LearningFlow.ai
Whiteboard Studio

One sentence in. A drawn, narrated video out.

A hand draws each scene while a voice explains it. Fix one scene without redoing the rest.

  • Explainer videos
  • Voice-over
  • Subtitles

Watch a few.

Each one started as a single sentence.

Why water is H₂O"Explain the water molecule to ten-year-olds."

Inside the studio.

Every scene is listed. Edit one, and only that one is redrawn.

Whiteboard Studio showing a two-scene video, one scene changed, with a Re-render button

An explainer today, not next term.

Fix one scene

Change a sentence or a drawing. Only that scene re-renders.

Check before it renders

You see every scene and every line first. Nothing is wasted.

Pick the voice

Several voices, subtitles included.

Landscape or portrait

A classroom screen or a phone. Download the MP4.

Hand the LMS a SCORM 1.2 package.

One button in the toolbar turns the finished video into a zip that Moodle, Canvas, Blackboard or your company LMS will take. The LMS records who watched it and how far they got, on the terms you choose.

The whole video is inside the zip

The MP4 and its player travel in the file. No link back to us, so it plays on an LMS with no internet.

Completion, at the mark you set

Anything from half the video to all of it. Below the mark the LMS says incomplete; at it, completed.

A score, if you want one

Switch it on and the percent watched goes to the LMS as the score, against a pass mark. Nobody is ever marked failed.

Watched, not skipped

It counts the seconds that actually played. Dragging the bar to the end earns nothing.

They carry on where they stopped

Position is saved every few seconds, so a closed tab costs a learner almost nothing.

Your rules on the player

Fence off skipping ahead, pause when the tab is hidden, hide the download button. Re-watching is always allowed.

Tested on a real LMS

The package imports and reports completion on SCORM Cloud, the service LMS vendors test against.

SCORM 1.2. Not SCORM 2004, AICC or cmi5. And the score is the percent watched — a video has no questions to mark, so if your LMS has to record a result on answers, a course tool like Articulate or iSpring is still the right buy for that part.

You type what a video should explain. The system writes a storyboard, draws one picture per scene, and then stops, so you can read and change every line and every drawing before anything is narrated or animated. When you say go, it records the narration, works out the order each part of each picture should be drawn in, times the drawing to the spoken words, and joins the scenes into one video you download as an MP4, as a separate subtitle file, or as a zip your learning system can import.

The start screen: a box for the topic, with shape, length, look and voice controls, and prompt starters grouped by kind Start screen
Under the box sit five lanes of starters — Explainers, Lessons, Stories, Product and how-to, and one for portrait — because a vertical video is a different video, not a cropped one.
  • 2 to 12Scenes in one video
  • 3Video shapes
  • 11Length stops, 15 seconds to 5 minutes
  • 8Tools in the picture editor
  • 50Undo steps kept
  • 5Settings on the SCORM package

What it looks like in use.

One scene being edited, with tabs for Text, Image, Fix an element and Drawing order, and the narration in a box One scene
Four tabs per scene: the narration in the box changes what is spoken, not the picture you already have — that only changes when you ask for the drawing again.
The library of finished videos Library
Each card shows the first scene's finished drawing rather than the video's blank opening frame, and landscape, portrait and square are packed together.

All of it, not the highlights.

From one sentence to a finished video

You describe it, read what it wrote, then let it record and draw.

  • Type what it should explainThe system writes a storyboard: a title, the line to be spoken, and a list of the things to draw, for every scene.
  • It works out the lengthScenes are the length divided by twenty seconds, held between 2 and 12. Write "in 90 seconds" in your request and it reads the length from your words.
  • It stops and waitsPictures are drawn four at a time, then the job pauses. Nothing is narrated, timed or animated until you have read every scene.
  • Or it does not stopTurn the review step off before you start and it runs straight through to a finished video.
  • Six steps, liveStoryboard, Images, Review, Narration, Animate, Done. The page refreshes itself every two seconds while a job is working.
  • The log is on screenThe last twelve lines of what the job is doing, with the newest one alone in the status pill. The full log is kept as a file.
  • Then it draws and speaksAfter you approve, it records the narration, plans the order each part of every picture is drawn in, animates each scene and joins them into one video.

What you choose before it starts

Set on the start screen, and checked against the live list of voices and models each time you open it.

