capture, masking, validation, or write-back checks.
Choose a service route
| Route | What you provide | Request path | Billing |
|---|---|---|---|
| UnimeType Balance | A signed-in UnimeType account with available credits | Device → UnimeType gateway → configured model provider | Uses your UnimeType credit balance |
| Your own API key | Provider, Base URL, model, and provider API key | Device → the compatible address you save | Billed separately by that provider; no UnimeType credits are used |
Choose UnimeType Balance when you do not want to configure a model-provider account. Choose Your own API key when you already have a compatible provider and want that provider to handle the request directly.
Both routes remain subject to the data terms of the model provider that receives the request. See Privacy and protected content for the text and local masks included in an action.
Select a model
The managed model catalog can change. The model list currently shown in UnimeType Settings is the source of truth for UnimeType Balance. Saving a different model changes the route used by later AI actions; it does not rewrite text that has already been converted.
Current managed catalog
The current shared client catalog contains:
- Seed 2.0 Mini
- DeepSeek V4 Flash
- GLM 4.7
- Grok 4.20
These names identify selectable managed routes. They are not a published quality ranking or a promise that one route is universally faster or more accurate. The in-app catalog can change between documentation releases.
Start with the model already selected by the app, then test the same short, non-sensitive examples for the input languages you use. Keep a model when the target script, preserved English, protected tokens, and response time meet your actual workflow. Switch or stop when representative inputs repeatedly return the wrong script, unchanged text, validation errors, or an unacceptable delay.
Different models can return different wording and can have different response times. UnimeType applies the same local output checks to every model, but it does not promise identical output or equal accuracy across models.
If a correctness-sensitive result matters, read the text before sending it. A successful model response proves that the service returned usable text; it does not prove that every statement in the text is correct.
Configure your own API key
UnimeType exposes three provider choices:
| Provider choice | Prefilled behavior | Required fields |
|---|---|---|
| Ark | Prefills the Volcengine Ark service address and offers recognized presets | Model and API key |
| OpenRouter | Prefills the OpenRouter address; one key can expose multiple models | Model and API key |
| OpenAI-compatible | Starts from a compatible default and allows a custom service | Base URL, model, and API key |
For OpenRouter and OpenAI-compatible services, Browse models asks the saved
service for an authenticated GET /models list. An empty or failed list does
not by itself prove that a manually entered model is unusable. Save the exact
model ID required by your provider, then use Test connection.
Do not infer a provider from the shape of an API key. Choose the provider explicitly and copy the Base URL, model ID, and credential from that provider.
What sign-in changes
Signing in is required for the UnimeType managed route because the account supplies the credit balance and short-lived service access. Signing in is not required to spend credits at your own provider, and your own service does not consume the UnimeType balance.
Current plans and purchase terms belong on the Pricing page, not in this setup guide. Check that page before purchasing; this document does not freeze a price, credit amount, or model catalog.
Service troubleshooting
| Symptom | Check | Failure condition |
|---|---|---|
| Complete AI service setup first | Confirm the selected route has every required field | Base URL, model, credential, account, or balance access is missing |
| Test connection fails | Re-copy the Base URL, model ID, and API key; confirm the provider account can use that model | The saved service rejects authentication, the model, or the endpoint |
| Model browser is empty | Enter a known model ID manually, then test it | GET /models is unavailable or returned no IDs |
| Managed route stops working | Refresh account and balance state, then retry | Sign-in, available credits, or short-lived access is unavailable |
| The request succeeds but text stays unchanged | Try a clearer target or another model | The model returned the same structurally valid text |
For host capture and replacement failures, continue with Status and troubleshooting. A successful connection test does not prove that a particular editor will allow replacement.