notepower2k1/CapCap

CapCap – AI‑powered video editing & localisation for Windows

What it does
CapCap is a Windows‑only desktop app that lets you take a raw video, automatically generate subtitles, translate them, synthesize new voice‑overs, and export a fully edited video. The whole workflow is split into five visual steps:

1. Prepare – import the source video.
2. Transcribe – speech‑to‑text (STT) or OCR for on‑screen text.
3. Translate – AI‑driven, context‑aware translation with support for several large‑language‑model services (Gemini 2.5, GPT‑4o, DeepSeek, Ollama, Google/Bing Translate).
4. Dub – text‑to‑speech (TTS) using both offline (Piper) and online (Edge, CapCut, VieNeu) engines, with optional background‑music separation.
5. Export – render the final video (GPU‑accelerated via NVIDIA NVENC) or output separate audio tracks or a CapCut draft.

Key features

• Adaptive UI for both landscape (16:9) and portrait (9:16) formats, automatically re‑layouting overlays for YouTube‑style or TikTok‑style videos.
• Context‑aware translation that analyses character relationships and can adjust pronouns/levels of formality.
• Smart voice‑over: offline Piper and several cloud TTS services; can re‑synthesize a line instantly after subtitle edits.
• Multi‑layer timeline: add subtitles, logos, blur, masks, text, etc., with a visual editor.
• Fast GPU rendering using NVIDIA NVENC; also supports exporting just the audio tracks.

How to run it (from the README)

# clone and set up a virtual environment
git clone https://github.com/notepower2k1/CapCap.git
cd CapCap
python -m venv venv
venv\Scripts\activate   # on Windows
pip install -r requirements-local.txt
python ui/gui.py

• Copy .env_example to .env to add your own API keys for the LLM/TTS services, though most settings can be changed inside the GUI.
• The preview player uses libmpv; you need the Microsoft Visual C++ 2015‑2022 Redistributable (x64) – the repo ships an installer for it.

Technical stack (as listed in the docs)

• Python backend handling STT/OCR, translation calls, TTS, and video processing.
• Large‑language‑model APIs: Gemini 2.5, OpenAI GPT‑4o, DeepSeek, Ollama (self‑hosted).
• TTS engines: Piper (offline), Edge TTS, CapCut TTS, VieNeu TTS.
• Video preview via libmpv; final encoding with NVIDIA NVENC.
• UI built with a Python GUI framework (details in docs/technical-stack.md).

Who might use it

• Content creators who need to repurpose a single video for multiple platforms (YouTube, TikTok, Shorts, Reels) without manual transcription and dubbing.
• Small studios or freelancers looking for an all‑in‑one localisation pipeline that works offline for privacy‑sensitive material.

License

• Apache License 2.0.

---

The information above is taken directly from the repository’s README and linked documentation.