Generation Modes: From Fastest to Finest

Overview

When creating a scribe on Coffeescribe, you choose a generation mode that controls how the AI writes your sections. The four modes range from fastest and cheapest to slowest and most coherent.

Lightning

Lightning generates all sections in parallel -- every section is written at the same time, with no cross-section context.

How it works

1. The AI receives the scribe structure (chapter titles, section titles, descriptions)
2. All sections are generated simultaneously, each working from the structure alone
3. The entire scribe can be ready in minutes

Best for

• Educational or reference scribes where each section stands on its own
• Quick first drafts to refine later
• Budget-conscious long scribes

Trade-offs

• Sections don't know what came before them, so there may be repetition
• Not ideal for fiction or content that needs a through-line

Lightning Medium

Lightning Medium runs chapters in parallel, but writes sections within each chapter one by one. Each section sees the previous sections from the same chapter.

How it works

1. All chapters start generating at the same time
2. Within each chapter, Section 1 is written first
3. Section 1 is sent as context when writing Section 2 in the same chapter
4. And so on -- but chapters don't see each other

Best for

• Scribes where sections within a chapter build on each other, but chapters are fairly independent
• A good balance of speed and intra-chapter coherence
• Non-fiction with structured arguments per chapter

Trade-offs

• Slightly slower and more expensive than Lightning
• No cross-chapter context -- Chapter 5 doesn't know what Chapter 3 said

Pour Over

Pour Over writes chapters one at a time, with each new chapter seeing all previous chapters. Within each chapter, sections run in parallel.

How it works

1. Chapter 1's sections are all generated in parallel
2. Once Chapter 1 is complete, all its content is sent as context for Chapter 2
3. Chapter 2's sections generate in parallel, with Chapter 1 as context
4. And so on, through every chapter in the scribe

Best for

• Scribes where chapters build on each other but sections within a chapter are independent
• Fiction and storytelling where plot develops across chapters
• A strong balance of coherence and reasonable speed

Trade-offs

• Slower than Lightning modes -- chapters must wait for previous chapters
• More expensive -- later chapters send more context tokens

Pour Over Slow Brew

Pour Over Slow Brew generates every section one at a time, each seeing everything that came before. Maximum coherence, slowest and most expensive.

How it works

1. The AI writes Section 1
2. Section 1 is sent as context when writing Section 2
3. Sections 1 and 2 are sent as context when writing Section 3
4. And so on, through every section in the scribe

Best for

• Fiction, memoirs, and narrative non-fiction where every detail matters
• Short scribes where quality matters most and cost is secondary
• Final polished versions where you want maximum coherence

Trade-offs

• Slowest mode -- sections are written one at a time
• Most expensive -- token usage grows with every section as the full scribe is sent as context

Cost Comparison

The token cost difference between modes depends on the number of chapters and the price of the model.

• Short scribes (5-8 chapters): The difference between all four modes is modest
• Long scribes (20+ chapters): Pour Over Slow Brew can cost 3-4x more than Lightning because later sections send the entire scribe as context
• Expensive models amplify the gap: A 25-chapter scribe with Claude Sonnet shows the biggest cost difference

If you are watching your token budget, Lightning is the most economical choice -- especially for longer scribes and premium models. Lightning Medium offers a good middle ground with only a small cost increase.

When to Use Each

Availability

Generation mode selection is available to all tiers in Americano, Cold Brew, and Espresso creation flows.

See Also

• Choosing an AI Model -- Model choice affects both quality and cost
• Token Costs by Model -- How model pricing and generation mode interact
• Cold Brew (Guided Creation) -- Step-by-step creation with full control