CARDBOARD / YC W26 / AGENTIC VIDEO EDITING
Cardboard AI video editorYour footage.
Your direction.
An AI first cut.
Cardboard wants to edit your videos before you touch the timeline. Can it take over the repetitive work while leaving the creative decisions with you?
Three directions.
Original storyboard simulation. No Cardboard connection.
THE BOTTLENECK AFTER YOU PRESS RECORD
Every company wants
to publish more video.
Somebody has to edit it.
Six takes of the introduction. Twenty minutes of screen recordings. Dead air. Captions. Another round of feedback. Somewhere in that folder is a good 30-second ad.
Cardboard combines a chat-directed first cut with a timeline you can edit. Its public materials show generated media, motion graphics, captions, reframing, and link-based feedback. [1]
The interesting question is what happens after the first draft. Can you change the opening without breaking the rest? Can someone else understand the revision? Can you finish the work in another editor?
01 / VIDEO CAMPAIGN EXPERIMENT LAB
One recording.
Three different ads.
Pick your footage. Set a goal. Compare the creative decisions. Leave with a brief you can use in a real editing session.
Educational simulation using original transcripts and illustrated scenes. It builds storyboards with local rules, not AI inference or rendered video. Custom instructions enter the brief; supported revision commands change the storyboard.
SOURCE BIN / ORIGINAL ILLUSTRATIONS
Inspect the sample transcript
Original versus edited timeline
Final quality checks
- Every quote keeps its original meaning.
- Screen recordings show the actual product, without private data.
- Captions are accurate and clear of interface controls.
- Music, stock, generated assets and voices have appropriate permissions.
- Watch the exported file, with sound and without sound, on a phone.
02 / CREATIVE REVIEW ROOM
The first cut
is only half the job.
“Make it better” is a difficult instruction. A timestamp, a reason, and an intended change give an editor something to work with.
A separate educational review simulation. Cardboard documents preview sharing and comments; the apply, reject and history controls below are our conceptual workflow. [1]
Revision history
- Original illustrative cut loaded.
03 / HOW DOES AN AGENT KNOW WHAT TO CUT?
An editor that
understands footage.
A transcript helps find the sentence. Visual context helps find the right shot. The timeline is where those choices become an edit you can inspect.
Conceptual pipeline, not Cardboard’s private architecture. Cardboard’s engineering page discusses WebGPU/WebCodecs, media synchronization, memory limits, multimodal search, and verification. These are engineering problems, not proof every planned capability has shipped. [5]
Frames have deadlines.
Decoding, compositing and audio playback must agree on time. A delayed frame or drifting soundtrack is visible immediately.
Media is heavy.
Loading a whole recording into memory is costly. Proxies, caches and loading around the playhead are common techniques for keeping an editor responsive.
A correct command can make a bad edit.
Trimming the requested seconds is a mechanical success. Cutting off the explanation is a storytelling failure. Human review still matters.
What changed after the original launch?
The launch HN post described WebCodecs/WebGL2 rendering in the browser, cloud video understanding, multi-track keyframes, and XML handoffs to Premiere, Resolve and Final Cut. [6] The current changelog adds cloud preparation, export checking, a Mac app, and external agents. A browser-only, entirely-local description would now be incomplete. [3]
04 / CARDBOARD AI REVIEW
Can it replace hours
of manual editing?
That depends on how much usable work remains after the agent stops.
Where I see the practical fit
A growth team already has footage and a recurring format: weekly product updates, testimonials, founder clips, or tutorial excerpts. An editable first cut could move their effort from finding and arranging moments toward choosing an angle and reviewing the result.
Three variations are useful only if they test different ideas. A new caption color is a weaker experiment than a different opening argument. Keep the audience, offer and distribution context visible in the brief.
The control that matters
A timeline gives a person a place to intervene. An export handoff gives an editor a way to finish. Those options matter when a reasonable first draft needs precise typography, sound, color, compositing, or changes the agent cannot make reliably.
What remains untested here
I have not used an authenticated Cardboard account, uploaded footage, timed a render, or inspected its exported project files. This is documentation-based analysis. First-cut quality, semantic retrieval accuracy, narrative coherence, caption accuracy, learning curve and long-footage performance need a hands-on trial.
The same goes for audio finishing, repeated brand consistency and ambiguous revisions. A published demo is a selected example. It cannot tell you the failure rate across your own recordings.
My recommendation
Try one approved recording against your current method. Include preparation, prompting, mistakes, revisions, export checks and human cleanup. Keep your existing finishing workflow until the new handoff works with your actual files.
A trial that tells you something
- Ask it to find a specific quote and a visually similar scene. Check both selections.
- Make three openings, then revise one without changing the approved middle.
- Test a vertical crop with multiple speakers and visible software controls.
- Send a preview to a colleague. Check whether the feedback survives the next revision.
- Open the exported edit in your finishing tool and inspect audio, fonts and media relinking.
No numerical ratings or claimed time savings. Cardboard’s performance claims have not been independently measured here.
05 / PRICING CHECKED OCTOBER 14, 2026
Check the plan.
Then check the credits.
The homepage and dedicated pricing page currently show the same four plans. The figures below are the displayed monthly launch offer, not an annual quote or a guarantee of future rates. [2]
| Plan | First month | Displayed recurring | Crossed-out reference | Generation / usage | Exports |
|---|---|---|---|---|---|
| Lite | $5 | $8/mo | $10/mo | None / standard | 3 · 1080p |
| Starter | $9 | $29/mo | $40/mo | 400 credits / 2.5× | Unlimited · 1080p · NLE |
| Creator | $30 | $59/mo | $100/mo | 1,500 credits / 10× | Unlimited · 4K · NLE |
| Pro | $89 | $175/mo | $250/mo | 5,000 credits / 30× | Unlimited · 4K · NLE |
Mac changes the entry point.
