Subtitle Translator

Batch-translate subtitles into 120+ languages, with TokenMix built in.

What this is

TokenMix Subtitle Translator is a free, browser-based batch subtitle translator for .srt, .ass, .vtt, and .lrc files. It is a TokenMix-customized build of the open-source subtitle-translator by rockbenben — the translation engine ships with TokenMix already wired in as the default, so you only add your own sk-tm- key and start translating. One endpoint reaches Claude, GPT, Gemini, DeepSeek, Kimi, GLM, Qwen and more; switch models any time to compare quality and cost.

Everything runs client-side: subtitle parsing, translation, and caching all happen in your browser. Subtitle content and your API key never touch a server.

Before you start

Sign up at tokenmix.ai and top up your wallet.
Open the console, go to API Keys, and create an sk-tm-... key.

That key is all you need — the app already points at https://api.tokenmix.ai/v1.

Deploy your own copy

This is a deploy-type app: you run your own instance from the open-source fork. Pick whichever way fits you.

Cloudflare Pages or Vercel (static, easiest):

Import TokenMixAi/subtitle-translator on Vercel or Cloudflare Pages.
Build command npm run build, output directory out.
Importing the repo and deploying with the defaults works out of the box.

Docker (self-hosted server):

git clone https://github.com/TokenMixAi/subtitle-translator
cd subtitle-translator
docker build -t subtitle-translator .
docker run -p 3000:3000 subtitle-translator

Then open http://localhost:3000.

Local (try it first):

git clone https://github.com/TokenMixAi/subtitle-translator
cd subtitle-translator
yarn install && yarn dev

Configure your key

Open the app, then open API Settings (the settings drawer on the translator page).
The translation engine is already set to TokenMix. Paste your sk-tm-... key.
The default model is DeepSeek V4 Flash — fast and cheap for whole seasons. For the most natural dialogue, switch to Claude Sonnet 4.6, GPT-5.5, or Gemini 3.5 Flash from the model dropdown.

Verify it works

Drop in a .srt file, choose a target language, and click translate. A short file should finish in a second or two. If you see a 401, re-check the key; if a few lines fail, the app fills them with the original text and lets you retry just those lines.

FAQ

Do I need to set the API endpoint? No. This build points at https://api.tokenmix.ai/v1 already. Just paste the key.

Which model should I pick? DeepSeek V4 Flash is the value pick for batch work; Claude, GPT, and Gemini read the most natural. You can switch per job.

Is my data private? Yes — parsing, translation requests, and caching all happen in your browser. Your key lives only in your browser's local storage.

Can I use a different provider? Yes. The app keeps every upstream engine (DeepL, Google, OpenAI, Claude, any OpenAI-compatible endpoint, and more); TokenMix is simply the default.

About this build

This is an open-source fork of rockbenben/subtitle-translator (MIT), with TokenMix wired in as the default provider and branding adjusted. All subtitle-processing and translation logic is the upstream author's work.