Movie Kick User Guide v5.1
A tool for designing feature-film and drama screenplays with AI · As of 2026.08
Movie Kick is a tool for completing feature-film and drama screenplays step by step with AI. The AI does not make creative decisions for you; it proposes options at each stage and helps you develop the work in detail.
The entire process has 7 stages — 1. Idea → 2. Synopsis → 3. Characters → 4. Plot → 5. Location → 6. Scene List → 7. Final Draft. If this is your first project, follow them in order; whenever necessary, use the sidebar to return to an earlier stage and edit it directly.
How is my work saved?
Every project is auto-saved to the browser you work in (IndexedDB). The save status (saving / saved) is shown in real time at the top of the sidebar while you work, so there's no need to press a separate save button. Your login account is used only for authentication.
The primary storage has changed from the cloud to browser-local (IndexedDB). It's faster and more reliable, but your data is kept only in this browser. To continue working on another device or to keep your work safe, we recommend the backup options below.
Automatic Folder Backup (Chrome · Edge) — Once you designate a backup folder in the "Open" modal on the home screen, your project is automatically saved to that folder as .mk.json (about every 15 seconds). If you choose a sync folder from Dropbox · Google Drive · iCloud, you can pick up the same work on another device with no extra setup.
For a one-off backup, you can always pull everything into a single file via Final Draft stage → Project Export (.mk.json).
Movie Kick is optimized for Chrome · Edge on desktop. Some features, such as automatic backup-folder saving, work only in these browsers and may be limited on Safari · Firefox or mobile. For a stable experience, we recommend desktop Chrome or Edge.
Deleting projects and the Trash
Deleting a project doesn't make it vanish immediately — it moves to the Trash (soft delete). From the Trash located below the Library button in the left sidebar, you can restore deleted projects or permanently delete them all at once with "Empty." When the Trash contains items, a count badge appears on the button.
Renaming a project
You can rename a project at any time. Even after taking an AI title suggestion at the Final Draft stage, just edit the name in the project list on the home screen.
Why we nudge you to back up
Once your work has piled up while it lives only in this browser, a backup notice appears once. From there you can turn on automatic backup with "Choose folder", or leave a single copy with "Save to file". In browsers that don't support automatic backup, only the file option is offered.
The sidebar shows your current backup state (Auto-backup folder / Choose a backup folder).
Working in dark mode
The account menu at the bottom of the sidebar offers "System light" and "System dark". The work screens, graphs and Boundary analysis all follow along.
If you open the same project in two places
Even with the same project open in two tabs or two windows, whatever was saved first is never silently overwritten. If a conflict is detected, you're told.
When a new version ships
When the app is redeployed, an "Update available" notice appears. It sits where it won't cover your work, and while you're idle the app refreshes itself to the latest version. You can see what changed under "What's New" in the account menu.
If a single item — one scene, one synopsis — exceeds the per-item save limit (about 900KB), that item isn't saved and you get a notice. The notice appears once when the state changes (it no longer repeats on every autosave), and autosave and manual Ctrl+S both tell you the same thing — something that failed to save will never show as "Saved." When this happens, split the scene.
How to sign in
- Sign up with email/password, then log in
- Quick sign-in with your Google account
If you forget your password you can reset it yourself from the sign-in screen — no waiting for an administrator. There is a short cooldown so reset mails aren't sent repeatedly.
One account, one device
One account can't be signed in on two devices at once. When you sign in on a new device, we first tell you another device is in use and ask whether to continue. The device being signed out is told why it was signed out.
Even when a device is pushed out, your writing stays on it — you can keep reading and editing; only AI generation pauses.
"Sign-in history" in the account menu shows which devices signed in recently, and "Sign out all other devices" clears them in one go.
Plans — Free and Pro
Plans are now two tiers: Free + Pro. The Education plan has been discontinued — its features were identical to Pro, so there was no reason to keep it as a separate tier. Automatic enrollment by school email domain is gone as well. Department and institution rollouts are handled as Pro group quotes.
| Plan | Key Offering |
|---|---|
| Free | 5 days on which you used AI generation · bulk scene-list generation once in total · unlimited projects |
| Pro | Everything, unlimited · Final Draft (FDX) export |
Actual amounts depend on where you are and can change, so this guide does not list them. See Pricing on the landing page for what applies right now — and the amount shown at checkout is always what you are charged.
Movie Kick prices according to purchasing power where you live. Charging everyone the same looks fair, but in practice it means charging more the less a region earns. Korea is the baseline.
Your billing region is set by the country you connect from, independent of the language you use the app in. Whether you work in Korean or English, pricing follows where you are. The EU and EEA share a single price across all member states, shown inclusive of VAT.
- 5 days on which you used AI generation — not 5 calendar days from signup. Only days you actually generated something count, so if you work weekends only, all 5 days are still yours.
- Bulk scene-list generation once in total — per account, not per project.
Project count, work size, plot-structure experiments and the trope engine are not restricted. And after the 5 days run out, reading, editing by hand and PDF/HTML export stay open forever — the only thing that stops is AI generation.
"Subscription" in the sidebar account menu shows the plan you're on and the planned price. Prices are shown with the term they cover, and the displayed figure is the total including VAT.
Billing opens on 9 September 2026. Until then everything is open on Free. Once billing opens, plan changes, cancellation and receipts are handled from this screen.
A Pro subscription does not include AI token costs. Every tier uses your own OpenRouter key. That also means that on Free, once you add credit to OpenRouter you can use paid-model features like bulk scene-list generation and Boundary analysis exactly as they are — it is a credit condition, not a tier restriction.
Movie Kick uses a BYOK (Bring Your Own Key) approach. To use the AI generation features, you need an OpenRouter API key.
The connection prompt appears once. You can skip it with "Look around first" and connect later from the sidebar whenever you like.
