User documentation

Kino Director and Kino Scribe share one account, one login, and one credit balance — sign up with Google, Discord, or an email and password. New accounts get a one-time $2 free credit automatically as soon as you verify your email address (check your inbox for the link) — no card required to try either app.

Everything in both apps runs on prepaid credits, denominated in USD. Renders, AI images, and subtitle jobs are priced up front (you always see the price before you commit), and you're only ever charged once a job actually succeeds — see Billing & credits for the full picture.

New to the editor? Open the ? menu in the top-right of either app and choose Take the tourfor a short guided walkthrough of the screen you're on — this page is the complete reference to come back to any time.

A Kino Director project is an ordered list of shots — the strip on the left of the editor. Each shot is a 5–15 second AI-generated clip you write, configure, and render independently; the project plays them back in order.

Shots

• Add a shot — the + button above the strip appends a new, blank shot.
• Reorder — drag a shot card by its handle to move it earlier or later in the sequence.
• Select— click a shot's thumbnail or title to open it in the settings panel and the preview.
• Clone Shot— the copy icon on a shot card duplicates it (prompt, references, and params included) as a new shot right after the original. Use this to preserve a shot's current state before you start changing it — there's no separate “retake with the same params” button; the shot's current settings are always what Render take uses.
• Delete — the trash icon on a shot card removes it (and its takes) from the project.

Each card shows a status dot for its most recent take (running/succeeded/failed), a progress bar while it renders, and its duration. Hover a card to reveal Clone/Delete.

Selecting a shot opens its settings panel. Top to bottom:

Title

A free-text label for the shot — shown on its card and in the takes queue. Purely organizational.

Mode: Animate vs. Reference

Every shot is one of two mutually exclusive modes — switching replaces the conditioning section below entirely:

• Animate— “this exact photo moves.” You supply a first frame (and optionally a last frame) and H3 animates between them.
• Reference— “same character, new scene.” You supply up to 9 images, 3 videos, and 3 audio clips as loose references (character look, voice, style) and describe the new scene in the prompt — H3 is not literally moving one photo.

See References & keyframes for how each mode's slots work.

Style / scene

Free text describing the overall look — cinematography, lighting, mood (e.g. “Cinematic, handheld, golden hour…”). This is also where you bind an audio reference's identity — see References & keyframes.

Beats

What happens in the shot, broken into ordered, time-ranged steps — beats are the shot's clock: a beat's start and end time in seconds is the only timing control the shot has. Each beat also has an optional label(e.g. “SHOT 1”) and a text description of the action. Drag a beat by its handle to reorder it; the trash icon removes it. Add beat appends a new one starting where the last one ended.

There is no separate dialogue field — you write a character's spoken line directly inside the beatthat covers the moment it's said, tagged to a speaker. The beat's own start/end range is what times the line:

She stops at the water's edge and the woman from <Picture 1> (S1) says: <d>[English] Come on! The water's amazing!</d> — then beckons with one hand.

(S1), (S2), … are speaker IDs — tag each spoken line to the speaker saying it. Stick to one speaker per shot for the cleanest lip sync.

Audio

Free text describing ambient sound, music, and sound effects (e.g. “wind, distant traffic, the score bursting at 4s…”). This is also where you bind a music reference — see References & keyframes.

Constraints (positive-only)

Free text for what you want kept OUT of frame or preserved — phrased positively (e.g. “No text or subtitles, keep the live-action texture…”), not as a list of banned words.

The : reference picker

Inside Style, Audio, Constraints, or any beat, type :to open a dropdown of the shot's attached references — keep typing to filter, Enter or click to insert that reference's tag (e.g. <Picture 1>) at the cursor, Escape to close without inserting. Any tag typed or inserted into a field shows as a small chip below it — hover the chip (or the matching reference tile) to see which one it points to.

Generation settings

• Duration— 5.2–15.1 seconds, snapped to H3's frame grid as you drag.
• Resolution — a preset picker: Landscape (864×480, the default), Landscape native (1344×768), Portrait native (768×1344), or Square (960×960). The 864×480 default renders at the cheaper Draft price; every other preset renders at HD — see Billing & credits.
• Seed — locked (a fixed number, reproducible) or unlocked (randomized every render). The lock/unlock button toggles it; the dice button rerolls a locked seed.
• Steps— sampling steps for the render, default 20. Higher can mean more detail at the cost of longer render time; most shots don't need to change this.

Animate mode

Two image tiles: First frame (required — the photo that animates) and Last frame(optional — where the motion ends up). Click a tile to pick an image from the project's asset library, or use its sparkles button to generate one — see AI assistance. Animate-mode slots have no reference tags; there's nothing to insert into the prompt.