  • Three shapesLandscape 16:9, portrait 9:16, square 1:1. The picture is generated at that shape's own size rather than cropped to it.
  • Three looksMarker draws by hand. Sticker pops cut-out art in. Slides is typeset presentation slides, with four typography themes: Editorial, Geometric, Elegant, Plain.
  • Eleven lengthsA slider with fixed stops from 15 seconds to 5 minutes, or "read it from my prompt" if you would rather say it in the request.
  • Pick the narratorSearch every voice on the account by name, gender, accent or use case, and play any of them without closing the list. Leave it unset and one is picked at random.
  • Captions, three waysOff, a separate track the viewer can switch on, or burned into the picture.
  • How the scenes joinFive moves: straight cut, camera pan, smooth pan, wipe, fade. The move knows the shape — on portrait the board slides up, never sideways.
  • Which models do the workTwo models can write the storyboard, three can draw the pictures. The drawing order is either read off the picture by a vision pass or left to right.
  • Remembered on the machineYour choices are kept in the browser, not in your account. A link can also arrive with the topic and the shape already filled in, for that visit only.

Changing a scene during review

Every scene is open: the words, the picture, and the order it will be drawn in.

  • Four tabs on every sceneText, Image, Fix an element, Drawing order.
  • Rewrite anything it wroteTitle, narration, composition notes and the element list. Editing the text never changes the picture you already have — it changes what drawing it again will mean.
  • Three ways to a new pictureDraw it again from the brief, edit it with a written instruction, or fix one named part of the drawing and leave the rest of it alone.
  • Your own picturesAttach reference images to a scene and edit against them, or upload a finished picture of your own. 20 MB each; the wrong shape is refused.
  • Nothing is overwrittenEvery attempt is kept as a numbered version, and any earlier one can be made the current one again.
  • Watch one scene firstRender a single scene on its own and play it before committing to the whole video.
  • The order, over the drawingNumbered boxes sit on the picture in drawing order. Click one to select it, drag the list to reorder, or type the order you want in plain words and re-run it.
  • It tells you what is staleFour chips per scene — Picture, Narration, Drawing order, Animation — return to "to do" after an edit, and a finished video is marked out of date with a Re-render button.

How the drawing is animated

The engine reads the picture, then reveals it in time with the words.

  • It finds the separate objectsThe picture is broken into its parts and each is revealed on its own, in order. Pictures are worked at 1280 pixels on the longest side.
  • Words are drawn as wordsA text detector finds writing, so a label is drawn as one whole word rather than broken into pieces of stroke.
  • Trace, fill or fadeEvery part carries a role and a reveal style: line art and text are traced along their own skeleton, solid shapes are filled, faint decoration fades in.
  • Cued to the narrationEach part starts on the phrase it belongs to, found in the spoken word timeline by that phrase's first six words. Anything unmatched is spaced evenly between the rest.
  • A quarter second aheadThe pen starts 0.25 seconds before the word, and no window is shorter than 0.6 seconds. Drawing finishes inside the narration and the frame is held to the end.
  • No hand, no cursorNothing is composited over the video. The ink simply appears as it would be drawn.
  • Sticker and SlidesSticker cuts each element out with a white edge and a soft shadow. Slides appears the same way without the border, so a slide reads as typeset rather than stuck on.
  • Rewording keeps the timingReorder or reword the lines and an element the storyboard already knows keeps the phrase it was cued to, so the animation does not drift out of step.

Narration, captions and what you get out

  • Every word is timedSpeech is generated with a time for each word. The same timings drive both the captions and the drawing.
  • Only the line you changedNarration is cached by voice, model and text, so rewriting one sentence re-records that scene and nothing else.
  • Captions from the real timingsAt most two lines and five seconds per caption, built from the word times. Lines are made shorter for portrait and square videos.
  • A caption file either wayAn .srt is written whenever captions are on, including when they are also burned into the picture.
  • What the file is1280 pixels on the longest side: 1280x720 landscape, 720x1280 portrait, 1280x1280 square. That is 720p, not 1080p, even though the scenes are generated larger.
  • Three quality settings20 frames a second at a low bitrate, 24 at standard, or 24 at double the bitrate.
  • Download, or a linkAn MP4 that saves rather than opening in the tab, the .srt separately, and a share link that works for up to an hour.
  • A library of what you madeFinished videos gather on one page that plays each in place, packed by shape. Each card shows the first scene's finished drawing, not the video's blank first frame.

Handing it to a learning system

One button turns a finished video into a SCORM 1.2 package.

