Real-Time Canvas and Alchemy

Lesson 2: Alchemy – Enhanced Generation Quality

Lesson Objectives

By the end of this lesson, students should be able to:

  • Enable and use Alchemy in generation settings
  • Understand what Alchemy improves compared to standard generation
  • Apply Alchemy to content types where its quality enhancement is most valuable
  • Balance Alchemy quality benefits against its higher token cost

Lesson Content

What Alchemy does.

Alchemy is Leonardo AI's enhanced generation pipeline – producing higher quality output than standard generation through additional processing steps that improve prompt adherence, image coherence, and visual quality. Alchemy-enabled generations typically show better understanding of complex prompts, more coherent subject rendering, and higher overall image quality at the same resolution.

Alchemy quality improvements.

Key Alchemy improvements: better prompt adherence – the generated image more accurately reflects the prompt's intent; higher compositional coherence – elements in the image relate more naturally; improved fine detail rendering – textures, materials, and lighting have more quality depth; and more realistic or appropriate subject rendering depending on model and content type.

Token cost of Alchemy.

Alchemy uses more tokens than standard generation – the enhanced processing pipeline is more resource-intensive. For production work where quality is the primary concern: Alchemy is worth the additional token cost. For exploration and iteration where quantity of variations matters more than maximum quality per image: standard generation produces more variations within the same token budget.

When to use Alchemy vs. standard.

Use Alchemy for: final production images where maximum quality is the priority; complex prompts that standard generation interprets inconsistently; content where subject coherence is critical (portraits, detailed product images); and client presentation quality images. Use standard generation for: exploration and iteration; large variation batches for selection; and content types where standard generation already produces acceptable quality.

Enroll to continue this lesson.

The preview above shows the lesson objectives and opening lesson content. Enroll to view the full lesson, complete the practice work, and take the lesson quiz.

Log In / Create Account
Back to Course