Reference mode

Three grids of slots, each with a fixed capacity and its own H3 prompt tag once bound:

• Images — up to 9 slots, tagged <Picture 1>, <Picture 2>, …
• Videos — up to 3 slots, tagged <Video 1>, <Video 2>, …
• Audio — up to 3 slots, tagged <Audio 1>, <Audio 2>, …

Slot numbers (image_3, etc.) never change once bound — clearing a slot leaves a hole. The tagshown and inserted, however, is always the contiguous, hole-skipping count of what's currently bound: if only your second and fifth image slots are filled, they submit to H3 as <Picture 1> and <Picture 2>, not 2 and 5. Each tile shows a sparkles button (generate an image into that slot — image slots only, see AI assistance), an insert-tag button (drops the tile's current tag into whichever field you last had focused), and an X to clear it.

Using references — the audio-reference recipe

An audio reference conditions H3's voice/soundtrack — it is nota transcript. H3 never reads text off an audio clip; you always have to tell it, in words, both who's speaking and what they say:

• Bind the voice in Style or Audio: “<Audio 1>is the speaker's voice — preserve its timbre.”
• Give the exact wordsright inside the beat that covers the moment, tagged to a speaker: “The woman with a warm voice (S1) says: <d>[English] exact words</d>” — see Writing a shot's Beats section for the full pattern.

One speaker per shot gives the cleanest lip sync. For a music reference, state its role directly: “use <Audio 1> as the soundtrack.” An audio reference with speech but no matching spoken-line text in a beat produces unintelligible mumbling — this is documented H3 model behavior, not a bug: always pair a voice reference with an explicit spoken line.

AI beat

The AI beatbutton above the Beats list proposes and appends a new beat, written from your shot's current context (project name, shot title, style, existing beats, and audio) — the same H3-dialect text you'd write yourself. Each beat row also has its own Improvesparkles button that rewrites just that beat's text in place.

Both are free — no credits charged — but require a verified email and are rate-limited per day to keep the feature sustainable. An inline error banner explains if either condition blocks a request.

Generate image

Any image-kind reference tile (first frame, last frame, or an Image slot) has a sparkles Generate image button that opens a dedicated dialog:

• Write a prompt, pick an aspect ratio (square, widescreen, or portrait), and choose a model — each model's per-run price is shown right on the Generate button.
• Generate as many times as you like — every run charges (only on success) and lands in the dialog's own runs queue at the top, with its status and price.
• Select one or more finished runs (checkboxes), then Add to project to bring them into the project's asset library.
• From there, pick the image into the frame/reference slot exactly like any uploaded asset.

Render take, at the bottom of the shot panel, renders the shot's current settings — there's no separate “retake with the same params” button; use Clone Shot first if you want to preserve the current version while you experiment. The button shows the full-run price before you click; a shot with pending changes reprices automatically.

Pricing tiers

H3 renders bill per output-second, in one of two tiers based on resolution: the default 864×480 resolution renders at Draft ($0.12/s); every larger preset (native landscape, portrait, or square) renders at HD ($0.25/s). You always see the exact quoted price on the Render take button before committing.

The Renders queue

Every take across the whole project — not just the selected shot — lives in the global Rendersqueue (the drawer icon in the top bar), newest first, each row tagged with its shot's number and mode icon. This is where more than one shot can render at once and you can watch all of them together. The shot panel itself only shows a one-line trace of that shot's own latest take, with a link to jump into the full queue filtered to it.

Each row shows a status badge:

• Pending / Submitted / Running — in flight, with a live progress bar and stage label.
• Ready — succeeded; click the row to preview it, or use the download button.
• Failed — charges nothing; click the error summary for details.
• Cancelled — the result of an abort (see below).

Abort

A still-running take can be aborted. Because the GPU time already spent is billed to us regardless, abort settles the executed portionof the original quote — never more than the full reserve, and $0 if the render hadn't really started yet — and releases the rest. The confirmation dialog shows both the estimated executed-portion charge and the original full-run price side by side. A row that ended up charged from an abort shows its charge in an amber, underlined “?” — click it for the exact amount. This is different from a Failed take, which always charges nothing; partial charging only ever happens on a user-initiated abort.