  • A zip, imported as a lessonThe video file travels inside the zip, so the lesson needs no internet connection and no link back to us once it is in your system.
  • Five settingsPercent watched to complete (50 to 100, default 90), report a score, allow skipping ahead, show a download button, pause when the tab is hidden.
  • Nobody is marked failedBelow the mark a learner stays in progress. With a score on, percent watched is the score, the status becomes passed, and the pass mark goes into the manifest.
  • Watched, not skippedOnly what actually played is counted; dragging the bar forward adds nothing. With skipping off a learner can go back and re-watch but cannot jump ahead.
  • It remembers where they stoppedPosition and watched stretches are saved every five seconds and on every pause. The next launch resumes, with a banner saying from where.
  • A real playerCaptions toggle, mute, full screen, keyboard seeking, chapter buttons named from the storyboard, a watched bar with a live percentage, speeds to 2x, and an Exit button.
  • Up to nine filesThe manifest, the player and the video, plus the captions as WebVTT and a poster picture shrunk to at most 1280 pixels on its longest side.
  • Built once, then instantAsk for the same options again on the same render and the download starts immediately. Anyone who can see the project can take the zip, as they could take the MP4.

Who sees it, and what happens when it breaks

  • Your organisation sees itA project is visible to its owner and to everyone in the owner's organisation. Only the owner can change it; to a colleague it opens read-only, with the owner's name on it.
  • Signed in alreadySign-in is your organisation's existing session — the app has no login of its own — and access is gated on a named permission, not merely on being signed in.
  • Five heavy jobs at onceOne organisation runs five projects with heavy work at a time by default; a sixth is refused with a message. A re-render, a redraw and a one-scene preview all count.
  • The hard limitsRequests up to 4,000 characters, uploads 20 MB a file, and at most twelve entries in a scene's element list.
  • Files are never publicEvery picture and video link is signed by the server and expires. Nothing sits at an address someone can guess.
  • A failed run resumesA retry keeps the storyboard and redraws only the pictures that are missing. One scene failing does not stop the others being drawn, and the message names the scene.
  • It retries the awkward onesA scene refused on copyright grounds is reworded and tried again. If two things come out touching, one stricter attempt is made. A rate limit is a wait, up to six times.
  • Delete means deleteDelete removes the records, the queued work and every file, and is refused while a worker is using the job. A name you give a project survives every later storyboard.

The parts that are bigger than they look.

A lesson your learning system marks, not just a video file

Any finished video downloads as a SCORM 1.2 package: the MP4, a player and a manifest in one zip, imported as a single lesson. The player reports how much was watched, counting only the seconds that actually played, and the learner resumes where they stopped. You decide the percent that counts as complete, whether a score is reported, and whether skipping ahead is allowed.

The SCORM package panel and its five settings SCORM package

The queue underneath, so a deploy does not cost you a video

Every change is a queued job, with quick edits in their own lane so a text change never waits behind a queue of full renders. When a worker stops, whether from a crash or a deploy, it spends up to ninety seconds finishing and saving what it holds, then puts the job back; another machine picks it up from the last checkpoint and the person sees a pause rather than a failure. Four separate caps limit how many picture, narration, storyboard and render calls run at once, which is why a busy hour rarely turns into a wall of provider errors.

A project open: the six-step progress across the top, every scene with its picture and narration, and the finished video with its download buttons A project, open

A drawing editor, not a crop tool

Clicking a scene's picture opens a full-screen canvas with eight tools — lasso, brush, eraser, line or arrow, rectangle fill, colour picker, text, hand — zoom from 5% to 800%, fifty undo steps, and three text faces. The lasso lifts only the ink inside the loop, so the paper underneath stays where it was. Save never overwrites: it becomes that scene's next numbered version, with the one before it still there.

What it does not do.

Worth knowing before you plan around it.

  • There is no language setting. Narration follows whatever language the storyboard is written in, and the caption track is always labelled English.
  • Scenes cannot be added, deleted or reordered once the storyboard is written. You can rewrite any scene; you cannot change how many there are.
  • There is no music, no sound effects and no second audio track. Exactly one narration track is mixed into the video.
  • Only SCORM 1.2 is produced — no SCORM 2004, xAPI or cmi5 — and the package's block on skipping ahead and its hidden download button are enforced in the page, not on a server, so a student using browser developer tools can get past both.
  • Downloads are the MP4, the .srt caption file and the SCORM zip. There is no GIF, WebM, audio-only, slide-deck or transcript export, and the MP4 is 1280 pixels on its longest side, which is 720p landscape rather than 1080p.
  • A copied share link stops working within the hour, so it is a way to show someone the video, not a way to publish it.

Free for teachers.

Sign up, get credits every month, make something today.