8h • BookNova.ai stuff
BookNova - Contructive feedback after creating 4 books
I’ve noticed a few recurring patterns across longer BookNova projects and wondered if some future tuning might help improve output quality even further:
- Chapter endings sometimes over-summarise emotional meaning after the scene has already demonstrated it. In many cases, the strongest moments are already present in the behaviour or dialogue, so a lighter ending style might create more impact.
- Certain atmospheric or behavioural words tend to recur frequently across long drafts (“watchful,” “careful,” “subtle,” “quiet,” etc.). A repetition-awareness layer across chapters could help reduce recursive vocabulary patterns.
- Emotionally intelligent characters can gradually drift into over-interpreting people’s feelings and motivations. The strongest passages often come from observable behaviour rather than direct emotional explanation.
- Supporting characters occasionally converge into similar conversational rhythms or observational styles over time. Stronger differentiation in speech cadence and perception styles could help ensemble scenes feel more distinct.
- Emotional intensity can become too evenly distributed across scenes, making everything feel equally significant. More tonal variation and genuinely neutral breathing-space moments might improve pacing and realism.
- Some of the strongest scenes are built around small behavioural contradictions rather than explicit exposition. The model seems especially good when tension emerges through gesture, interruption, silence, posture, or routine rather than direct explanation.
- Longer projects occasionally show continuity drift around timelines, established facts, or character details. Some kind of lightweight continuity memory across chapters and sequels could be extremely valuable.
- Dialogue is often strongest early in generation, but later chapters sometimes become slightly too polished or mutually explanatory. More conversational asymmetry, evasion, interruption, and incomplete responses could improve realism.
- Scenes sometimes continue slightly beyond their natural endpoint, especially after the emotional or narrative turn has already landed. Stronger scene-end awareness might tighten pacing automatically.
- The biggest strength of the newer models seems to be narrative coherence rather than decorative prose. The emotional continuity and structural consistency are noticeably improving, especially in longer manuscripts.
- Object-based storytelling works surprisingly well when objects carry emotional meaning rather than just plot relevance. The model seems particularly effective when recurring objects become emotional anchors across scenes.
- The most believable tension often comes from characters trying to remain socially functional while privately wanting incompatible things. BookNova is at its strongest when conflict stays restrained rather than overtly dramatic.
Honestly, the improvement in long-form coherence compared to earlier models is already very noticeable. The newer generations feel much better at sustaining emotional architecture across an entire novella rather than just producing isolated strong scenes.
1
0 comments
BookNova - Contructive feedback after creating 4 books
powered by
skool.com/digital-product-creators-club-9979
Create, promote, and sell digital products with AI. One keyword. One click. A complete product ready to sell.
Suggested communities
Powered by