BYOK Cost Control: Budget a Full Novel with API Keys

Bring-your-own-key pricing is honest: you pay for what you generate. It is also easy to burn thirty dollars on unplanned regenerations. Budget like a producer, not like a chat user.

Rough math

An 80,000-word novel might be 35–45 chapters at 1,800–2,500 words each. Each chapter generation includes: bible context, prior summary, outline chunk, and output. Expect 3k–8k input tokens + 2k–4k output tokens per chapter depending on context size and model.

Multiply by your provider's per-million rates. Add 30% buffer for rewrites and outline iterations.

Task-based model choice

• Outlines and beat sheets — capable mid-tier model; structure matters more than prose
• Final prose — your best model for voice
• Line polish — smaller/faster model with tight prompts

Do not use flagship models to regenerate entire chapters because one paragraph failed—scope prompts to the broken scene.

Cost control habits

1. Freeze the bible length; trim instead of appending
2. Accept "good enough" first drafts; revise in diff editor
3. Log each job's token estimate in a spreadsheet
4. Set provider billing alerts at $10 / $25 / $50

Compare providers on our AI model comparison page. Entangled Text's free sample book lets you test workflow before you spend on keys.