Option A
Cloud-stored projects
Project files, source media, autosaves, and exports live on a vendor server. Multiple devices see the same project; the vendor handles backup, sync, and quota.
COMPARE
Where your project lives shapes what it costs you in privacy, durability, quota, and dependency. The right answer depends on whether you're optimizing for sovereignty or for syncing.
Option A
Project files, source media, autosaves, and exports live on a vendor server. Multiple devices see the same project; the vendor handles backup, sync, and quota.
Option B
Project files, source media, autosaves, and exports live on the user's device — typically through OPFS, IndexedDB, or the local filesystem. The device is the source of truth.
Where a video project lives is one of the most consequential decisions in a modern editing workflow, and it's the one creators most often make by accident. Cloud storage is convenient — your project is on every device, the vendor handles backup, and collaborators can join with a link. Cloud storage is also a privacy posture: your raw footage now exists on a server you don't control, subject to whatever retention, breach, and access policies that vendor maintains. Local storage flips both columns. Your media never leaves the device, so there's no third-party copy to leak, but you're responsible for backup, your project doesn't follow you between devices automatically, and your free disk space caps the project. The honest 2026 view is that neither is universally better — it depends on your workflow's sensitivity, your team's collaboration needs, and how much you trust the vendor with footage you may not be able to take back. Below we break down the dimensions that actually matter, and we explain why Skrrol AI's editor is local-first as a core design decision rather than a feature flag.
| Dimension | Cloud-stored projects | Local-stored projects |
|---|---|---|
| Privacy | Vendor-hosted; your media exists on disks you don't control. | Device-only; no third-party copy of your media exists. |
| Cross-device access | Sign in anywhere, project follows you. | Project lives on one device; cross-device requires manual move or sync. |
| Backup | Vendor-managed; usually redundant across data centers. | User-managed; you back up yourself or accept the risk. |
| Storage quota | Vendor pricing per GB; quotas climb as projects accumulate. | Bound by local disk; can fill up on long projects with 4K source. |
| Cost over time | Recurring monthly storage fees; never paid off. | One-time disk cost; no ongoing fees. |
| Vendor dependency | Vendor shutdown or account ban can lose your work. | Files survive any vendor's choices; you own the medium. |
| Sync speed | Bound by upload bandwidth; large shoots take hours. | Instant — files are already on the disk you're editing on. |
| Collaboration | Real-time, link-shareable, multi-cursor workflows. | Asynchronous; share project bundles or use a sync layer. |
| Offline editing | Limited — many cloud editors require connectivity to scrub. | Full — disk-based projects don't care about the network. |
| Disaster recovery | Strong if the vendor's redundancy is real; weak if you're locked out. | Depends entirely on your own backup discipline. |
Multiple editors, producers, and reviewers all need the same project. Cloud is the natural surface for live, shared editing.
If you start on a desktop, continue on a laptop, and review on a tablet, cloud storage saves you the constant manual sync. The project is always where you are.
If you don't trust your own backup discipline (and most people shouldn't), cloud storage with a serious vendor is more durable than your laptop's SSD.
Working from a Chromebook, a tablet, or a thin laptop with a small drive? Cloud offloads the storage burden to someone else's hardware.
Legal evidence, medical footage, brand-confidential pre-release material, journalist source files — these don't belong on someone else's server. Local-only is the correct posture.
Local storage is a one-time disk purchase. Cloud storage is forever. Over five years on a heavy workflow, the cost difference is meaningful.
If the editor closes shop or your account is locked, files on your own disk are unaffected. Cloud-stored projects can be unrecoverable in either scenario.
Field shoots, travel, rural connections — uploading a 50 GB shoot is a non-starter. Local keeps the bandwidth math out of the workflow.
Skrrol AI's editor is local-first by design, not by toggle. Every project file, source clip, autosave snapshot, undo entry, thumbnail, waveform, and exported MP4 is written to the user's device through the Origin Private File System (OPFS) and IndexedDB — sandboxed local storage that browsers expose specifically for this kind of workload. Skrrol's servers never see a frame of your footage or a byte of your project JSON. That's a deliberate architectural commitment: editor data is local-only, and we've documented it as a core policy. The honest tradeoffs are real. Browser quota is finite, so we surface preflight warnings before large imports and a global banner when free space drops under two gigabytes — your project doesn't silently fail. Cross-device editing requires you to export a project bundle and reopen it on the new device — there is no automatic sync. And if you clear the browser's storage for our origin, you delete the work; we autosave aggressively, but the medium is your disk. Our generators (Sora, Veo, GPT Image, ElevenLabs, Lyria) do round-trip to our cloud because the underlying models need data-center GPUs, but everything you do with the generated assets afterward — editing, composing, exporting — stays on your device. We think this is the cleanest privacy posture a real video editor can offer in 2026.
Different, not equivalent. OPFS persists across browser sessions and reboots, but if a user clears site data, runs an aggressive cleanup tool, or fills the disk past quota, files can be evicted. For long-term archival, exporting a project bundle to a backed-up local folder is the right pattern. Cloud storage is more durable on its own, in exchange for the privacy cost.
A well-designed local-first editor warns you before that happens. Skrrol surfaces a preflight dialog before large imports, a global banner when free space drops below two gigabytes, and a friendly QuotaExceededError toast with cleanup guidance if it does fill. Your work is never silently dropped.
Yes — Skrrol exports project bundles that include the timeline, your settings, and references to source media. You can move the bundle to a new device and reopen it. There is no automatic cross-device sync, and that's intentional.
Encryption helps, but the vendor still controls the keys in most setups. If they're served a subpoena, get breached, or change their privacy policy, your media is reachable in ways device-only storage isn't. End-to-end encrypted cloud is closer to local from a privacy standpoint but still doesn't beat "never left the device."
Because the AI models — Sora, Veo, GPT Image, ElevenLabs, Lyria — physically can't run on a phone or laptop GPU. Generation has to round-trip to the data center where the model lives. Skrrol sends the prompt and any reference images you attach, and the generated asset comes back to be saved locally. Editing those generated assets afterward stays on your device.
The local-first architecture doesn't preclude an opt-in sync layer in the future. If we ever add it, it would be opt-in per project, end-to-end encrypted, and disabled by default. The principle stays: your media doesn't leave your device unless you explicitly choose to send it.
Online cloud editors vs fully offline editors compared on speed, privacy, collaboration, and reliability. Plus the hybrid approach Skrrol AI takes.
Browser-based vs desktop video editors compared on performance, privacy, install friction, AI features, and cost. Find which fits your workflow and where Skrrol AI lands.
Free vs paid video editors compared on features, watermarks, export limits, AI tools, and total cost. Where Skrrol AI's free editor + paid generation model fits.
The editor is free, runs in your browser, and stores your projects locally on your device. AI generation is metered as credits when you need it.