Creating Brand Profiles

August 26, 2026 · View on GitHub

Brand profiles let you maintain consistent visual identity across videos. Each brand defines colors, fonts, and voice settings.

Quick Start: Run /brand in Claude Code to create a brand interactively. It can extract colors from a website URL, generate a palette from a single color, or let you specify colors manually.

Brand Structure

brands/
├── default/
│   ├── brand.json     # Colors, fonts, spacing
│   ├── voice.json     # Voice settings
│   └── assets/        # Logo, backgrounds, watermarks
│       ├── logo.svg
│       └── background.mp4
└── my-company/
    ├── brand.json
    ├── voice.json
    └── assets/

Creating a New Brand

Copy the default brand
```
cp -r brands/default brands/my-company
```

Edit brand.json

{
  "name": "My Company",
  "description": "Corporate brand for My Company videos",
  "colors": {
    "primary": "#2563eb",
    "primaryLight": "#3b82f6",
    "textDark": "#1e293b",
    ...
  },
  "fonts": {
    "primary": "Inter, sans-serif",
    "mono": "JetBrains Mono, monospace"
  }
}

Configure voice settings (optional) Edit voice.json to use a different ElevenLabs voice:

{
  "voiceId": "your-voice-id",
  "settings": {
    "stability": 0.75,
    "similarityBoost": 0.9
  }
}

Add brand assets Place logo, background videos, or watermarks in assets/.

Using a Brand in a Project

When creating a new project with /video, you'll be prompted to select a brand:

Which brand?

  1. default - #3B82F6 (blue)
  2. digital-samba - #3771e0 (blue)
  3. my-company - #your-color
  4. Create new brand → /brand

The selected brand's colors, fonts, and assets are automatically applied to your project via src/config/brand.ts.

Manual Brand Loading

If you prefer manual setup, reference the brand in your project's config:

// brand.ts (auto-generated by /video command)
import { loadBrand } from '../../../lib/brand';

export const brand = loadBrand('my-company');

Brand Properties

Colors

Property	Description
`primary`	Main accent color
`primaryLight`	Lighter variant for hover/highlight
`textDark`	Primary text color
`textMedium`	Secondary text color
`textLight`	Tertiary/muted text
`bgLight`	Light background
`bgDark`	Dark background
`bgOverlay`	Semi-transparent overlay
`divider`	Border/divider color
`shadow`	Shadow color (with alpha)

Fonts

Property	Description
`primary`	Main font for headings and body
`mono`	Monospace font for code

Typography

Define sizes and weights for each text style:

h1, h2, h3 - Heading styles
body - Body text
label - Labels and captions

Voice Settings

ElevenLabs

Property	Description
`voiceId`	ElevenLabs voice ID
`stability`	Voice stability (0-1, higher = more consistent)
`similarityBoost`	Voice similarity (0-1, higher = more similar to original)
`style`	Style exaggeration (0-1)
`model`	ElevenLabs model to use

Qwen3-TTS

Add an optional qwen3 section to voice.json for self-hosted TTS:

{
  "voiceId": "...",
  "settings": { ... },
  "qwen3": {
    "speaker": "Ryan",
    "language": "Auto",
    "instruct": "",
    "clone": {
      "refAudio": "assets/voice-reference.m4a",
      "refText": "Exact transcript of the reference audio."
    }
  }
}

Property	Description
`qwen3.speaker`	Built-in speaker name (Ryan, Aiden, Vivian, etc.)
`qwen3.language`	Language hint (Auto, English, Chinese, etc.)
`qwen3.instruct`	Emotion/style instruction (e.g., "Speak warmly")
`qwen3.clone.refAudio`	Path to reference audio, relative to brand directory
`qwen3.clone.refText`	Exact transcript of the reference audio

Both qwen3 and qwen3.clone are optional. Set clone to null or omit it to use built-in speakers only.

Setting Up Voice Cloning

Run /voice-clone for a guided workflow that:

Records or imports reference audio
Tests the clone quality
Saves the profile to your brand's voice.json

Once saved, use --brand to load it automatically:

uv run tools/voiceover.py --provider qwen3 --brand my-company --scene-dir public/audio/scenes --json