The key is never uploaded to our servers — it is stored inside this browser only. On another computer, press the connect button once more; the key on this device stays where it is.
Keys are kept per account, so you don't have to enter one again after signing out and back in. The button in the notice right after connecting opens the OpenRouter settings page for the key you just connected, where you can set a spend limit or delete keys you no longer use.
The free model you can try without adding credit is NVIDIA's Nemotron. In exchange for being free, what you type may be used to train the model. We tell you this before you connect.
The API key is stored only inside your browser. So if your browser is set to clear site data on exit, or a cleaner extension is installed, the key is wiped and you have to enter it again next time.
When the "Install app" prompt appears on the home screen, take it. Installing makes the browser treat this data as protected and exclude it from cleanup — no more re-entering the key. You don't have to hunt through browser settings; one button does it.
Available models
Choose from the model selector at the top of each stage. Switch between them to match the nature of your task.
Prices are per 1 million (1M) tokens for input / output (USD). The table is ordered by the app’s blended-cost measure, highest first. Source: OpenRouter official pricing (as of 2026-08).
| Model | Input ($/1M) | Output ($/1M) | Strengths | Category |
|---|---|---|---|---|
| Claude Opus 4.7 | $5.00 | $25.00 | Complex narratives and character depth, 128K output | Premium |
| Gemini 3.1 Pro | $2.00 | $12.00 | Top of the latest benchmarks, multimodal | Premium |
| Claude Sonnet 5 | $2.00 | $10.00 | Distinct voices per character, dialogue depth | Premium |
| Gemini 2.5 Pro | $1.25 | $10.00 | Stable and high-performing, 1M context | Standard |
| Gemini 3.5 Flash | $1.50 | $9.00 | Fastest inference, ideal for chat and quick iteration (default) | Standard |
| GPT-5.6 Terra | $1.00 | $6.00 | Strong at plot and logical structure, 1M context | Premium |
| Kimi K2.6 | $0.589 | $2.48 | Top Chinese model on creative-writing benchmarks, long-form narrative | Chinese creative |
| MiniMax M3 | $0.30 | $1.20 | 1M context · unbeatable value | Chinese creative |
| DeepSeek V4 Pro | $0.43 | $0.87 | Great value, useful for first drafts | Standard |
| Nemotron 3 Ultra | $0 | $0 | Try it without adding credit · strong instruction-following (IFBench 81.7) · official KO/JA support · 1M context | Free |
Claude Sonnet 5 is on OpenRouter introductory pricing ($2/$10) until 2026-08-31, after which it returns to list price ($3/$15). Model prices can change, so also check the current OpenRouter listing before paying.
Chinese creative models such as Kimi and MiniMax are tuned for long-form prose and can be selected only while writing scene scripts. Other stages show only the models suited to that task.
The one exception is bulk scene-list generation, which has to receive dozens to 100+ scenes as a single JSON payload, so it only uses models whose format reliability has been verified (see Stage 6).
Only one free model, Nemotron 3 Ultra, is shown in the picker (to reduce choice fatigue). If it expires or goes down, the app automatically falls back to another free model so your work isn't interrupted. It lets you try the full workflow without adding credit; paid top-tier models are more consistent in prose quality and structural completeness.
Fit-based recommendations (★ marker)
While you work, Movie Kick automatically learns which model and expressive-variety setting leads you to edit the results the least. Once enough data accumulates, a ★ recommended marker appears in the model selector, and you can apply it instantly with the “Switch to these” button.
A recommendation is calculated only after at least 10 tracked samples have accumulated in version history. It appears in the Scene Editor as a ★ marker in the model selector and a “Switch to these” button. Until then, use the global default, Gemini 3.5 Flash. Bulk Scene List generation does not support that model and automatically falls back to Gemini 2.5 Pro.
This is the stage where you set your story's starting point. Choose one of three paths to match the way you work.
Adjust the tone with sliders
Tune the mood directly with sliders for dark/light, fast/slow, commercial appeal, and more. Enter it together with a one-line idea, and it expands into a synopsis.
Build structure through dialogue with AI
The design interview asks, one by one, about the protagonist, the person standing opposite, want and need, the trap the events set, and the turning points. As your answers accumulate, the story's skeleton takes shape.
Pick and combine the events you love
Select 1–15 elements that appeal to you from the narrative-element library. The AI analyzes their compatibility and conflicts and uses them as the seed for your synopsis.
The design interview — including who stands opposite your protagonist
The interview doesn't stop at the protagonist. It builds the person who stands opposite them with you. You choose one of three:
- The one standing opposite — the side that blocks the protagonist
- The one who listens — the side the protagonist confides in
- The one with their own story — a co-lead. Here you also decide who carries the climax
The figure who came to you first doesn't have to be the protagonist. If you bring up the second figure first, the interview starts there.
The interview gives a turn back at each layer and reaches for films as examples as it goes — not one example at the start, but a fitting work each time the question changes.
What you settle in the interview, and the films you touched on along the way, gather into the closing summary and carry through synopsis, characters, plot and script — you don't have to explain them again downstream.
Discover suits you when you start from a vague image or mood; Design when you have a faint sense of characters and conflict; Combine when a particular scene from a film you love won't leave your head. Whichever path you choose, everything merges into the same structure from the next stage on.
The Synopsis stage is where you refine and finalize the story itself. Edit the text directly or regenerate it with AI.
Regenerating the synopsis does not automatically overwrite your Characters. Design and refine them in the next stage.
What to do
- Generate a synopsis from the idea you created in Stage 1.
- Check that the causal chain of events and the ending read as one complete story, then refine the text yourself.
- Once the content is settled, use the sidebar to move to the next Characters stage.
Regenerate on the spot and compare
You can regenerate right here on the synopsis screen. No need to go back to stage 1 — draw several versions, compare them, and keep the one you like. Earlier versions stay in Version History.
