Explainer videos you can actually edit

Sketchie turns any prompt or document into a whiteboard explainer video in about a minute. Then it does the part one-shot tools leave out by design. Click any scene. Change it. Re-render in seconds, for cents.

7 days or your first 3 videos, whichever comes first. Full product, scene editor included. Card required, no charge during the trial. Cancel anytime.
Made with Sketchie
Sketchy, the Sketchie mascot Built by Celinda, a psychology student, in a single day. Read the whole story →

One-shot AI video tools hand you a finished MP4. If one sentence is wrong, you pay to regenerate the whole thing and hope. Sketchie gives you a video made of scenes you can open.

How it works

Prompt in. Video out. Fix in seconds.

1

Drop anything in

A prompt, a PDF, a Word doc, a Markdown file, your product docs. Sketchie reads it and writes a script built on story structure.

2

Watch it draw

Every element draws on stroke by stroke, timed to the exact word of narration it belongs to. The biggest retention lever in the research.

3

Fix what is off

Open the scene list, click a scene, edit the text, visual, or narration. Sketchie re-renders that scene only, in seconds. Cents, not another full video.

4

Ship it

Export MP4 in any of three aspect ratios, or pull the scene graph through the API and do whatever you want with it.

recursion-explainer.mp4  · 6 scenes · 1:12
Scenes, not takes

Fix one scene. Not the whole video.

The one-shot tools that dominate this category work one way: input goes in, MP4 comes out, and the workflow ends. Sketchie’s videos are living documents. Every scene is addressable and editable after generation.

  • Edit narration, headings, icons, and timing per scene
  • Re-render a scene in seconds, for cents
  • Reorder, delete, or duplicate scenes like slides
  • Fix one sentence in a 5-minute video without re-spending 5 minutes of credits
Why Sketchie

Built different where it counts.

Scenes, not takes.

Every scene is addressable and editable after generation. Fix one sentence in a 5-minute video without re-spending 5 minutes of credits.

Built on learning science.

Progressive drawing beats finished diagrams for retention, attention falls off after 6 minutes, ending with a question beats a summary. We wired those findings in as hard rules with an eval harness behind them.

Any voice. Including yours.

A voice library designed for the audience, like a kid-explains-to-kids voice for young learners. Or paste an ElevenLabs or Fish Audio voice ID and Sketchie narrates in your voice.

Length you control.

Set length in 30-second steps, or let Sketchie pick. Videos default to 6 minutes or less because that is where the engagement data says attention lives. Longer material becomes a chaptered series.

Priced like we mean it.

A real 7-day trial on every plan, not a locked demo. Edits re-render one scene for seconds of minutes, not a whole regeneration. And API access is included at 1x on Max and Business, not a 3x markup.

API and MCP, day one.

One endpoint in, video plus editable scene graph out. Plus an MCP server, so any AI agent can generate and edit explainers directly.

Sticky, not viral

We optimized for whether they remember it tomorrow.

Most AI video tools optimize for the demo: how fast, how long, how flashy. We optimized for whether the viewer remembers the thing tomorrow. Sketchie’s generator enforces rules drawn from published, peer-reviewed research.

  • Everything draws on, always. Content built up stroke by stroke beats showing completed scenes for both retention and transfer. This is the core mechanism, so we never break it.
  • Six minutes is a cliff, not a guideline. Analysis of 6.9 million online-learning sessions shows engagement near 100% under 6 minutes, roughly half at 9 to 12 minutes, and a fifth past 12. Sketchie caps videos at 6 minutes and chapters anything longer.
  • Every video ends with a question. A retrieval beat: the question appears, the video pauses, then a one-line recap. Making the learner generate the answer improves learning.
  • Pauses are on purpose. After a scene finishes, the narration holds before the cut. Processing time at boundaries is part of how segmenting improves learning.
  • No decoration. Every visual element on screen is referenced by the narration. Our eval harness treats interesting-but-extraneous content as a failure, not flair.

Read the full spec we built from the research

Pricing

Four plans. Every one starts with the trial.

Plus
$19.99 /mo
100 minutes / month
  • 100 rendered minutes every month
  • Scene editor, all voices, all aspects
  • Full 7-day trial to start
See full pricing
Max
$99.99 /mo
750 minutes / month
  • 750 rendered minutes every month
  • Scene editor, all voices, all aspects
  • API + MCP access at 1x
See full pricing

Business adds custom branding. Annual billing takes 40% off. See all four plans.

Questions

The honest FAQ.

How is this different from Simi, Knowlify, or Golpo?
Simi’s own launch copy promises "a finished video, not a draft." Sketchie’s premise is the opposite: every video ships as editable scenes. Fix one, re-render one, pay cents. Full comparison, sourced and dated, on the vs page.
How does the trial work?
Pick a plan and start. The trial runs 7 days or your first 3 videos, whichever comes first, with the full product including the scene editor. Edits and scene re-renders do not count toward the 3 videos, so fix as much as you like. Then the plan you picked simply continues. Nothing expires, nothing gets taken away. Your plan starts when you’ve seen it work.
How fast is it?
About a minute for a typical video, and a few seconds per scene when you edit. Some competitors generate faster. We think the speed that matters is time to a video you would actually publish, including the fixes. One-shot tools lose that race.
What can I feed it?
A plain prompt, PDF, Word, PowerPoint, TXT, or Markdown. Product docs, lesson notes, SOPs, PRDs, changelogs.
Can it use my voice?
Yes. Paste an ElevenLabs or Fish Audio voice ID, or pick from the library. The library includes designed voices matched to the audience, not just a male/female toggle.
How long can videos be?
Up to 6 minutes per video, in 30-second steps, or auto length. Longer material becomes a chaptered series. That is a deliberate design decision backed by engagement data, not a limitation we are working around.
Is there an API?
Yes, plus an MCP server for AI agents. API access is included on the Max and Business plans, and API generations draw from your plan minutes at 1x, the same as the app. One POST returns the video and the full scene graph, so edits are in the API contract too.
Who is behind this?
A small team building from Lima, Peru. No venture funding. We publish our comparisons and our reasoning, receipts included.
Sketchie mascot holding a marker

The first draft takes a minute.
So does the fix.

Start free trial
Edit one scene during the trial and you’ll understand why one-shot tools feel broken afterward.