EpisodeReel

Serial work: what each tool documents about continuity

One vendor, one column, one page

A reading is the smallest unit this file has: what a single vendor put in writing about a single step of serial production, what the sentence names, and what it declines to say. Readings exist because a column of twenty-five cells is read down, while a purchase is decided one cell at a time. As of 2026-09-22.

What a single reading is made ofA reading takes one vendor sentence and separates three things: the words as published, the noun or verb inside them, and the question the sentence declines to settle. The third part is the one a production discovers late.The sentenceQuoted as the vendor publisheditWords a reader can check against thepageGraded, not paraphrasedWhat it namesA noun, a verb, a number or astageSomething a buyer can ask about bynameOr an adjective, which is notWhat it leaves outThe part no page on this fileanswersA question for a sales callMarked, never guessed atOne cell of the registerWhy a reading is narrower than an entry and deeper than a table
Fig. 1 The same three questions are put to every reading, which is what makes sixty-four of them comparable down a column.

The register answers a wide question badly and a narrow question well. Anybody comparing the whole market wants the table; anybody who has already shortlisted two products wants the sentence, the noun inside it, and an honest account of what the sentence leaves for them to find out later. The readings are written for the second reader.

A reading is opened when the vendor's own sentence carries something to argue with: a noun, a number, a named stage, a documentation genre, or a specific refusal. Where a page offers only an adjective that three other pages also offer, no reading is opened, because four pages arguing with the same adjective would be four pages of the same argument.

Every reading carries the column it belongs to, the entry it belongs to, and the day the sentence was read. Nothing on a reading page is inferred from an interface, a demo reel or a support thread.

1On the cast

  • AI Dramas on the cast — Each character is saved with a name, a face and a voice style, then referenced by name in the script, so the resolution happens wherever the name appears.
  • Anijam on the cast — Faces are said to be locked across every episode, dialogue synced to lip movement and everything assembled on a timeline, all as actions the tool performs.
  • Arcloop on the cast — Face, outfit, personality and visual identity are listed as locked, with locations and style held scene to scene, and no object named.
  • character.app on the cast — Two capitalised objects, a recurring Cast and a World, are presented as the things the series rests on rather than as results the model achieves.
  • Crano AI on the cast — Stable appearance, personality and emotional presence are stated to hold across episodes, which is a claim about the outcome with no step behind it.
  • Dramily on the cast — The reusable assets described here include character dynamics rather than only appearances, which is a claim about relationships as well as faces.
  • Genra on the cast — Cast cards are said to hold a face and a voice so characters keep their look and speak in their own voice, scoped to every scene.
  • Google Flow on the cast — The help centre describes reusing the same images for a character from one clip to the next, which is a smaller promise than one about a season.
  • invideo AI on the cast — Character sheets and locations are said to be held in agent project memory, which names a place without describing what is kept there or for how long.
  • Kling AI on the cast — Characters defined as elements carry their appearance and their voice between generations, which puts audio inside the same object as picture.
  • LTX Studio on the cast — Elements are called reusable across projects rather than only within one, which turns a character from a project asset into a company one.
  • Luma Agents on the cast — Characters, products and brand elements can be defined, with the agent told to preserve them through generation, localisation and iteration.
  • Medeo on the cast — Character consistency is attributed to the underlying video model rather than to the product, which moves the answer outside the thing being bought.
  • Novi AI on the cast — The lock is placed at creation time and the span given is the entire project, which makes the moment of the lock the interesting part of the sentence.
  • OpenArt on the cast — Leads are said to stay the same across every episode down to face, wardrobe and chemistry: a specific promise, and still only a promise.
  • Pexo on the cast — Hyperframe is given a name and a job: a face, a wardrobe and a voice held across every episode, and the identity of a product or brand held in every shot.
  • PopShort AI on the cast — Characters, scenes and props are called reusable production assets, and the span is given in episode numbers rather than as an adjective.
  • SceneMixer on the cast — The sheet is made while the script is read, and its face and body crops travel into every segment using that character, so a missing sheet shows up later.
  • Showrunner Studio on the cast — Persistence is claimed for sets, characters and canon together, held in the product as the show's world, which is a wider object than a cast list.
  • Topview on the cast — Characters and scenes are described as built before video generation, which the page says helps improve visual continuity, with no artefact named.
  • TrendShort on the cast — Locking identity, wardrobe, face and emotional range appears as one item in a numbered workflow, which places the work without describing it.
  • Vidu on the cast — The API reference takes up to seven reference images per request and documents no stored character, so the filing of a cast sits with the caller.