You can pick the model that regenerates and the model used for the next stage separately — a structurally strong model for the synopsis, something else downstream.
The synopsis box now stretches to the full height of the window, so a long draft doesn't have to be scrolled through a narrow slot.
Synopsis and plot are steered away from translationese and stock closing lines at the moment they are generated, not polished afterwards. Overlong sentences and trailing, unfinished endings are filtered at the same point.
The synopsis is the source of truth for the story text. If you revise it after creating Characters, return to the Characters stage and select “See what changed”. Only this synopsis-change flow lets you choose which proposals to apply; it does not overwrite existing Characters automatically.
Create a set of Characters from the finalized synopsis, then design their inner lives and relationships.
What to do
- Select “Create characters”. The AI generates Characters and immediately applies all unique new results as a set.
- Build on goal, flaw, and arc with contradiction, wound, and true need so each Character’s choices become clear.
- Select a Character in the relationship graph to reveal counterpart options below the graph.
- A counterpart with no relationship yet appears as a dashed-border option. Select it to begin a new relationship design.
- Graph edges can show relationship information you have designed and saved. When no relationship data exists, the graph may also draw solid faction-based supporting connections as a reference. To find an undefined counterpart, use the dashed-border options below the graph, not the line style.
- Edit the AI draft’s bilateral viewpoints, voices, tension, and defining actions, then save it.
A relationship draft becomes part of the project only when you select Save.
For each relationship you can set who speaks casually and who speaks formally, including relationships where only one side talks down. Once set, the register doesn't flip from scene to scene as the script is written.
In the same place you can write down what is taut between them and the decisive act. The temperature of scenes where the two meet then matches what you designed.
Role and structure terms carry explanation tooltips. Hover a term on the Characters, Treatment or scene script screens and the meaning and an example appear right away.
To check the synopsis again
The Characters screen does not display a second, read-only copy of the synopsis. Select “Go read the synopsis” to return to the original Synopsis stage.
Enriching and managing Characters
Edit each card’s name, age, gender, occupation, and description directly. You can also choose a separate model on this screen for generating and enriching Characters. Deleted Characters can be restored from the Trash together with their relationships.
Once the Characters and relationships are ready, move to the next Plot stage.
This is the stage where you design the whole story as a beat structure. Rather than just two options, you choose from 5 proven structures to fit the nature of your work.
After analyzing over 500 reference films, we found that these five structures cover roughly 95% of all works. Two structures alone (the Hero's Journey and Save the Cat) can't accommodate narratives like tragedy, relationship-centered stories, temporal variation, or situation-driven momentum. By expanding to 5 options, we cover the full diversity of genre, ending shape, and narrative momentum.
The five plot structures
Each structure has a defined per-beat weight (%), and these weights are used directly to allocate the number of scenes. For example, the "Finale" of STC15 is 20%, so in a 60-scene work about 12 scenes are allotted to the finale.
Want vs Need — a structure where the protagonist's external desire and inner need collide and transform. The standard skeleton of genre film, integrating theme, character, and event. Blake Snyder (2005).
A parallel structure of external adventure × inner growth — an archetypal journey that departs from the ordinary world, passes through ordeals, and returns as a transformed self. Campbell (1949) → Vogler (1992) organized it into 12 stages for film.
A structure where the pressure of an external situation forces the protagonist into motion. It holds up even with little character change, and is the only structure that can accommodate a tragic ending. The "situation arc" is the core driving force.
A structure driven by the change in a relationship between two people. Instead of a single protagonist, two people transform each other. Two-way change + flaws on both sides are required conditions.
A structure where subtle variations within repetition accumulate to create meaning. It's not events but time itself that drives the story. Within 6 phases (Ritual Established → Small Variation → Cracks Accumulate → Climax → Resolution → Continuation), the writer declares the micro-units directly.
P1 Ritual Established · P2 Small Variation · P3 Cracks Accumulate · P4 Climax · P5 Resolution · P6 Continuation
If external adventure takes up a large share, choose HJ12; if the character's inner change is central, STC15; if tragedy or external pressure is the driving force, SD11; if a relationship between two people is the center, REL10; if repetition and the accumulation of time is the theme, TV6. Click a structure's name on screen to open a detailed modal with beat-by-beat descriptions and reference films.
When the AI recommends a plot design, it shows its reasoning alongside — which qualities of your story fit this structure — so you can judge before choosing. You are free to ignore the recommendation and pick another structure.
Editing beats
The beats of your chosen structure automatically expand into beat cards. You can edit the content directly on each card, or make natural-language requests through the Kick Assist chat on the right, like "Make beat 3 more suspenseful." Beat weights are preserved, ensuring consistency in scene-count allocation.
Hover a beat and a tooltip shows the job that beat is doing along with its content — on both the Plot and Treatment screens. Beat bodies no longer repeat their own title.
Expansion modes — three rhythms for turning a beat into scenes
You choose this from the "Expansion" dropdown on each beat card. It sets how many scenes, and in what manner, this single beat unfolds into. If you change nothing, it runs as Sequence.
Builds up across several scenes — the natural flow of setup → escalation → payoff. This is the standard that fits most beats.
Example — "The protagonist settles into a new department" → first day / friction with a colleague / a small win / an unexpected mistake, unfolding across four scenes.
Resolves this beat in a single scene. It shows only the decisive moment (discovery, betrayal, reversal), while the tension-building work is handled in advance by the immediately preceding beat. This keeps the punch from being diluted across multiple scenes.
Example — "Witnessing a spouse's affair" → the single scene of opening the door. The suspicion and clues were already laid down in the preceding stretch.
Links the scenes strictly by "therefore → but". Each scene must be the result of the one before it; a flat "and then… and then…" list of events is forbidden. Use it for a short connective stretch you pass through quickly and logically, rather than building a big arc.
