PopLingo speaks the OpenAI-compatible API, so any service that offers
/chat/completions works. Pick a service in Settings, paste your key, done.
Right-click the tray icon → Settings → the Translation tab.
| Field | Description |
|---|---|
| Provider | Dropdown. Choosing one auto-fills the API URL and model list. |
| API URL | Editable (used with the "Custom…" provider). |
| API Key | Your key from the provider. |
| Model | Dropdown, but you can type any model name the provider offers. |
| Get API key | Opens the provider's signup page. |
Good Chinese and English quality at a low price; the best default choice.
| Field | Value |
|---|---|
| Provider | DeepSeek |
| API URL | https://api.deepseek.com |
| Model | deepseek-flash (cheap and fast)or deepseek-v4-pro (higher quality) |
DeepSeek is a paid service billed by usage. Translation uses very few tokens, so typical monthly cost is tiny.
| Provider | API URL | Example model |
|---|---|---|
| OpenAI | https://api.openai.com/v1 | gpt-4o-mini |
| Alibaba Qwen (DashScope) | https://dashscope.aliyuncs.com/compatible-mode/v1 | qwen-plus |
| Zhipu GLM | https://open.bigmodel.cn/api/paas/v4 | glm-4-flash |
| Kimi (Moonshot) | https://api.moonshot.cn/v1 | moonshot-v1-8k |
| SiliconFlow | https://api.siliconflow.cn/v1 | deepseek-ai/DeepSeek-V3 |
| Tencent Hunyuan | https://api.hunyuan.cloud.tencent.com/v1 | hunyuan-turbo |
| Volcengine Doubao | https://ark.cn-beijing.volces.com/api/v3 | doubao-1.5-lite-32k |
| MiniMax | https://api.minimax.chat/v1 | abab6.5s-chat |
| Groq | https://api.groq.com/openai/v1 | llama-3.3-70b-versatile |
| Mistral | https://api.mistral.ai/v1 | mistral-small-latest |
| OpenRouter | https://openrouter.ai/api/v1 | openai/gpt-4o-mini |
All of these are built into the dropdown, so you never type the URL by hand. Model names change over time, so the model field accepts free text.
Don't want to send anything to the cloud? Run a model on your own machine with Ollama — free, offline, and nothing leaves your PC.
ollama pull qwen2.5:7b
ollama list
| Field | Value |
|---|---|
| Provider | Ollama (local) |
| API URL | http://localhost:11434/v1 |
| API Key | ollama (ignored locally, any value works) |
| Model | qwen2.5:7b |
| Model | Memory | Notes |
|---|---|---|
qwen2.5:3b | ~4 GB | Lightest, acceptable quality |
qwen2.5:7b | ~8 GB | Recommended balance of quality and speed |
qwen2.5:14b | ~16 GB | Better quality, needs a stronger machine |
gemma3:4b | ~5 GB | Decent multilingual support |
Note: the first response from a local model is slow (it has to load into memory), and quality is usually a bit below cloud models like DeepSeek or GPT. Best for privacy-conscious users, or if you'd rather not pay.
You can also use LM Studio: start its local server,
then pick "LM Studio (local)" in Settings. The API URL is http://localhost:1234/v1.
| Field | Description |
|---|---|
| Source language | Defaults to "Auto detect" and lets the model decide. |
| Target language | Defaults to English. |
| Second language | Used when the source already is the target language. "(None)" disables it. See the guide. |
| Style | Academic (default) / Business / Casual. |
| Delay (ms) | How long to wait after you stop typing before translating. Default 500. |
Only the text you type is sent, and only to the API you configured. PopLingo itself collects nothing and contains no tracking or telemetry. With a local Ollama model, your text never leaves your computer.