alesha-pro/tools

最近记录的来源活动: 2026年9月28日 12:49

SkillsMP 收录数据更新: 2026年11月2日 18:02

已收集 skills: 6

GitHub 星标: 695
GitHub Forks: 79

用 AI 助手安装

复制这段 Prompt，发给你正在使用的 AI 助手。

Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/alesha-pro/tools.

仓库 README

仓库 README 来源：README.md

查看 README 源文件

展开完整 README

<div align="center"> # tools Things I build while running models on my own hardware, cleaned up enough for other people to use. [![License](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE) [![Rig](https://img.shields.io/badge/measured%20on-4x%20RTX%203090-76b900.svg)](https://x.com/superalesha) [![X](https://img.shields.io/badge/@superalesha-follow-black.svg)](https://x.com/superalesha) </div> --- ## Sections Every section stands on its own, with its own installer and its own docs. There is no shared setup to get through first. ### [minimax-h3](minimax-h3) Video with sound, generated at home. [MiniMax H3](https://huggingface.co/Comfy-Org/MiniMax-H3) writes the picture and the synchronized audio in one pass, so voice, effects and music come out of the same generation as the frames. Open weights. Thirteen ComfyUI workflows in two sets. One set runs on any single 24 GB card. The other puts every card in a multi-GPU box to work and hits **3.12x on four cards**, taking a 15-second shot from 19 minutes of sampling down to 6:41. Ships with the custom nodes, the patch that makes the turbo LoRA actually load, and a three-command installer. ```bash cd minimax-h3 && ./install/setup.sh check ``` ### [qwen-image-2.1](qwen-image-2.1) Image generation, editing and transparent PNGs on **one GPU**. The baseline is an RTX 3090 with 24 GB VRAM, using official INT8 ConvRot weights. Five native ComfyUI workflows, an isolated installer, SHA-256 model verification, and a script that submits a generation and checks the resulting PNG. Includes an [English guide](qwen-image-2.1/README.md), a [Russian guide](qwen-image-2.1/README.ru.md), and an [installation brief for an agent](qwen-image-2.1/INSTALL-WITH-AGENT.md). For smaller GPUs, use the **[low-VRAM guide](qwen-image-2.1/LOW-VRAM.md)**: GGUF Q4_K_M, CPU offloading, 7 workflows, and measured memory requirements. ```bash cd qwen-image-2.1 python3 install/setup.py check ``` ### [skills](skills) Agent skills that stand on their own. **[hand-drawn-canvas-animation](skills/hand-drawn-canvas-animation)** makes short films with authored poses and expressive strokes in JavaScript and Canvas 2D. It covers pencil, ink, risograph, screen print and drawings interacting with real photos, plus sand animation and projected paper. The [60-second phoenix film](skills/hand-drawn-canvas-animation/examples/becoming-phoenix/) combines the materials through one story, including a moving storm and wings that unfold from a book. The skill includes editable examples, original sound, an MP4 renderer and browser regression checks. No Blender or video-generation model is required. ```bash cp -r skills/hand-drawn-canvas-animation ~/.agents/skills/ ``` **[motion-reel](skills/motion-reel)** makes product films, showreels and UI morphs from a brief and visual references. It includes a Canvas renderer, spring motion, original music and SFX, beat analysis, contact sheets and MP4 checks. The skill covers planning, animation, review and export, with source files you can edit. More lands here as I clean it up. --- ## The rig everything is measured on 4x RTX 3090, 96 GB VRAM total, PCIe 3.0 x16, no NVLink, 320 W per card. Ubuntu. Numbers in these READMEs come from real runs on that machine, not from datasheets. Your absolute times will differ, the ratios usually hold. There is one thing worth checking on your own box before anything else. Cards ship at whatever power limit the vendor set, and it is often well under the default. Mine were at 220 W against a 350 W default, and raising them to 320 took an identical run from 220 seconds down to 170. ```bash nvidia-smi --query-gpu=name,power.limit,power.default_limit --format=csv ``` --- <div align="center"> I post the benchmarks behind all of this. Local inference on 4x RTX 3090, open-weights models, argv and raw logs attached. **[@superalesha](https://x.com/superalesha)** MIT. Take it, change it, ship it. </div>

正在加载仓库 README…

这个仓库中的 skills

特效艺术家和动画师影片与视频编辑

2 个职业分类 · 已分类 67%

skill

职业分类

描述

更新

lyric-music-video

未分类

Make a code-rendered lyric music video for a finished song: every sung word lands on screen on its own syllable, inside dense animated graphics, with a recurring character built from generated stills. Covers the intake interview, style exploration with…

原文语言：英语

2026-09-28

motion-reel

未分类

Create finished code-rendered motion graphics: showreels, product launch films, kinetic typography, continuous UI morphs and animated explainers, with music, frame critique and verified MP4 exports. Use for motion-video production or an explicit motion-reel…

原文语言：英语

2026-09-27

hand-drawn-canvas-animation

特效艺术家和动画师

Create hand-drawn short films with whole-pose drawings, expressive strokes and intentional exposure in JavaScript and Canvas 2D: ink, pencil, risograph, screen print, or brush-pen doodles on photos. Includes performance animation, real-motion tracing, sand…

原文语言：英语

2026-09-25

h3-long-video

影片与视频编辑

Builds long videos (minutes) from a chain of 15-second MiniMax H3 clips while preserving characters, scene, voice, and style. Use when the user wants a video longer than 15 seconds, asks about stitching H3 clips, character consistency across generations,…

原文语言：英语

2026-08-09

h3-ref-prompt

特效艺术家和动画师

Turns a user idea plus reference media (images, video, audio) into a full full-reference prompt for MiniMax H3 (Ref2VA). Use when the user describes a video for H3/MiniMax and attaches or points at images (first/last frame, character, style, storyboard),…

原文语言：英语

2026-08-09

h3-t2v-prompt

特效艺术家和动画师

Turns a raw user idea into a ready production prompt for MiniMax H3 text-to-video (T2VA, no reference media). Use when the user describes a video idea, scene, or clip for H3/MiniMax and asks for a prompt, and has NOT attached any images/video/audio. Handles…

原文语言：英语

2026-08-09

已展示 6 / 6 个已收集 Skill。

alesha-pro/tools

用 AI 助手安装

仓库 README

相关分享

这个仓库中的 skills

lyric-music-video

motion-reel

hand-drawn-canvas-animation

h3-long-video

h3-ref-prompt

h3-t2v-prompt

仓库 README