Example — "Shaking the pursuit and reaching a hideout" → dodging a checkpoint (therefore) into a back alley → a dead end (but) into the sewer → swept by the current (therefore) arriving at an abandoned factory. Passing through fast, cause to effect.
Based on plot signals (whether it's a turning point, the size of the inner shift, the beat's length), the app shows a "Suggested: ___" chip. It is never enforced, though — the final decision is yours. Feel free to ignore the chip and pick any mode you want.
Treatment — reading beats as one stretch
Once the plot exists, the Treatment tab opens. There, a few beats are grouped into one sequence — roughly two beats per sequence. You don't build sequences yourself; they are derived from the plot, and a grouping boundary only ever falls on a beat boundary. The plot is always the original.
Each sequence carries three things.
- Question — the one question that pulls the whole sequence along. It has to be answerable yes or no, and the answer arrives as the sequence ends.
- At Stake — how what's on the line changes between entering and leaving ("calm → past the point of no return"). Leaving must weigh more than entering. In a temporal-variation (TV6) structure, Accumulated takes its place.
- Prose — narrative writing that fills the gaps between beats. When the scene list is built, this prose stands in for the beat text.
You don't have to regenerate a whole sequence's prose. You can pick just the paragraph you dislike and rewrite it. Prose is written to flow from one sequence into the next, and each sequence ends where you want to know what happens next.
The default is two beats per sequence, but you can adjust the density with Fewer · Normal · More. How many it splits into shows up in the sequence names — you don't have to press it and then count.
Every beat sits in a structural position — some are placed to flip the situation, others to pass through lightly. That job is now sent along with the story when a sequence is written, so the question comes out as the one those jobs add up to. In works where you added beats yourself or reordered them, the positions can't be matched with certainty, so writing falls back to the story alone as before.
Treatment and Boundary are connected
Boundary analysis doesn't read beats alone — it also reads the question, what's at stake and the weight you settled in the Treatment tab. "At Stake" is the same story the Boundary measures, told in different words, so without it the two screens can tell you different things.
- Curves are still drawn per beat. No steps appear at sequence boundaries.
- Editing a sequence marks the Boundary analysis outdated, so you know it is time to run it again.
- Polishing only the prose does not — prose is not used by Boundary analysis, so you never have to repeat an analysis that would come out the same.
So the natural order is Plot → Treatment → Boundary. You can skip the Treatment and go straight to Boundary, but then it judges without anything to go on for "At Stake".
17-Boundary analysis
Once the plot is generated, the AI analyzes the conflict structure of your story using a 17-boundary framework (four domains: Self, Bond, Order, Fate). It identifies tensions between opposing poles — like "Truth vs Lie" or "Greatness vs Ordinariness" — and visualizes them as a graph across beats.
From the analysis, 1 primary boundary + 3 secondary boundaries are automatically selected.
- Primary boundary — the core axis of conflict running through the entire story. A boundary question the audience will ask themselves is presented along with it.
- Secondary boundaries — 3 additional axes of conflict that complement the primary one
The Boundary graph — reading by polarity
The intensity (0–100) of the 4 boundaries is plotted as a line graph across beats. What up and down mean depends on each boundary's polarity — for value-type boundaries, up (100) = resolution · growth; for threshold-type boundaries, the center is healthy tension while both extremes are catastrophe; for complementary-type boundaries, up/down simply indicate which pole dominates (not good vs bad). The points where a line crosses the baseline (Y=50) are clues to twists and climaxes.
This graph lets you check at a glance whether your tension structure is too monotonous and whether multiple conflicts erupt simultaneously at the climax.
Tab layout
This stage is run across four tabs. Left to right is the working order.
- Plot Design — structure selection and design options
- Plot — edit the text of each beat
- Treatment — group beats into sequences and read them as prose (opens once the plot exists)
- Boundary — the 17-boundary analysis graph
Trope (narrative element) work has moved to the Combine path in Stage 1.
Once the plot exists, pressing "Next" takes you through Treatment, then Boundary, rather than straight to Space Design. Picking locations the moment the structure is up would send an unverified story downstream.
Design locations as devices that drive events and relationships, not merely as backdrops. Define each place’s story purpose, mood, and related Characters to build a location palette for scene generation in the next stage.
What to do
- Choose the location strategy that fits the work: Genre Space, Everyday Space, or Displaced Space.
- Generate a location palette from AI proposals, or add locations yourself with "+ Add manually".
- Read the seven-identity strip at the top of the list to see which slots are filled and which are empty.
- Edit each location’s name, description, category, mood, related Characters, appearance frequency and shooting zone to fit the work.
- Pick the locations you'll use ("Select All" / "Deselect All") and move on.
The strip on top — the lens behind this palette
Seven space identities stretch across one row above the location list. You can see at a glance how many sit in each slot and which are empty, and clicking a slot shows only spaces of that kind. Return to everything with "Show all".
Click an empty slot and we fill just that gap. The old "Generate More" made similar spaces again without knowing what already existed; now only the missing identities get filled.
After "Select All" you can undo it from the same spot with "Deselect All", and spaces you added yourself stay untouched.
Not how many spaces — how many places you'll shoot in
How many spaces to make is decided by the story. A journey-shaped story can go past 40 and that is fine. There is no cap and no warning on the number of spaces.
Instead we show how many places those spaces will actually be shot in (shooting zones) — pull several scenes out of one building and a large number of spaces still means few moves. If the zone count feels high, don't cut spaces: group spaces from different zones into the same zone.
Zones can be edited directly on a location card (e.g. "downtown district").
