MemoryWorks: Grace

MemoryWorks: Grace is a portrait realism LoRA designed to create natural-looking female portraits with a candid selfie aesthetic. It emphasizes lifelike skin texture, flattering facial features, and the imperfect realism found in everyday photography rather than a polished studio look.

The model was trained on a diverse collection of female portraits spanning multiple skin tones and ethnic backgrounds. While the training set is limited, the LoRA is intended to generalize across a wide range of prompts during inference.

Unlike heavily stylized beauty models, MemoryWorks: Grace aims to produce portraits that feel authentic—capturing the look of spontaneous phone selfies, casual snapshots, and real-life photography.

✨ Best Use Cases

• Natural female portraits

• Casual phone selfies

• Candid photography

• Lifestyle photography

• Social media style portraits

• Amateur camera aesthetics

• Everyday realism

📝 Recommended Prompting

This LoRA performs best when using a prompt structure similar to the one it was trained on.

Recommended format:

candid_selfie_style, [subject], [hair], [skin tone], [clothing], [lighting (optional)], [background]

Example

candid_selfie_style, a young South Asian woman with long dark brown hair, medium brown skin, wearing a black top, soft window light, spacious indoor restaurant background

The anchor trigger:

candid_selfie_style

is strongly recommended when using MemoryWorks: Grace by itself, as it activates the intended selfie composition and portrait style.

🔀 Using with Other LoRAs

MemoryWorks: Grace also stacks well with other LoRAs.

When combining it with additional style or character LoRAs, prompts that reinforce photographic realism tend to produce the best results. Useful keywords include:

• real life

• candid

• raw

• amateur photography

• natural lighting

• smartphone photo

• casual portrait

• realistic skin

Although it can work with many photography styles, the LoRA has a stronger bias toward selfie-style compositions, since that was the primary focus during training.

⚙️ Recommended Settings

LoRA Strength: 1.0 (Recommended)

The model was trained with a standard strength distribution and is intended to be used at full weight.

Suggested range:

• 0.6–0.8 → Subtle influence

• 1.0 → Recommended

• Above 1.0 → Stronger stylistic effect (results may vary depending on your base model and LoRA stack)

💡 Tips

• Works best with realistic checkpoints.

• Use descriptive details for hair, clothing, lighting, and background to achieve more natural results.

• The LoRA is designed to enhance realism while preserving the identity and characteristics described in your prompt.

If you create something using MemoryWorks: Grace, I'd love to see your results. Enjoy creating!