anixlynch/bossbaby-decaf

Bossbaby Decaf Dataset (Bchan Multimodal)

B-Tone voice samples + expression images for multimodal avatar synthesis

📦 Dataset Contents

Audio (Voice Samples)

• audio_voice_formal.m4a - Formal/professional tone
• audio_voice_informal.m4a - Casual/informal "B" tone
• audio_voice_seed.m4a - Reference seed voice

Images (Expression References)

Cartoon Style

• image_cartoon_grumpy.jpg - Grumpy expression
• image_cartoon_neutral.png - Neutral baseline
• image_cartoon_smiley.jpg - Happy/smile expression
• image_cartoon_sneaky.jpg - Mischievous expression

Real Photo Style

• image_real_angry.jpg - Angry expression
• image_real_happy.jpg - Happy expression
• image_real_kimono.jpeg - Cultural/formal attire
• image_real_professional.png - Professional setting

🎯 Use Cases

1. Voice Tone Classification - Train models to distinguish formal vs informal speech patterns
2. Multimodal Similarity - Match voice tone to facial expression
3. Avatar Animation - Generate talking head videos synced to B-Tone voice
4. Style Transfer - Cartoon ↔ real photo expression mapping
5. Emotion Recognition - Cross-modal emotion detection

🔧 Usage

from datasets import load_dataset

# Load dataset
ds = load_dataset("anixlynch/bossbaby-decaf")

# Access audio files
audio_formal = ds['audio_voice_formal.m4a']

# Access images
image_grumpy = ds['image_cartoon_grumpy.jpg']

🎬 Pipeline Integration

This dataset supports the Boss Baby multimodal pipeline:

1. B-Tone Detection → Voice tone classification
2. Expression Mapping → Match tone to facial expression
3. Animation Synthesis → AnimateDiff/Pika for motion
4. Voice Sync → Combine audio + animated visuals
5. Espresso Integration → Final talking avatar video

📊 Dataset Stats

• Total Files: 11 (3 audio + 8 images)
• Modalities: Audio, Image
• Language: English
• Character: Bchan (Boss Baby persona)
• License: CC-BY-4.0

🤝 Citation

If you use this dataset, please credit:

@dataset{bossbaby_decaf_2025,
  title={Bossbaby Decaf: Multimodal B-Tone Voice and Expression Dataset},
  author={anixlynch},
  year={2025},
  publisher={Hugging Face},
  howpublished={\url{https://huggingface.co/datasets/anixlynch/bossbaby-decaf}}
}

🔗 Related Projects

• Boss Baby Pipeline: Full avatar synthesis workflow
• B-Tone Classifier: Voice tone detection model
• Espresso Integration: Video output for Americanoverse

---

Created for multimodal avatar research and creative AI applications 🎭✨