Seven location roles
| Category | Purpose |
|---|---|
| Character Origin | Reveals the protagonist’s identity and past |
| Conflict Arena | Hosts the story’s central clashes |
| Emotional Refuge | Gives a Character room to catch their breath |
| Power Center | Where domination and control operate |
| World Expansion | Widens the scope of the story world |
| Relational Space | Where a relationship between two Characters forms or changes |
| Antagonist's Domain | The antagonist force’s stronghold or controlled territory |
Answer the AI’s questions to clarify the place’s atmosphere and dramatic purpose. Every location card can be edited directly, and the completed location palette is used in the next Scene List stage.
This is the stage where you'll spend the most time. You manage the overall structure in the Scene List and write each scene's actual script with AI in the Scene Editor.
Four tabs
| Tab | Role |
|---|---|
| Scene Design | Compare scene intentions for each beat and choose the direction that fits the work |
| Scene List | View scenes per beat as cards, reorder them by dragging, or enter the Scene Editor with the “Write” button |
| Plot Analysis | A creation and design feedback view that runs automatically once all scenes of a beat are written |
| Preview | Review the entire screenplay manuscript at once |
Bulk scene-list generation — 3 methods
The Scene List doesn't just fill in blanks — it's generated after automatically allocating the number of scenes according to each beat's weight. Choose one of three paths depending on the nature of your work.
Intent-recommendation auto
The AI proposes a scene-intent card for each beat (e.g., "This beat must show that the first response doesn't work"). Pick the card you like, and the Scene List is generated in bulk according to that intent. The most systematic path.
Chat-based writing
Describe a scene in natural language in Kick Assist on the right — like "This scene is about ~" — and the AI reflects it in that scene. Good for partial touch-ups.
Add by hand
Insert a blank scene yourself with the + Add Scene button. The beat position is sorted automatically. For when you need full manual control.
If you add or remove beats and then generate scenes, the scene list assigns them against the sequence spans as they are now. No more scenes built to the old spans with numbers that no longer line up.
If a scene runs long or two events are tangled in it, just ask the chat to split it. The app does the cutting — and the original script isn't thrown away; it carries over to both halves.
The rail that shows where you are
A position and progress rail runs down the right side of the scene list. You can see where in the whole you're looking and how far the script is written, and click the rail to jump straight there. Past 100 scenes you no longer have to hunt by scrolling.
The Scene List must be generated systematically for "plant and harvest" (setup and payoff) to work accurately. If scenes are scattered at random, it's hard to track where a seed is planted and where it's harvested. Intent-recommendation generation distributes scenes according to each beat's weight and role, so the Library's Seed Map works correctly afterward.
Choosing a generation model — you can use Claude too
From the model selector at the top of the scene-list generation screen, you can choose exactly which AI builds your scenes. Claude models are among the options.
- Gemini 2.5 Pro (Scene List fallback) — The global default, Gemini 3.5 Flash, is not supported for bulk Scene List generation, so the app selects this model instead. Its format remains stable when generating dozens to 100 scenes at once.
- Claude Opus 4.7 · Sonnet 5 — For when you want a literary, auteur grain with living character texture and narrative density. Internally it splits work by beat and generates in parallel, so it completes reliably even at large volumes.
- GPT models — JSON output can be enforced, so the list comes back fast in a single call.
DeepSeek and the free models are excluded from bulk Scene List generation because their reliability with large JSON payloads has not been verified. Kimi and MiniMax are available only for scene script writing, so they do not appear in bulk Scene List generation or other stages. Gemini 3.5 Flash is also excluded from bulk generation; Gemini 2.5 Pro is selected as the fallback.
A failed run doesn't take the scenes with it
If scene-list generation breaks off partway or leaves gaps, the scenes already created stay. You don't have to start over.
You can fill in just the empty slots — filled scenes are left alone and only the gaps are regenerated. Scenes you already edited by hand are never overwritten.
Non-destructive scene-list versions
Even if you go back to the design and pull a different version of the Scene List, the existing list isn't lost — it's kept as a version. Scene scripts are preserved along with it, so switching versions restores the work from that point exactly as it was.
Version chips appear below the "Refine Scene List" button (once there are 2 or more). Chip actions:
- Click — switch to that version (whatever you're currently editing is auto-captured into the active version → prevents loss)
- Double-click — edit the version name inline
- × — delete the version
Undo scene deletion
When you delete a scene card, a "Scene deleted · Undo" toast appears at the bottom of the screen for 5 seconds. Click the toast and the deleted scene is restored to its original position and original scene number. You can instantly bring back a scene you removed by mistake.
Before it goes, you're told what disappears with it — a seed planted in that scene, a bit part who appears nowhere else — so you can decide with that in view.
Scene Editor — Settings
The left panel of the Scene Editor is organized into two tabs: Settings / Characters. Key items on the Settings tab:
| Setting | Details |
|---|---|
| Scene length | Set as screen time (15s–5min) — an approximate character count is shown alongside |
| Stage direction vs Dialogue ratio | Adjust with a slider — stated in words as well as a number (scenes with a character alone start at 0% dialogue) |
| Dialogue length | Short / Medium / Narration |
| Tone | Light / Heavy / Neutral / Light→Heavy / Heavy→Light |
| Dialogue pattern | Pick one of the 9 below |
| Creativity (Temperature) | Lower is more stable, higher is more experimental |
| Custom instructions | Anything you want to ask of this scene only — the field tells you what it takes, with examples |
The settings block is shorter now, so the generated script isn't pushed off the bottom of the screen. Scene cards no longer print their purpose twice.
The "Style (Dry · Natural · Poetic)" field is gone — picking it never reached the output, so it was removed. Prose style is set through the dialogue pattern and Style Preferences instead.
The scene length you set is a target, not a quota to pad out. A scene with little event won't be stretched to fill the time. And if you set it short, the script won't get cut off mid-sentence.
