linyqh/NarratoAI

NarratoAI – AI‑powered video narration & automated editing

What it is – An open‑source desktop/web app that lets you turn raw video footage into narrated short videos. It strings together a large‑language model (LLM) for script generation, speech‑synthesis (TTS) engines for voice‑over, automatic subtitle creation, and a simple video‑cutting pipeline. The whole flow runs locally (or via the hosted cloud version) and is controlled through a Streamlit UI.

Key capabilities

• Script writing – Prompt an LLM (e.g., GLM‑5.3, Kimi‑K3, DeepSeek, MiniMax, Doubao, Qwen2‑VL) to produce a narration script for a given video or topic.
• Automatic cutting – The tool extracts key frames, selects highlights, and assembles a trimmed video clip.
• Voice‑over – Supports multiple TTS back‑ends, including IndexTTS‑1.5/2 (MLX on Apple Silicon), OmniVoice, Tencent Cloud TTS, and others. Voice cloning is available.
• Subtitle generation – Integrated ASR (Fun‑ASR) produces subtitles that can be previewed and edited.
• AI‑generated background music – Optional Sonilo AI music track.
• Export – Can output a final video file or a draft for the Chinese “剪映” editor.

Installation options

1. All‑in‑one package – Download a pre‑built zip for Windows x64 or macOS Apple‑Silicon, run update‑*.bat/.command then start‑*.bat/.command. The UI appears at http://127.0.0.1:8501.
2. Docker (macOS) – docker compose up -d after cloning the repo.
3. From source –

   git clone https://github.com/linyqh/NarratoAI.git
   cd NarratoAI
   uv sync                     # install deps
   cp config.example.toml config.toml
   # edit config.toml with your API keys
   uv run streamlit run webui.py --server.maxUploadSize=2048
   

   Then open the same localhost address.

Typical workflow

1. Upload or point the app to a source video.
2. Choose a language model (or let the app pick the best one) and generate a narration script.
3. Select a TTS engine, optionally provide a reference voice for cloning.
4. Run the automatic cut‑and‑merge step; the app will also run ASR to produce subtitles.
5. Review/edit script, subtitles, or timing if needed.
6. Export the finished video or a draft for further editing.

Hardware & software requirements

• CPU 4 cores+, 8 GB RAM (GPU optional).
• Windows 10/11 or macOS 11+.
• Python 3.12+ (if running from source).

Community & support

• Official wiki, Discord, and a GitHub community page for announcements.
• Frequent releases (latest v0.8.6 adds Apple‑Silicon TTS engines and OmniVoice packs).
• Sponsorship links to Volcengine, APIMart, and SiliconFlow provide low‑cost API credits for video‑understanding and TTS.

License – See the LICENSE file in the repository (standard open‑source license, details not specified in the README).

Bottom line – NarratoAI is a genuine, actively maintained AI tool that integrates LLM‑based script generation, multi‑engine TTS, ASR, and video editing into a single, easy‑to‑run package for creators who want automated film‑style narration.