Luganda Text-to-Image Model for African Language Prompts

BULaMU Dream is a text-to-image model trained from scratch to generate images from prompts in Luganda, an African language. The model was developed by mwebazarick using modified scripts from the imagen-pytorch repository and was last updated in February 2026.

Use Cases

• Fine-tune the model on Luganda text prompts to generate culturally specific imagery.
• Evaluate the model's performance on African language prompts compared to multilingual models.
• Use the model's architecture as a baseline for training text-to-image models on other low-resource languages.

Strengths

• First text-to-image model trained from scratch for an African language, specifically Luganda.
• Model development is documented with technical details available in a referenced white paper.
• Last updated in February 2026, indicating recent maintenance.

Limitations

• Specific dataset details like training data size, image count, or column structure are unknown.
• Model performance and bias are untested without access to the underlying training dataset.
• Geographic and linguistic scope is limited to Luganda, which may not generalize to other African languages.

Provenance

Source

huggingface

Collection Method

Trained from scratch using modified scripts from lucidrains/imagen-pytorch.

Time Range

null

Freshness

Last updated 2026-02-01.

Geography

Focus on Luganda, an African language.