9 dialogue patterns
You can choose your scene's dialogue style from 9 patterns.
| Pattern | Characteristics |
|---|---|
| Subtext | The gap between what's said and what's meant — default |
| Power Game | Every line is domination, submission, or resistance |
| Verbal Acrobatics | Witty intellectual exchange |
| Silence | The nonverbal and the still carry the meaning |
| Naturalistic | Transcribes the rhythm of everyday life as is |
| Ritualistic | Conventional, repetitive patterns build tension |
| Pop | Short, punchy lines of genre convention |
| Poetic · Metaphorical | Based on sensory imagery |
| Information War | Who knows what drives the conversation |
Per-Character acting direction — it belongs to the Character, and you switch it off per scene
You can specify acting direction (speech style, reaction speed, mannerisms, etc.) for each Character. You can add it from either the card in the Character stage or the Scene Editor — wherever you enter it, it belongs to the Character and shows up on the other side immediately, so a note that occurs to you mid-script can go in right there.
Directions are on by default in every scene the character appears in. To drop one in a particular scene, uncheck it there — only the switched-off state is recorded on that scene, so other scenes are unaffected, and directions you add later still apply to existing scenes automatically.
The scene editor expands the characters in that scene and collapses the rest. Even if a scene's character list is empty or a name doesn't match exactly, the directions don't disappear.
So characters aren't buried under events
Scripts that only push events along, leaving characters as scenery, are steered away from. A character's flaw, contradiction and wound are made to surface as behaviour inside the scene.
And that surfacing is distributed across the whole work — a character isn't fully explained in one scene and then reduced to errands in the rest.
The dialogue tone you set, and the casual/formal register you gave each relationship, now actually reach the script.
You don't pick a script layout
The Korean/Hollywood selector is gone. The layout follows where you write — in Korea and Japan your script uses A4 and the Korean layout; in English it uses US Letter, 12pt Courier and industry-standard indents. Character names sit centred in the text column in the Korean layout.
The scene editor and the exported file use the same layout — what you saw while writing is what you download.
Recording bit parts that show up while writing
Writing a script turns up people you never planned (“convenience-store clerk,” “Detective 2”). You can record those bit parts on the spot. They do not join the Character stage (the main-cast relationship graph); they are managed in the Character tab in the Library, so the relationship and faction graphs do not get buried under bit parts.
Inline editing & AI rewrite of a selection
You can edit the generated script directly. Drag-select a portion and an "AI Rewrite" button appears, letting you strengthen or replace just that part. Highlighted sections are passed to the Kick Assist chat on the right, so you can discuss edits with full context understood.
Plot Analysis (auto-triggered)
Once all the scenes of a beat are written, Plot Analysis runs automatically in the background while you stay on the Scene List. There is no need to press a separate “Start Analysis” button.
Go to the Plot Analysis tab to visually check the following for each beat.
- One-line design feedback + readiness badge (Pass / Needs work / Block)
- Emotion curve and conflict curve sparklines (intensity per scene)
- Hook density, thematic coherence, and average length statistics
- Areas to strengthen (weak scenes, duplicates, un-harvested setups) and next design actions
When you edit a scene, a "Re-analysis recommended" indicator appears. You can force a re-run with the "Re-analyze" button on the right.
This is the stage where you give your finished screenplay a title and export it in the format you want.
AI title suggestions
Press the “Get Title Suggestions” button and the AI comprehensively analyzes your synopsis, Characters, Plot, and script to generate title candidates in 4 directions.
- Theme-centered — implies the story's theme and message
- Character-centered — foregrounds the protagonist or a key character
- Image · symbol-centered — uses a core visual motif
- Genre-appeal-centered — a title that stirs the audience's curiosity
Each title is shown with a one-line rationale. Click a title you like and it's applied immediately as the project title; you can also enter your own.
Export
Choose the sections you want to export from among the synopsis, Plot structure, treatment, Character roster, and full scene script.
| Format | Description |
|---|---|
| TXT | Plain text |
| Word (DOC) | Word-compatible document |
| Print-optimized | |
| Final Draft (FDX) | Continue in the industry-standard screenwriting software — Pro |
| HTML | A single file that opens straight in a browser |
The treatment is generated automatically by AI based on the Plot. To back up or migrate the entire project, you can save it as a single file via Project Export (.mk.json).
To move the script into another tool, Copy all puts the whole thing on the clipboard at once.
TXT, Word, PDF and HTML all use the same colours and the same heading grammar. Layout follows where you write, automatically (see You don't pick a script layout in stage 6).
Casting roster
You can include a roster of every character, main cast and bit parts alike, as a table in the final draft. Role, gender, age, personality and the scene numbers they appear in are laid out, so the directing and production teams use it directly as casting material.
Sharing by link
You can share the finished work with a single link. Press "Share" at the top of the preview and a URL is created and copied to your clipboard. The recipient reads it in the browser, no login required.
Shared pages block search-engine indexing. To stop sharing, revoke the link from the same place and it stops opening immediately. You'll be asked to confirm once the first time you share.
If the draft is too large to fit in a link, we say so — you won't be handed a link that pretends to work and then doesn't. Share fewer sections, or export a file instead.
When something is generated, we read it back right then and point out what snags. The report isn't a separate screen — it arrives as writing in the chat on the right.
We don't hand you answers; we ask — as in "this choice runs against the desire you set earlier; is that intended?" The call is yours. Click an item and you're taken to the spot.
What gets read
| Target | When |
|---|---|
| Synopsis | as soon as a synopsis is generated |
| Plot | as soon as beats are generated |
| Boundary graph | as soon as Boundary analysis finishes |
| Space design | as soon as the location palette is generated |
| Scene list | as soon as the scene list is generated |
| Script | as soon as a scene script is written |
What can be checked by comparison is checked by comparison
Reading-and-judging is kept apart from what a straight comparison already answers. The latter is stated as fact, without guesswork.
