YeJe-cpu/SeeCut

An AI editor that watches its own cut: talking-head / AI-avatar video → auto-edited short video. 网感口播精剪：数字人/真人口播 → AI 自动精剪成网感动效短视频 → 可选剪映分层草稿

What it solves

SeeCut is an AI-powered editing tool designed to transform raw talking-head videos (from humans or digital avatars) into high-quality, engaging short videos. It automates the process of finding visual evidence, arranging clips, adding dynamic effects, and performing self-quality checks to ensure the final output feels professional and "on-trend."

How it works

SeeCut functions as a skill for coding agents (like Claude Code). The process follows a three-step pipeline:

  1. Preparation: It generates or accepts a talking-head video.
  2. Editing: The agent uses an AI model (via Antigravity CLI) to "watch" the video, understand the content, and fetch relevant screenshots or data as visual evidence. It then uses the HyperFrames rendering engine to create dynamic visuals based on a structured design system.
  3. Self-Iteration: The agent performs a quality check by comparing different versions of the edit against each other, selecting the best version, and refining it over multiple iterations before final delivery.

Who it’s for

This project is for creators, developers, or content teams who want to automate the production of high-quality, information-dense short videos. It is specifically designed for users who are comfortable running coding agents and managing local development environments on macOS.

Highlights

  • AI Self-Correction: Uses a "two-way comparison" method where the AI evaluates its own edits against previous versions to ensure continuous improvement.
  • Visual Evidence Integration: Automatically captures real-world screenshots and data to match the spoken content, avoiding generic filler footage.
  • Structured Design System: Uses a systematic approach (design tokens, component libraries, and shot lists) rather than random effects to ensure a consistent, professional look.
  • Flexible Output: Produces a final rendered video and can optionally generate a layered draft project for further manual fine-tuning in Jianying (CapCut).
  • Digital Human Support: Provides a complete workflow for combining high-quality voice synthesis (Doubao) with digital avatars (HeyGen) for automated video production.

Related

  • Project
    Anil-matcha/Open-AI-UGC

    Open AI UGC is an open‑source, self‑hostable web app that lets users generate short vertical UGC video ads using multiple state‑of‑the‑art text‑to‑video models. It includes Google OAuth, a credit system with Stripe billing, async webhook handling for renders, and a dashboard of past creations. The stack is Next.js 16, React 19, Tailwind, Prisma, and Muapi.ai for the underlying video generation.

  • Project
    pytraveler/MiniMax-H3-Prompt-Rewriter-ComfyUI

    ComfyUI nodes that use LLMs and LoRAs to rewrite short prompts into structured, production-ready audio-video descriptions for the MiniMax-H3 video generator.

  • Project
    ModelTC/Minimax-H3-Turbo

    A collection of distilled LoRA checkpoints for MiniMax-H3 that enables fast, few-step video and audio generation via Diffusers and ComfyUI.

  • Project
    liyupi/ai-promo-video-maker

    An open‑source tool that lets you describe a product in one sentence and, using LLM agents, automatically researches, designs style frames, captures 4K screenshots, generates music/voice (ElevenLabs optional), assembles scenes with Remotion, validates timing and audio, and outputs a finished promotional video.