convez.ai
PluginAutomaçãoIntermediário
Nusscookie

Clautter

Nusscookie/clautter

Automatiza edição de talking-head no DaVinci Resolve com IA

Ver no GitHub
InicialPluginScore · de 0 a 100

Plugin para DaVinci Resolve que automatiza edição de vídeos talking-head com IA, removendo silêncios, gerando legendas, aplicando zooms inteligentes e inserindo B-roll, música e efeitos sonoros. Funciona em Windows, macOS e Linux nas versões free e Studio do Resolve.

Instalação

$npx convez-ai Nusscookie/clautter

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
3
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Remover silêncios e pausas com detecção VAD automática
  • Gerar legendas via Whisper local ou ElevenLabs
  • Aplicar zooms dinâmicos baseados em detecção facial ou picos de áudio
  • Inserir B-roll automaticamente via busca em Pixabay/Pexels ou sugestão de LLM
  • Adicionar música de fundo e efeitos sonoros sincronizados com cortes

Veredito

Clautter é um plugin maduro e ambicioso que reimagina a automação de edição de vídeos talking-head: em vez de apenas sinalizar candidatos, ele toma decisões e aplica edições. Beta estável com suporte a múltiplos LLMs (OpenAI, Gemini, Anthropic, NVIDIA) e pipeline completo de produção.

README

A tradução para o português deste README ainda está na fila.

Clautter

CLAUDE + CUTTER — built by Claude Code. A DaVinci Resolve plugin for editing talking-head videos faster.

Status: BETA — core features work, bugs expected.

Works with DaVinci Resolve free and Studio on Windows, macOS, and Linux.


Philosophy

Clautter is built around one idea: the editor shouldn't have to babysit the tool.

Most Resolve plugins for talking-head cleanup are assistants — they surface candidates, mark regions, or highlight pauses, then wait for you to confirm or discard each one. That's still a lot of clicking. Clautter goes the other direction: it analyzes the timeline, makes the call, and applies the edit.


Features

TabStatusDescription
Smart Cuts✅ FullRemove silences with a pace slider (1–10) + fine-grained controls. VAD-based detection. Padding scales with pace preset and is applied as total (not per-edge).
Subtitles✅ FullElevenLabs or local Whisper STT → styled Fusion Title subtitle track. Hormozi/TikTok word-by-word mode. Re-running replaces existing subtitle clips automatically.
Auto Zooms✅ FullFace Detection (OpenCV) or RMS peaks → zoom cuts with Fusion ease-in/out.
B-Roll✅ FullManual: folder scan + keyword match + online search (Pixabay/Pexels). Autonomous: one-click end-to-end pipeline. The LLM mode button is a provider picker — Off plus one segment per cloud LLM you have a key for (OpenAI / Gemini / Minimax / NVIDIA / Anthropic); pick one to let the LLM choose clips and positions.
Music & SFX✅ FullMood-matched background music (Jamendo / local folder), keyword or LLM mood engine, single-track or per-segment placement. Auto-placed sound effects (Freesound / Pixabay) keyed off cuts, zooms, and B-roll. Volume, fades, and ducking baked in via pydub.
Motion Graphics🔧 BetaRule-based graphic suggestions. AI generation planned for V2.

Clautter supports OpenAI, Gemini, Minimax, Anthropic, and NVIDIA. NVIDIA grants free access to many open-source models via an OpenAI-compatible API, so you can run the cloud-LLM features without paying — paste an NVIDIA key in Settings → LLM Keys and the model id (e.g. moonshotai/kimi-k2.6) in Settings → LLM Models.


Quick Start

Python 3.10 / 3.11 / 3.12 only. DaVinci Resolve's compiled scripting module segfaults on 3.13+.

py -3.12 install.py             

See INSTALL.md for the full guide.


Architecture

Two-process model via a localhost HTTP bridge. main.py runs inside Resolve's process (where the resolve object lives), starts a bridge server, then spawns gui.py as a subprocess. The GUI talks to Resolve through the proxy — no direct scripting access needed.

See CLAUDE.md for full design notes and gotchas.

Visual flow maps of every feature — steps, decision points, and the real thresholds — live in docs/whiteboard/ as editable Excalidraw diagrams, so you can see how a feature works without reading the code.


Design

Brand color palette documented in design/palette.md. Primary accent: #D97757 (terracotta orange, from the app icon).


Dependencies

customtkinter, pydantic, faster-whisper, silero-vad, torchaudio, soundfile, onnxruntime, spacy, yake, keybert, joblib, sentence-transformers, opencv-python, scenedetect, pydub, requests, numpy, Pillow — plus ffmpeg (system).

Install: py -3.12 -m pip install -r requirements.txt


Compatibility

  • DaVinci Resolve free and Studio (v18+)
  • Python 3.10 / 3.11 / 3.12 (3.13+ segfaults on import)
  • Windows, macOS, Linux

Tested on Windows only. macOS and Linux are untested. API key status: ElevenLabs ✅ · NVIDIA ✅ · OpenAI ⚠ untested · Gemini ⚠ untested · Anthropic ⚠ untested.


Support

This is a part-time project. A paid installer is planned — pay once, get a one-click setup that handles Python, ffmpeg, and the plugin. No fiddling required. The installer is a way to support the project; the plugin itself stays free on GitHub.

No website or community yet. Issues and PRs welcome.

Testing help wanted. If you run Clautter on macOS or Linux, or test an API key (OpenAI, Gemini, Anthropic), please open an issue and report whether it worked. This helps update the compatibility table above.


License

GPLv3

Ver o original no GitHub

Bom para

Relacionados