- a lead character who appears nowhere in the scene list
- gaps in the scene numbering
- a run of scenes with the same location and the same cast
- a beat with no scenes at all
- a location in the palette that no scene uses (and the reverse)
- someone listed in the cast who never appears in the script
- how long the whole thing runs at your current pace
Turning it off and on
The "Auto review" toggle above the chat switches it off. With it off, nothing arrives on its own — reviews run only when you ask for one.
While a review runs you see "Reading…", and it lands as a piece of writing titled "[target] review".
If one model comes back blank, another model picks it up. If that fails too, we say we couldn't get it and tell you something you can actually do — never "switch to a model" you can't even select.
Reviews talk without theory jargon and reach for films as examples at any stage. The old check cards and the manual "Review result" button are gone.
An AI chat assistant that stays open at all times on the right side of the screen. It answers with awareness of your project’s full context (Synopsis, Character, Plot, Scene List, and the script you’re currently editing).
Key uses
- Edit the project directly — request something like "Change the protagonist's name from Cheolsu to Minho" and the AI updates it across the whole project at once. A rename now carries through the relationship map, treatment prose, scene-card descriptions and acting directions
- Add a seed (add_seed tool) — request something like "Show a watch in scene 10 and harvest it at the end," and the setup and payoff are automatically registered in the Library's Seed Map (with plant/remind/harvest points specified)
- Add Story Ground material — paste an article or reference into the chat, and it's automatically classified and registered
- Discuss a selection — a section you highlighted in the Scene Editor is passed to the chat, so you can discuss it freely
- Works on the home screen too — usable on the home screen before you open a project, e.g. for guidance on setting up your API key
It fixes what it points at
Chat used to be able to name a problem, but fixing it meant regenerating the whole plot. Now it changes only the part that needs changing.
- One beat — say "beat 7 is weak" and only that beat is swapped out; the rest stays, and sequence treatments you already wrote survive.
- A sequence's question and prose — the treatment can be edited through chat too.
- The cause, not the curve — when a Boundary curve is flat, it doesn't tell you to fix the curve; it points you at the beats and sequences that produced it.
When you accept a suggestion, "Overwrite" applies it right away.
Retracing what the AI changed
Open "Change history" at the top of the chat and you see when the AI changed what, in order. Everything that alters your draft is recorded here. The button appears once there is something to show.
How much was changed is shown as a number too — you can tell a single fixed sentence from a wholly rewritten paragraph without opening it.
In a conversation carrying text brought in from outside — an article, a reference — the AI asks for confirmation before editing your project. That way a "change this to…" sentence buried in the pasted material can't rewrite your draft on its own.
It reads the Boundary graph exactly as you see it
Chat reads the Boundary analysis as the graph on your screen. What it won't do is talk in numbers ("it rises to 62 at point 3") — it translates that geometry into what it means for the story.
- Cost — what was paid for this rise
- Reversal — where the direction turns
- Predictability — whether the audience already knows what comes next
- Rhythm — whether the same intensity holds for too long
If you edit the plot or a sequence after running Boundary analysis, an "outdated" marker tells you it's time to run it again. That marker appears no matter which screen you made the edit on.
It says so when it couldn't do it
When chat handles a request only partly, or not at all, it tells you it didn't — instead of claiming it did. It reports exactly what was applied and what is still outstanding.
Beyond that: it won't create a character that already exists, and when discussing the synopsis it looks up earlier versions on its own (nothing to switch on). Characters deleted through chat can be restored from the Trash.
A panel for systematically managing your story's supporting structures. Enter it via the Library button at the bottom of the left sidebar, where four tools are organized into tabs.
| Tab | Role | Active when |
|---|---|---|
| Character | Every Character — main cast plus bit parts — in one place | After Characters are created |
| Seed Map | Visually manage setups and payoffs (plant and harvest) | After the plot is complete |
| Page-Turning Techniques | Apply a catalog of 36 narrative techniques | After the scene list is generated |
| Reference Films | Match similar films based on a synopsis + plot hash | After the synopsis + plot are complete |
The Library manages your story's internal structure (seeds, techniques, references). Story Ground stores your story's external reference material (articles, notes, worldbuilding settings). They are separate panels with different roles.
A view of every character in the work, in one place. The character stage (relationship map) only handles the main cast, so a full roster including bit parts needs its own screen.
Main and supporting
Shown exactly as in the character stage. Read-only here — edits happen in the character stage, because letting the same information be edited in two places blurs which one is the source of truth.
Bit parts
Characters you add in the scene editor collect here. Bit parts are shown as a scene-order map — the horizontal axis is the progression of scenes and each dot is a scene they appear in. Click a dot or a name to view and edit it below, attach a note, or remove it from the list.
Deleting a bit part also deletes the acting directions attached to them.
This roster can be exported at the Final Draft stage as a Character roster table. Role, gender, age and personality plus the scenes they appear in — directing teams use it directly for casting prep.
A tool for visually managing setups and payoffs — that is, "plant and harvest." Track at a glance which foreshadowing is planted where, where it's reminded, and where it's harvested.
Two-layer structure
- Macro seeds — large foreshadowing at the beat level (theme, relationship, identity)
- Micro seeds — detail foreshadowing at the scene level (props, lines, images)
Lane visualization — one shared ruler
Each seed is shown as a lane graph. The horizontal axis is plot progression order, and the dots represent each seed's phase.
Every seed now sits on a single beat ruler. Lanes used to carry their own axis, which made them impossible to compare; now you can see at a glance which seed is planted before which.
The remind points you set actually reach scene generation at those spots.
● plant ━━━ ● remind ━━━ ● harvest
Adding seeds — chat-based
Seeds are added from the Kick Assist chat on the right. Request something in natural language like "Show a watch in scene 10 and harvest it in the final scene," and the AI calls the add_seed tool to register it in the Seed Map. Scene numbers are automatically converted into beat numbers.
User customization — Phase toggle
You can toggle a seed's phase directly on its card.