Cardboard advertises complimentary Lite on Apple Silicon, with three 1080p exports and no card needed. Connecting your own Claude or Codex requires an eligible paid plan with that provider. Local media processing does not establish that every AI request stays on-device. [7]
Ask before committing.
Usage multipliers are not an absolute video-minute allowance. I could not establish storage quotas, top-up prices, precise per-task consumption, minimum macOS version, or team seat charges from the retrieved pages. Confirm those details in your account. Cancellation retains access through the billing period. [2]
Start with desktop Chrome for the browser product. The homepage excludes phones and tablets; its Mac app targets Apple Silicon. A March changelog documents Edge export, but I would not infer identical support for every browser or codec. [1] [3]
06 / A USEFUL IN PRACTICE CAMPAIGN
Could my Field Notes
become video ads?
A new operator.
Use the existing Voyager article as the editorial source. Record Stephen explaining one practical workflow, capture an approved tool demonstration, and turn it into channel-specific cuts.
An original production proposal. No footage has been edited in Cardboard. Capture the real product before making a product demonstration.
07 / WHICH WORKFLOW FITS?
Start with the footage.
Or start with the idea.
| Tool | Primary workflow | What to check |
|---|---|---|
| Cardboard | Existing footage and generated assets; chat plus timeline | Precision, cleanup, credit use, export handoff [1] |
| Vyra | Upload footage and describe the edit; aimed at non-editors | Fine-grained control and repeated revisions YC |
| Voyager | Local creative harness for models and applications | Your application connections and editable output YC |
| Palette | Connected multimodal generation and conversational edits | Continuity of references and generation costs Official |
| Descript | Transcript editing, Underlord, recording and clips | Spoken-word accuracy and layout control Official |
| CapCut | Social editing with templates and AI features | Device-specific features and asset permissions Official |
| Adobe Premiere | Professional timeline editing with AI-assisted features | Team workflow, finishing requirements and subscription Official |
| DaVinci Resolve | Editing, color, Fusion effects and Fairlight audio | Hardware, learning curve and Free versus Studio features Official |
Conversational editing is no longer unique to one vendor. Compare your actual task, required control and total work. Competitor pricing and feature parity are not asserted here.
08 / SPECULATIVE CREATOR PORTFOLIO
Three hooks.
One editing problem.
Original UGC concepts for a possible collaboration. Illustrated storyboards, not commissioned ads or completed Cardboard outputs.
09 / THE PEOPLE BUILDING THE EDITOR
They had footage.
They had the same problem.
Saksham Aggarwal / CEO
YC describes a BITS Pilani computer science background, AI work at Iterate AI, research published at ACL, early engineering at PYOR, and growth work at Flint. His content creation gave him a reason to want a faster edit.
Ishan Sharma / CTO
YC describes roughly 4.5 years at HackerRank building high-performance web products. He also produced internal launch videos. The founders say they have been friends since school, approximately 15 years.
YC lists Cardboard as founded in 2025, Winter 2026, team size eight, Bengaluru, with primary partner Ankit Gupta. Current product pages say “Made within SF”; jobs include Bengaluru and San Francisco. These are public listings, not an independently verified headquarters or headcount. No funding amount or customer revenue figure was established. [4]
Watch the original product demonstration.
The official YC launch links this video. Watch the actual editor alongside the educational simulation.
Open official demo ↗Loads a third-party video only when selected.
FOR FOUNDERS & MARKETING TEAMS
Your product.
Three videos.
One good story.
I turn complex software into something people can understand and explore: product walkthroughs, creator-led software ads, tutorials, technical explainers and interactive launch features.
This feature is independent. A proposed paid pilot would be separately scoped and disclosed. The YC launch invited growth teams to collaborate; I found no verified public Cardboard affiliate or creator program in the reviewed sources. No affiliate links or existing partnership are claimed. Original invitation ↗
PRACTICAL QUESTIONS
Before you make the first cut.
What is Cardboard AI?
Cardboard is an agentic video editor that accepts footage and creative direction, builds an editable cut, and supports revisions through chat and a timeline.
Is this a hands-on Cardboard review?
No. It is independent analysis of public product materials. Output quality, export reliability and editing performance have not been tested in an authenticated account.
Does this experiment lab edit real videos?
No. It is a deterministic educational simulation using original illustrated scenes and transcripts. It produces storyboards, creative briefs and editing prompts, not media files.
Is Cardboard free?
Its Mac page advertises free Lite access on Apple Silicon with three 1080p exports. Browser plans and paid generation have separate limits. Check the live plan before starting.
Can I move a Cardboard edit into another editor?
Paid plans list NLE exports. Public launch materials describe XML handoffs to Premiere, DaVinci Resolve and Final Cut. Test your project, fonts, audio and media relinking before relying on the handoff.
PRIMARY SOURCES / CHECKED OCTOBER 14, 2026
Check the work.
- Cardboard homepage: current product, device and sharing claims.
- Dedicated pricing page: displayed monthly offers, credits, export limits and cancellation.
- Changelog: Mac and agents, cloud processing, export checking and editor updates.
- YC profile and jobs: founders, batch and self-reported company details.
- Hard problems: engineering goals and open problems.
- Launch HN: founders’ launch-era technical description.
- Cardboard for Mac: local processing, free Lite and eligible agent subscriptions.
Official demo linked but not independently played for this review. No personal product-testing or measured performance claims are made.