本文へスキップ
noveltodrama

noveltodramaについて — 小説起点のパイプラインを作った理由

多くのAI動画ツールはクリップ止まり。私たちは原作から始めます。筋の順序を守り、配役を固定し、課金前に絵コンテを承認し、クレジットを明細化します。

この製品の目的

There are more finished novels than there are screenwriters to adapt them. A web-novel platform with a hundred thousand titles can option maybe a hundred a year; the rest sit in a catalogue earning nothing. Meanwhile short-drama channels need roughly a hundred episodes a year, in more than one language, in more than one aspect ratio, and cannot staff that cadence with human writers, casts and editors.

noveltodrama is built for the gap between those two facts: not "generate a clip from a prompt", but "take a manuscript you already own, keep its plot and its characters, and ship a season of 60–90 second episodes with one locked cast, dubbed into 12 languages, exported in three ratios". ≈ 430 credits per episode is the unit economics that makes the cadence possible at all.

It is deliberately narrow. It compresses prose; it does not improve plotting, so a weak middle act stays weak. It does not do stunts, crowd scenes, long-form, or a recognisable human lead across forty episodes. And it cannot tell you whether you hold the rights to what you paste in. Where it is the wrong tool, we say so on the 活用例ページ rather than in a refund thread.

5つの運用原則

Published prices, no opaque invoicing

Every render step has a fixed credit price on the pricing page. If a change makes a step more expensive, the price page changes before the pipeline does.

Failures cost a shot, not an episode

Shots render independently and refund automatically when no file is produced. A queue that dies at shot 14 of 15 should not cost you an episode.

Every shot traces back to a chapter

Shots store the chapter and beat they came from, so a reviewer can ask "where in the book is this?" and get an answer. Adaptation without traceability is just improvisation with a familiar name.

Say what the tool cannot do

Cast lock is stable, not identical. Stunts, crowd work and broadcast VFX are out of scope. Those limits are on the use-cases page, not buried in a changelog.

Disclosure is not optional

Exports carry metadata identifying the content as synthetic. Removing it to pass off AI video as live action is a terms violation.

コンプライアンスをやさしく説明

These are the four areas that actually get creators in trouble. None of them are edge cases.

Synthetic media labelling

Masters include a content-credential metadata block. Platforms that require AI disclosure (TikTok, YouTube, Meta) accept it directly; you still need to tick their disclosure box at upload.

Rights and source material

You must hold the rights to any novel, comic or script you paste in. We do not run a plagiarism check against published works, and a generated season is not a defence against an infringement claim.

Likeness and voice

Generated cast members are synthetic. Uploading a photo or voice sample of a real person to clone their likeness is prohibited, including for "internal tests".

Data handling

Manuscripts, chapters and renders are stored in your account region and are never used for model training. Payment credentials never reach our servers — WeChat and Alipay are handled by their own gateways, cards by Creem and Waffo Pancake as merchants of record, USDC by on-chain verification only.

Full text lives in the 利用規約, 許容される使用 and プライバシーポリシー.

運営者について

A small independent team. We build the pipeline, run our own test seasons with it, and publish the credit numbers from those runs. There is no sales team and no enterprise tier that unlocks features shown on this site — every plan runs the same renderer.

Product & pipeline

Chapter splitting, episode mapping, storyboard review, render queue

Localisation

Dubbing timing, subtitle packaging, RTL layout for Arabic

Payments & compliance

Five channels, sandbox/live switching, disclosure metadata

最初の1話は300クレジットから

短編の取り込み、10章の要約、1話分の絵コンテ作成に足ります。カード不要。

300 credits · No card required · Commercial rights on paid plans