Other row actions

The kebab (⋮) menu on every take row, plus a few dedicated buttons:

• Restore settings from take — copies that take's frozen params back onto the shot, without re-rendering.
• Delete — removes a finished take (disabled while it's still running).
• Download — the succeeded take's video file.
• Report (flag icon) — flag a take with a reason (bad quality, wrong result, charged but failed, other) and an optional message, sent straight to us for review.
• Associate with shot…(queue-only) — move a take under a different shot in the same project, optionally also loading that take's frozen settings onto the target shot.
• Open shot(queue-only) — jumps straight to the take's shot in the editor.

A take whose original shot no longer exists shows an italic “deleted shot” badge in the queue instead of a shot number — it's still there for reference/download, it just has nowhere to jump back to.

Below the preview video, the sequencegroup is the project-wide timeline — a play/ pause button, an elapsed/total timecode, and a ruler with a block per shot (click a block to jump to that shot; drag the playhead to scrub across the whole project). It's visually distinct from the per-shot transport underneath it, so it's always clear whether you're watching one shot or the full sequence — only one of the two is ever “armed” (highlighted) at a time.

Transitions

Hover the gap between two shot cards in the strip to reveal a small +— click it to add a transition (crossfade or fade to black, with a duration) between those two shots; click an existing transition's chip to reconfigure or remove it. Configured transitions also show as small marks between shot blocks on the sequence timeline.

Export movie

The Export moviebutton next to the sequence timeline stitches every playable shot's chosen take (with transitions applied) into one MP4 and downloads it — a single fixed, good-quality format, no format picker. Its price is shown on the button before you click: free if no shot's subtitles are burned in, otherwise priced the same way as a per-shot subtitle export (see below), over the whole movie's duration.

Each shot has its own collapsible Subtitles panel in the settings lane, independent from the full Kino Scribe app:

• Enabled — turns the subtitle track on/off for this shot.
• Style — Movie (plain lines) or Highlight (word-by-word active-word highlighting, Instagram/TikTok-style captions).
• Build from chosen take— runs automatic speech recognition on the shot's chosen take to generate subtitle segments, priced at $0.10/minute of audio, capped at $1.00 per build. Requires a chosen (rendered) take first.
• Segments — the resulting lines, each editable (text, start/end time), with split, merge, and delete controls per row, plus Add line to add one manually.

Edits autosave a moment after you stop typing — a status label next to the segment list shows Saving/Saved/Save failed.

The standalone Kino Scribe app (a separate product from Kino Director, one shared account/credit balance) is purpose-built for captioning an existing video, start to finish:

Import

From the Kino Scribe dashboard, New video uploads a video file straight into a new project and opens the editor — no storyboard, no shots to configure, just your video.

Transcribe

The Auto-subtitles button runs ASR (automatic speech recognition) on the imported video, detecting the spoken language and generating segments — priced at $0.10/minute of audio, capped at $1.00 per build, shown on the button before you run it.

Edit

The segment list below (and the segment timeline under the video) let you retime, edit, split, merge, or delete any line, and drag the caption box directly on the video preview to reposition or resize it.

Styles, including the word-highlight look

Two styles, same as Kino Director's per-shot panel: Movie (plain lines) and Highlight— the Instagram-style look where each spoken word lights up in a distinct highlight color as it's said. Text color, highlight color, font size, letter spacing, word-wrap (with a draggable caption box), and an optional background box with adjustable color/opacity are all configurable here.

Export & prices

• Export — a good-quality MP4 with subtitles burned into the video, $0.02/minute of source video, capped at $5.00 per export.
• Export subtitles — a plain SRT, VTT, or ASS subtitle file (never a video), $0.01/ minute of source video, uncapped. Each format shows its own price in the dropdown.

Both apps let you flag a problematic result with the same Report (flag) button described under Rendering.

Credits are USD-denominated and never expire. Buy creditsfrom Settings → Credits: fixed packs ($1 / $5 / $15 / $50 / $100 — the $50 and $100 packs include a +25% bonus), or a custom amount (minimum $0.50). Checkout runs through Stripe; if you're buying from Brazil, Stripe charges your card in BRL and credits the equivalent USD to your balance — we never see or store your card number.

Every priced action in either app follows the same rule: charge on success only. The price you see before you click is a reservation, not an immediate charge — a failed render, ASR build, or export charges nothing at all, and the reservation is released. The one exception is an aborted in-flight render, which settles only the executed portion (see Rendering).

Full terms, refund conditions, and how we handle your data are in the Terms of Service, Refund Policy, and Privacy Policy.

• Report a take— the flag button on any take row (Kino Director's Renders queue or Kino Scribe's export bar) sends us the take, your chosen reason, and an optional message.
• Email support — support@getkino.io for anything else, including billing questions.
• Policies — Terms of Service, Privacy Policy, Refund Policy.