Use Immersive Translate with Models Hub: Custom AI Service Setup
Switch Immersive Translate's browser extension to a custom OpenAI service, enter your Models Hub URL and key, and translate web pages on your Models Hub balance, with verification and error fixes.
Immersive Translate is a browser extension that overlays a translation directly on the original page, and it can be switched to any OpenAI-compatible AI translation service. Point that service at Models Hub and your everyday page translations, papers, and video subtitles all draw on your Models Hub balance, no separate OpenAI account needed. This page covers only the connection step.
Three things you need
A Models Hub API key, the endpoint https://modelsok.com/v1/chat/completions (the full path, not just a base URL), and a model name.
Prerequisites
- The Immersive Translate browser extension installed. If not, get it from the official site for your browser.
- A Models Hub API key from the Tokens page, in a group that can access the models you want.
Recommended models
Page translation generates a lot of frequent requests, so start with something cheap and fast; switch to a stronger model for long documents or technical material where accuracy matters more.
| Use case | Model | Pricing | Live status |
|---|---|---|---|
| Default for everyday page translation, fast | gpt-5.4-mini | Price | Availability and latency |
| Cheap and fast, alternative | claude-haiku-4-5 | Price | Availability and latency |
| Long documents or technical content needing accuracy | claude-sonnet-4-6 | Price | Availability and latency |
Model names are authoritative on the pricing page.
Setup
-
Click the Immersive Translate icon in your browser toolbar and open Settings (the gear icon opens a dedicated settings page).
-
Find Translation Service on the left and select OpenAI. Note: as of the current version this section may be labeled "Translation Service" or "AI Translation Service" depending on your build; go by what you actually see.
-
Expand More Settings, which reveals a few new fields:
Field What to enter API Key Your Models Hub API key; separate multiple keys with commas for load balancing Custom URL https://modelsok.com/v1/chat/completions(the official examples all use this full path, not just/v1)Custom model One of the model names above, e.g. gpt-5.4-miniNote: these are the official field names from the docs (
api_key,Custom URL,Custom model); your version's on-screen labels may vary slightly. -
If translations feel unstable or you see frequent errors, lower maximum requests per second; the official recommendation is to drop it from the default of 10 to 5.
Verify
After saving, open any non-English page and trigger translation (via the extension icon or the bilingual display). Seeing the translated text appear means you're connected. Then check Usage logs in the Models Hub console; a request under that model name appears within seconds, confirming traffic went through Models Hub.
Common errors
| Symptom | Cause | Fix |
|---|---|---|
| Page keeps showing "translation failed" or spins forever | Wrong Custom URL, most often ending at /v1 instead of the full path | Go back to More Settings and use the full https://modelsok.com/v1/chat/completions |
401 or invalid key message | Incomplete key, stray whitespace or a line break, or a disabled token | Re-copy from the Tokens page, watch for extra spaces |
429 or a burst of failures | Maximum requests per second is set too high | Go back to More Settings and lower it from the default to 5 |
| Model not found message | Misspelled custom model name, or the token's group can't access it | Check the spelling on the pricing page |
| Translations don't change, still using the default service | Translation Service wasn't actually switched, or settings weren't saved | Reopen settings and confirm OpenAI is selected and all three fields are saved |
More in Troubleshooting.
FAQ
How is it billed? Per input and output token actually consumed per translation request, at the price on each model's pricing page. A single page translation splits into several small requests; usage logs show the detail, so one page won't rack up a large cost.
How is this different from using OpenAI's official endpoint directly? Same fields and format, just a different domain. You don't need to sign up for an OpenAI account or add a credit card.
How do I switch models? Go back to More Settings, change Custom model to another name from the table, and save. It takes effect immediately, no browser restart needed.
Related
Use FluentRead with Models Hub: Custom Translation Engine Setup
Point the FluentRead browser extension's translation engine at Models Hub, enter your key and endpoint, and translate pages on your Models Hub balance, with recommended models and error fixes.
Use LunaTranslator with Models Hub: LLM Interface Setup
Add an LLM interface in LunaTranslator pointed at Models Hub, enter the API URL and key, and translate visual novels on your Models Hub balance, with recommended models and error fixes.