Home › Methodology: how the seven criteria are scored

Methodology: how the seven criteria are scored

We score the ability to deliver a vertical short-drama series, not the look of a single clip. Seven criteria, 1–5 each, averaged into an overall score; any criterion without enough information is left blank (shown as —) rather than guessed.

Criterion531
Script & novel parsingWhole novel auto-split; reproducible appearance; scenes, props, hooksScript → storyboard, no splittingNo parsing; single prompts
Character consistencyCast library + references injected per segment + continuity rulesReference images / multi-subject reference, manualNo reference mechanism
Storyboard & shot controlAuto storyboard, per-shot editing, timing and cut controlCanvas / manual arrangementNo storyboard layer
Video generation qualityFrontier model; stable motion, physics, facesUsable with occasional artifactsClearly behind
Voice & audioControllable character voices with correct lip-sync attributionNative audio, not controllableNo audio
Episode compositingOne-click episode assemblyTimeline / canvas chainingExport segments, assemble by hand
Cost & free tierLow per-second price with a usable free tierMid-rangeExpensive or no free tier

Derived figures: stage coverage and engine strength

Stage coverage: of six production stages (parsing, reference art/consistency, storyboard, video, audio, compositing), a stage counts as covered when its criterion scores 3 or higher, giving 0–6. It measures how much of book → series a tool does without manual work; it is a mechanical roll-up of the scores, not a new judgment. Engine strength: the mean of video generation quality and voice & audio — how good each generated segment is. The positioning map on the home page uses exactly these two figures; every mean carries an "n/7 scored" badge and unscored criteria are excluded.

Sources and cadence

Facts come only from official pages (URLs and last-checked dates are listed at the end of every tool page); we re-verify monthly. The basis for each score is written in the tool page's "Basis" column and open to challenge.

What we don't do

  • No second-hand prices, no invented numbers for undisclosed facts.
  • No absolute "X is better" verdicts: comparison pages say "pick A if… / pick B if…".
  • No reviews of products in closed beta without a public entry point.