- Macro seeds — pending → planted → reminded → harvested
- Micro seeds — suggested → planted / reminded / harvested / rejected (5 states)
If you don't want to accept a seed the AI proposed, you can mark it rejected to exclude it from future generation.
A catalog of narrative techniques that make the audience want to see the next scene, and the screen where you plan what carries each seam between beats. It activates once the scene list is generated.
Seam planning — what carries this gap?
For each seam between one beat and the next, we propose what should carry the audience across it and what that looks like in this particular story, one sentence each.
Act breaks and turning points always get filled, and we keep the stretches between them from running too long. Some seams are left open on purpose — tightening without release wears the audience out.
Only proposals you adopt ride along when you write the scene that owns that seam. Anything you don't adopt stays out of generation.
Categories — six of them, 19 techniques
| Category | Techniques |
|---|---|
| Question (5) | Spark Curiosity · Cliffhanger Placement · Gradual Reveal · Layered Questions · Character Secret / Past Hints |
| Subversion (4) | Twist (Unexpected Ending) · Irony & Paradox · Unpredictable Character Choices · Set Up, Then Subvert |
| Pressure (4) | Time Pressure · Sharpened Conflict · Raising the Stakes · Point of No Return |
| Emotion (2) | Emotional Flash · Relatable Touchpoints |
| Rhythm (2) | Emotional Rollercoaster · Breathing Beat |
| Weave (2) | Motif Repetition · Information Asymmetry |
The list was rebuilt for features in 2026-08 (36 → 19). Short-form items (thumbnails, challenges) are gone, and what long stories need was added — letting the audience know what the character doesn't, raising what's at risk. What makes a scene well written (metaphor, rhythm) doesn't belong here; the scene-script side handles that.
Three-state management
- In use — techniques currently applied to your work
- Recommended — techniques the AI judges to suit your work's tone and structure
- Unused — unselected techniques that exist only in the catalog
Techniques marked "in use" are injected as context into the scene-generation prompt, so they're applied consistently across the whole work.
Automatically recommends films similar in structure, tone, and theme to your current work. It activates once your synopsis and plot are complete.
How matching works
Using the synopsis text + the hash (fingerprint) of your selected plot structure as a key, it matches the closest films from the internal reference DB. When the synopsis or plot changes, the cache is automatically invalidated and new recommendations are refreshed.
How to use it
- Structure validation — a benchmark for how acclaimed works using the same plot structure resolved their beats
- Genre tone reference — check whether the genre tone your work intends has settled in
- Reference awareness — if you're too close to one work, you can consciously create distance
The internal DB consists of over 500 real, fit-verified films.
A set of slots for gathering the external reference material that will ground your story. If the Library manages your story's internal structure, Story Ground stores material brought in from outside (articles, notes, worldbuilding settings).
It's organized into 4 slots.
- Worldbuilding (worldBuilding) — era, place, rules
- Characters (characters) — character details
- Plot · Synopsis (plotSynopsis) — storyline material
- Tone · Style (toneStyle) — voice and visual references
To add material, paste an article, setting, or note into the Kick Assist chat on the right and ask "Use this as reference material." The AI automatically determines the category and registers it in the corresponding slot.
Registered material is automatically referenced throughout synopsis, Plot, and scene generation, maintaining consistency of settings. There’s no need to paste it in every time yourself.
Text pulled off the internet can carry instructions aimed at the AI ("ignore the previous instructions and…"). Uploaded material is scanned first, in isolation from your story context, and you're told if such a sentence is found.
When material does reach generation it goes in marked as reference material — never on the same footing as your own instructions.
The Synopsis and Plot are automatically stored by version, and their version trees appear beneath the corresponding stages in the left sidebar. Each scene script has a separate version history inside its Scene Editor. Characters are not automatically versioned.
- Each version's source is distinguished by a colored dot — AI Edit Chat
- You can protect an important version by marking it with a star (★)
- You can restore and switch to a previous version at any time
- You can delete unnecessary versions (with a deletion warning provided)
As your work progresses, the AI automatically learns your creative style. It analyzes the differences between AI-generated text and your edited text to extract a profile across 10 axes.
- Dialogue style / Description style / Tone / Structure / Vocabulary
- Transition style / Rhythm pattern / Scene-length preference / Ending style / Contrast intensity
Open the "Style Preferences" modal from the user menu in the sidebar to review your learned profile, and apply it to future generation with the "Apply Preferences" toggle.
Activation condition — at least 2 AI-generated/edited pairs and at least 2 accumulated scripts are required.
The copyright strength of AI-generated work varies according to the degree of the user's creative contribution. Movie Kick supports this in three ways.
- Terms of Service — clearly state content ownership and rights relationships
- Per-version source records — AI generation, user edits, and chat application are automatically distinguished and can be used as evidence of contribution
- Authorship Contribution Report — generated from measured data at the Final Draft stage
This tool is AI generation + writer direction, so the "ratio of characters you typed yourself" comes out structurally low. Read alone, that number makes the contribution look small — which isn't what happened.
So the report leads with the number of chat directions and the number of iteration versions, the metrics where creative drive shows up honestly. Per-field edit contribution is shown alongside as supplementary evidence. The three metrics are shown as they are, never collapsed into a single score.
Merely opening a screen no longer counts as an edit — the ledger records only what you actually changed. What the AI changed is kept separately under "Change history" in the chat.
For the specific rights relationships, please refer to the Terms of Service.
.mk.json, and if you choose a Dropbox · Google Drive · iCloud sync folder, you can continue the same work on other devices. A one-off backup is also possible via Project Export in the Final Draft stage.
add_seed tool to register the seed in the Library's Seed Map and automatically sets the plant, remind, and harvest points. After registration, you can toggle the phase (pending → planted → reminded → harvested) directly on the card.
© 2026 Story Kick · User Guide v5.1 (as of 2026.08)