<div align="center">
<img width="1200" height="475" alt="GHBanner" src="https://github.com/user-attachments/assets/0aa67016-6eaf-458a-adb2-6e31a0763ed6" />
</div>

# Run and deploy your AI Studio app

This contains everything you need to run your app locally.

View your app in AI Studio: https://ai.studio/apps/a38a3cd5-7f82-49f0-a26e-99be4d77f863

## Run Locally

**Prerequisites:** Node.js, system `ffmpeg`, and system `ffprobe`


1. Install dependencies:
   `npm install`
2. Configure [.env](.env) with:
   `ARK_API_KEY`
   `GEMINI_API_KEY`
   `MINIMAX_API_KEY`
3. Optional defaults:
   `DEFAULT_LLM_PROVIDER=doubao`
   `DOUBAO_MODEL=doubao-seed-2-0-pro-260215`
   `VITE_API_BASE_PATH=/api` (optional)
   `FFMPEG_PATH=/path/to/ffmpeg` (optional)
   `FFPROBE_PATH=/path/to/ffprobe` (optional)
4. Run the app:
   `npm run dev`

## Model Switching

1. Subtitle generation now runs through the server and supports `Doubao` and `Gemini`.
2. The editor shows an `LLM` selector and defaults to `Doubao`.
3. `TTS` stays fixed on `MiniMax` regardless of the selected LLM.
4. All provider keys are read from `.env`; the browser no longer calls LLM providers directly.

## Deploying Under a Subpath

1. If your site is hosted under a subpath (for example `https://ai.yantootech.com/video_translate/`), set:
   `VITE_API_BASE_PATH=/video_translate/api`
2. Keep backend routes unchanged (`/api/*` on the server); the frontend will prepend the configured base path.

## Subtitle Generation

1. Subtitle generation is now driven by server-side multimodal LLM calls on the uploaded video file.
2. No separate local alignment/ASR backend is required for `/api/generate-subtitles`.