video-toolkit

Generates explainer videos with AI voiceovers, images, music, and Remotion rendering.

Updated Jul 16, 2026
One-click install
npx skills add https://github.com/X-manist/Cohmira --skill video-toolkit-x-manist
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: video-toolkit
Source: https://github.com/X-manist/Cohmira/tree/main/src/builtin-plugins/openmontage/.agents/skills/video-toolkit
Command: npx skills add https://github.com/X-manist/Cohmira --skill video-toolkit-x-manist

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires modal, remotion, ffmpeg.

What problem does it solve? Producing professional explainer or product demo videos normally requires separate tools for voiceover, image generation, music, animation, and video editing. This Skill orchestrates the entire pipeline from a text brief to a rendered MP4 using cloud GPU models and Remotion. ## Core Features & Use Cases - AI Asset Generation: Create per-scene voiceovers (Qwen3-TTS), background music (ACE-Step), scene images (FLUX.2), b-roll clips (LTX-2), and talking-head narrators (SadTalker) on Modal cloud GPUs. - Remotion Composition: Assemble scenes with per-scene audio, narrator picture-in-picture, and transitions using React-based Remotion templates like product-demo. - Timing Sync & Rendering: Measure actual audio durations with ffprobe, sync scene durations, review still frames, and render the final MP4. - Use Case: Given a product brief, generate a 60-second marketing video with narration, background music, animated scenes, and a call-to-action, rendered to out/ProductDemo.mp4. ## Quick Start Ask the agent to create a 60-second product demo video for your product using the video toolkit, starting from a short text brief.

Frequently Asked Questions about video-toolkit

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create an AI-generated explainer video from a text brief?

Copy a Remotion template like product-demo into the projects folder, write a scene config and voiceover script, then generate per-scene audio, images, and music with the toolkit's Python tools. Finally sync timing with ffprobe and run npm run render to produce the MP4.

How do I generate AI voiceover audio for video scenes?

Use the qwen3_tts.py tool with a speaker like Ryan and a tone like warm, generating one MP3 file per scene. It supports nine speakers, eight tones, and voice cloning from a reference audio file with its transcript.

Does the video toolkit require a local GPU?

No, all AI models run on Modal cloud GPUs accessed through endpoint URLs configured in the .env file. Modal includes $30 per month of free compute, and a typical 60-second video costs about $1-3.

Why is my SadTalker talking-head video square or cropped?

SadTalker defaults to crop preprocessing, which outputs a square video with the wrong aspect ratio. Re-run the command with --preprocess full and --still to get a properly framed, professional-looking result.

What are the limitations of LTX-2 video clip generation?

LTX-2 clips are limited to about 8 seconds at 24fps, with dimensions divisible by 64, and take roughly 2.5 minutes per clip. Generated audio is ambient only, and about 30% of outputs may contain artifacts, so re-running with a different seed may be needed.