Skip to content
View wpp666's full-sized avatar
  • AH
  • beijing

Block or report wpp666

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
wpp666/README.md

王平 · wpp666

本地 AI 视频与 Agent 工具链的维护者。做的事情大多在离线、单机、不依赖云服务这一侧: 把图像生成、视频生成、TTS 这些能力串成能跑完的管线,并写成可复用的 Agent skill。

Maintainer of local-first AI video and agent tooling. My work sits on the offline, single-machine side: wiring image/video generation into pipelines that actually finish, and packaging them as reusable agent skills.


主要项目 / Featured

mv-local-pipeline — 本地 MV 五段式管线

从分镜到成片全在本机 ComfyUI 上跑,不向任何托管服务提交素材:

锚点图 (Krea 2) → 分镜 (Qwen-Image 2.1 + Krea 2) → 视频 (MiniMax H3)
      → 按母带精确拼接 → 中英双语歌词字幕

关键在于每段的时长对齐母带音频,而不是先出片再硬凑;字幕走 ASS 排版, 不用逐帧贴字。整条链路离线可复现。

Local, five-stage MV pipeline: anchors → storyboards → video → master-audio-aligned assembly → bilingual ASS subtitles. Nothing leaves the machine.

cloud-comfyui-ops — 把 ComfyUI 搬到租来的 GPU 上

总结了一个容易踩错的判断:搬管线到云 GPU,难点几乎从不是算力, 而是那台机器上的镜像到底装了什么、配置怎么改、文件怎么进出。 仓库里是完整流程加各家平台档案,重点是提交前逐节点、逐模型校验依赖是否存在, 避免排两小时队才发现缺一个自定义节点。

A procedure plus per-provider profiles for moving a local ComfyUI pipeline to a rented GPU box: probe the image, adapt config, verify every node and model exists before you submit a job.

VideoSubtitleRemover140mix — 给上游补回 GPU 支持

上游 video-subtitle-remover v1.4.0 只发了 CPU 安装包,但源码里的 CUDA 检测是完整的 —— 换 CUDA 版 torch 即可。 本仓库用嵌入式 Python 把这个组合打包成绿色便携版,适配 Windows + NVIDIA。

Restores GPU acceleration for upstream v1.4.0, which shipped CPU-only despite having intact CUDA code paths. Embedded Python, portable, no registry writes.

其它 / Also

  • dsh-lazy-studio — DSH 侧栏「导演台」: 四级串行评审(文本 → 锚点 → 分镜 → 视频),计数在客户端现算,不读宿主的过期快照。
  • clio-style-krea2 — Krea 2 提示词成稿工具, 内置 398 个视觉风格,遵守官方的自然语言写法规范。
  • trae-skills — 自研 Agent skill 合集备份。

环境 / Environment

Windows 10/11 · RTX 3070 Ti (8 GB) · CUDA 13.0 · ComfyUI (portable) · Python via pymanager · 主要语言 Python / JavaScript

8 GB 显存是这些项目共同的约束条件:上面多数设计取舍(分块加载、模型分档、 本地试片再上云)都是为了让管线在这张卡上跑得完。


联系 / Contact

Issues 和 PR 都欢迎。提问题时带上显存、CUDA 版本和完整的报错信息,能省一轮往返。

Issues and PRs welcome. Please include your VRAM, CUDA version, and the full error output — it saves a round trip.

GitHub

Popular repositories Loading

  1. sinaweibopy sinaweibopy Public

    Forked from michaelliao/sinaweibopy

    新浪微博Python SDK

    Python

  2. dubbox dubbox Public

    Forked from dangdangdotcom/dubbox

    Dubbox now means Dubbo eXtensions, and it adds features like RESTful remoting, Kyro/FST serialization, etc to the Dubbo service framework.

    Java

  3. api2go api2go Public

    Forked from manyminds/api2go

    JSONAPI.org Implementation for Go

    Go

  4. api-front api-front Public

    Forked from hidu/api-front

    api统一前端、转发代理、协议分析、流量复制

    Go

  5. oauth2 oauth2 Public

    Forked from go-oauth2/oauth2

    OAuth 2.0 server library for the Go programming language.

    Go

  6. rpcx rpcx Public

    Forked from smallnest/rpcx

    rpcx is a distributed RPC service framework based on net/rpc like alibaba Dubbo and weibo Motan. One of best performance RPC frameworks.

    Go