Kinetic Video Creator

Kinetic Video Creator

Create stunning kinetic typography videos with AI-generated speech, music, and dynamic animations.

Workflow Overview

1. Script → Craft emotionally compelling speech text
2. Speech → Use /speech-generator skill for TTS
3. Transcribe → Use /transcribe skill for word timing
4. Music → Use /music-generator skill for background
5. Merge → Combine speech + music
6. Animate → Create kinetic typography in Remotion
7. Render → Produce final video
8. Publish → Use /youtube-uploader skill (optional)

Step 1: Craft the Script

Language Selection

Hebrew (Recommended for aviz's voice):

• Use Hebrew emotional directions in brackets
• Add natural Hebrew filler words
• See "Hebrew Script Guidelines" section below

English:

• Use English emotional directions
• See "English Script Guidelines" section below

Hebrew Script Guidelines

aviz's cloned voice is optimized for Hebrew. Use these Hebrew directions:

Hebrew Emotional Directions

Hebrew Filler Words (for natural flow)

• אממ... - hesitation
• אהה... - thinking
• כאילו... - like
• נו... - well
• יאללה... - come on
• בקיצור... - in short
• ... - pause

Hebrew Script Example

[נשימה עמוקה] יש רגע...
[לאט ובבירור] רגע שהכל משתנה.

[ברצינות] אבא שלי חלה בפוליו כשהיה תינוק.
כל חייו הוא היה על כיסא גלגלים.

[בהתלהבות] אבל אבא שלי? הוא היה ספורטאי מצטיין!
[בחום] הוא תמיד האמין... שאפשר להגשים כל חלום.

[בעצב] כשהייתי בן חמש עשרה... אבא נפטר.

[בכוח] והכאב הזה? הפך למשימה שלי.
[בחום] לעזור לאנשים אחרים להגשים את החלומות שלהם.

English Script Guidelines

English Emotional Directions

English Script Template

[HOOK - 5-10 seconds]
[dramatic pause] Opening line that grabs attention.
[slowly, with weight] The provocative statement.

[BUILD - 20-40 seconds]
[building intensity] Establish the context.
[pause for effect] Key insight moment.

[PEAK - 20-30 seconds]
[powerful, emphatic] The main message.
[pause] Let it land.

[RESOLVE - 15-25 seconds]
[warm, inspiring] Paint the vision.
[final beat] Memorable closing.

Step 2: Generate Speech

Use the speech-generator skill:

/speech-generator [path/to/script.txt] -o [path/to/speech.mp3]

Or invoke directly:

cd ~/.claude/skills/speech-generator/scripts
npx ts-node generate_speech.ts -f script.txt -o speech.mp3

Important: The speech-generator uses aviz's cloned voice, which works best with Hebrew text and Hebrew emotional directions.

Step 3: Transcribe for Timing

Use the transcribe skill:

/transcribe [path/to/speech.mp3] --json

Or invoke directly:

cd ~/.claude/skills/transcribe/scripts
npx ts-node transcribe.ts -i speech.mp3 -o transcript.srt --json

Output: transcript_transcript.json with word-level timing data.

Step 4: Generate Background Music

Use the music-generator skill:

/music-generator [composition.json] -o background_music.mp3

Music Composition Template

{
  "duration_ms": 75000,
  "instrumental": true,
  "positive_global_styles": ["cinematic", "inspirational"],
  "negative_global_styles": ["aggressive", "chaotic"],
  "sections": [
    {
      "section_name": "Hook - Mysterious",
      "duration_ms": 12000,
      "positive_local_styles": ["suspenseful", "soft"],
      "negative_local_styles": ["loud"],
      "lines": []
    },
    {
      "section_name": "Build - Rising",
      "duration_ms": 25000,
      "positive_local_styles": ["hopeful", "building"],
      "negative_local_styles": ["slow"],
      "lines": []
    },
    {
      "section_name": "Peak - Triumphant",
      "duration_ms": 20000,
      "positive_local_styles": ["triumphant", "uplifting"],
      "negative_local_styles": ["quiet"],
      "lines": []
    }
  ]
}

Step 5: Merge Audio

ffmpeg -y \
  -i speech.mp3 \
  -i background_music.mp3 \
  -filter_complex "[0:a]volume=1.0[speech];[1:a]volume=0.15[music];[speech][music]amix=inputs=2:duration=first[out]" \
  -map "[out]" -c:a libmp3lame -q:a 2 \
  final_audio.mp3

Step 6: Create Remotion Composition

Project Location

cd /Users/aviz/remotion-assistant

Default Template: SequenceComposition (One Word Per Screen)

Recommended: Use SequenceComposition for maximum impact - displays one word at a time with full-screen typography.

import { SequenceComposition } from '../templates/SequenceComposition';
import transcriptData from '../../projects/[project]/transcript_transcript.json';

const WORD_TIMINGS = transcriptData.words
  .filter((w) => w.word.trim() !== '')
  .map((w) => ({
    word: w.word,
    start: w.start,
    end: w.end,
  }));

export const MyVideo: React.FC = () => {
  return (
    <SequenceComposition
      wordTimings={WORD_TIMINGS}
      audioFile="[project]/final_audio.mp3"
      baseFontSize={200}
      dustEnabled={true}
      lightBeamsEnabled={true}
      centerGlowEnabled={true}
      glowIntensity={1}
      anticipationFrames={5}
      colorSchemeStart={0}
    />
  );
};

Alternative: MultiWordComposition (Word Cloud)

Use for faster-paced content with multiple words on screen:

import { MultiWordComposition } from '../templates/MultiWordComposition';

Hebrew Font Support

For Hebrew text, use Heebo font:

import { loadFont } from '@remotion/google-fonts/Heebo';

const { fontFamily } = loadFont('normal', {
  weights: ['400', '600', '700', '900'],
  subsets: ['hebrew', 'latin'],
});

Add RTL styling:

style={{
  direction: 'rtl',
  fontFamily,
}}

Step 7: Render

cd /Users/aviz/remotion-assistant
npx remotion render CompositionName output.mp4

Step 8: Upload (Optional)

Use the youtube-uploader skill:

/youtube-uploader [video.mp4] --title "Title" --description "Description"

Project Structure

remotion-assistant/
├── public/[project]/
│   └── final_audio.mp3      # Audio for Remotion
├── projects/[project]/
│   ├── speech.txt           # Script
│   ├── speech.mp3           # TTS output
│   ├── transcript_transcript.json  # Word timings
│   ├── music_composition.json
│   ├── background_music.mp3
│   ├── final_audio.mp3      # Merged audio
│   └── output.mp4           # Final video
└── src/compositions/
    └── [ProjectName].tsx    # Composition

Quick Reference