---
title: BrainGPT
emoji: 🎙
colorFrom: purple
colorTo: indigo
sdk: gradio
app_file: app.py
pinned: false
suggested_hardware: a10g-small
---

# BrainGPT

Replaces the voice in a screen recording with a clean AI voice, synced to the video.

**Pipeline:** Whisper (GPU) transcription → Claude grammar/term correction (reads the
screen) → MiniMax AI voice → FFmpeg re-sync.

## Setup (owner)

This Space needs three secrets under **Settings → Variables and secrets**:

- `ANTHROPIC_API_KEY`
- `MINIMAX_API_KEY`
- `MINIMAX_GROUP_ID`
- `MINIMAX_VOICE_ID` *(optional — prefills the voice box)*

Set the hardware to **Nvidia A10G** under **Settings → Hardware**.