2On the split into episodes

  • AI Drama Generator on the split — The narrative is said to be automatically structured into episodes, scenes, dialogue and beats, with automatic doing most of the work there.
  • Arcloop on the split — Recurring episodes with consistent characters, voices and worlds are given as what the product makes, answering the column with its own question.
  • character.app on the split — Cast plus World, story setup, scenes and film assembly are named as the stages, which describes the order of work instead of promising a result.
  • Crano AI on the split — Repeatable episode structures are described as something the user builds to publish on a schedule, which moves the work back to the buyer.
  • DramaSo on the split — A season of up to sixty episodes is stated to be plannable from one premise, which puts the input at a sentence and the output at a whole run.
  • Dramily on the split — Each project is stated to support serialized dramas of up to a hundred and twenty episodes, which is the highest published ceiling anywhere on this file.
  • Genra on the split — Structured episode outlines and scene breakdowns appear in the vendor's own blog post rather than on the product page, and the venue is recorded.
  • invideo AI on the split — An idea, a script or a show bible is described as becoming a season plan episode by episode, with the planning attributed to an agent and no step named.
  • LTX Studio on the split — Work is documented as organised per production with no episode structure, which is a usable answer rather than a silence, and a coherent design.
  • Luma Agents on the split — Characters, story arcs and episode boards, then bringing the pilot to life, is given as a use of the board: one episode short of a season.
  • Medeo on the split — One larger story becomes numbered parts with recap clips and next-episode teasers, described as a repeatable shape rather than a one-off.
  • Novi AI on the split — Serialized mini-dramas are stated to be creatable from scripts, novels or story outlines, with the input list precise and the split itself undescribed.
  • OpenArt on the split — A director feature is stated to structure a story into episodes and scenes so it becomes a serial vertical drama, which puts the work in a named place.
  • Pexo on the split — An open loop at the end of one cut is said to carry the cast into the next: continuity between neighbours rather than a plan for a run.
  • PopShort AI on the split — A script or a novel uploaded into a named canvas is stated to have characters, scenes and props extracted from it automatically, with no manual entry step.
  • SceneMixer on the split — One upload is analysed into characters, scenes, props and episode structure together, so the split is a by-product of reading rather than a separate run.
  • Showrunner Studio on the split — Writing a scene is stated to return an episode, and a show begun in any genre to be shaped into a season, which inverts the usual direction.
  • StoryShort on the split — The storyboard is described as built in a named order, hook then escalation then reveal then cliffhanger, which is a shape a writer can argue with.
  • TrendShort on the split — An existing script, an AI draft or a plot outline is stated to be broken into shots, beats and visual prompts, which is a finer division than an episode.

3On assembly

  • AI Dramas on assembly — A draft pass is stated to start at one credit per second and an episode cut from three, which prices rehearsal separately from delivery.
  • Anijam on assembly — Assembling everything on a timeline is stated as part of the flow, which names a specific surface rather than an abstract step.
  • character.app on assembly — Film assembly appears as the last of four named stages, which places the join inside the product rather than in an editor somebody has to open.
  • DramaSo on assembly — The stated output is a captioned vertical episode at 720P or 1080P, which names the deliverable and two resolutions in the same sentence.
  • Dramily on assembly — Scripts, assets, storyboards, prompts, previews, covers and export-ready episodes are listed as the outputs of a single workspace.
  • Genra on assembly — Export is stated as 1080p HD, which names a resolution while leaving the question of who joins the shots unanswered on the page.
  • SceneMixer on assembly — Episodes are stated to be composited from segments and billed at one credit per second of output, which turns the join into a line in a budget.
  • Showrunner Studio on assembly — Publishing straight from the product is named as a step, after which the audience is described as stepping into the world, which reaches past a file.
  • StoryShort on assembly — Clips in which the characters speak their lines are stated to be assembled into one vertical episode, putting dialogue and the join in one sentence.
  • TrendShort on assembly — The export target is stated as a 9:16 vertical cut with captions and AI voiceover, which bundles picture, text and audio into one deliverable.

4On the shape that ships

  • AI Drama Generator on the shape — The generator exposes an episode count, an aspect ratio and a language as inputs, which answers the question by handing it to the user.
  • AI Dramas on the shape — Clips are stated to generate in vertical 9:16 by default and to run four to fifteen seconds each, which puts a frame and a duration together.
  • Anijam on the shape — Every episode is stated to export vertical at 9:16, which is a property of the product rather than a setting a project can change.
  • DramaSo on the shape — The output is given as a captioned vertical episode at 720P or 1080P, the one cell here pairing a frame with two named resolutions.
  • Genra on the shape — Export is 1080p HD on the product page and vertical at the platform's highest resolution on the blog, which is two answers in two venues.
  • Google Flow on the shape — Aspect ratio, output count, model and generation length are listed as preferences chosen before generating, which puts the shape at clip scale.
  • Luma Agents on the shape — Extending a video's duration and reframing its aspect ratio are listed among the operations available, which changes the shape rather than setting it.
  • Novi AI on the shape — Short dramas are stated to run from under a minute to several minutes, with nothing anywhere on the page about the frame they are delivered in.
  • Pexo on the shape — Widescreen, vertical and square are all stated as native deliveries, up to 4K, with a length range from five seconds to two minutes.
  • PopShort AI on the shape — Video length is given only as an estimate that rises with the subscription tier, from about one or two minutes to about eight.
  • SceneMixer on the shape — Episodes default to widescreen with vertical available, and outlines are stated to be written with cliffhanger pacing, which is a claim about the writing.
  • Topview on the shape — Vertical is called the format usually suited to short-drama platforms, which tells a reader what the vendor thinks, not what the software does.
  • TrendShort on the shape — The export target is a vertical cut with captions and AI voiceover, which states a shape and bundles two audio and text decisions into it.

Readings are grouped by column rather than by vendor because that is the axis on which they disagree. A vendor's four readings are linked from each other and from its entry.

Also filed here: Production tools, Season, Stages, Contrasts, Terms, Questions, Learn, Data. What earns an entry is set out under how read.