REWYND

Technology

Your Mac does the work.
Your footage stays with you.

The video stays on the Mac, and the AI works on the Mac.

  1. 01In productLocal-first AIAI processing on the creator’s Mac.Transcription, scene analysis, highlight selection and cutting all run on the creator’s Mac. Once the models are downloaded, a video is made with no internet connection (checked on 2026-10-05 with the network cut).
  2. 02In productPrivacyRaw footage stays on the device.Raw footage is not uploaded to a server, which is what teams handling unreleased material require.
  3. 03In productUnit economicsCustomer-owned compute, not a cloud bill per hour of footage.Heavy video processing uses customer-owned compute, rather than making every additional hour of footage an additional cloud inference bill for REWYND.

Long-term R&D

  1. 04BuildingPersonalizationLearn how each creator edits.Each channel’s way of editing becomes a profile.
  2. 05ResearchEfficiencyMake local processing faster and more efficient.A light pass over everything, deeper analysis only where it is needed.

How it works

Every step runs on the Mac

  1. 01Long raw videoHours of footage
  2. 02Local analysisA light pass over everything, on the MacSceneSpeechSilenceTranscriptSpeakerVisual change
  3. 03Candidate segmentsOnly the parts worth a closer look remain
  4. 04Deeper local analysisA closer look at what remains, still on the Mac
  5. 05ProfilePer-genre today, per-creator next
  6. 06Editing decisionKeep, trim or drop

Current research focus

  • Faster local processingResearch
  • More efficient on-device computeResearch
  • Raw-to-edit alignmentBuilding
  • Editing intent understandingResearch
  • Personalized creator profilesBuilding
  • Continuous feedback learningResearch

Models are swappable tools. What REWYND accumulates is the processing architecture, the Creator Profile, the workflow data and the feedback loop.

Why local

More editing shouldn’t mean proportionally more cloud AI cost.

A structure where company cost does not rise at the same speed as usage.

The question is not “can AI edit a video?” It is “as usage grows, who pays for the compute?”

In a cloud-based AI editor, server cost tends to rise with every video processed, and it comes back to the user as usage-based pricing. SwitchLog does the video processing on the user’s Mac.

More video, in the cloud →Cloud AI cost Usage-based AI cost Server processing cost

Usage-based cloud AI

  1. More footage
  2. More inference
  3. More API, credit and cloud processing
  4. Provider cost increases

SwitchLog

  1. More footage
  2. More local compute on the user’s Mac
  3. Raw footage remains local
  4. Provider-side video processing cost grows much more slowly

For creators

More predictable editing cost.

Using it more does not raise the price. It is a flat fee, so editing cost is predictable.

For REWYND

Better SaaS unit economics.

Designed so that server-side AI processing cost does not rise in step with usage.

Your Mac does the work.
Your footage stays with you.

Passing long video to a general multimodal AI again and again can incur inference, API and compute cost that grows with footage length and usage. SwitchLog is designed to perform the video processing itself on the user’s Mac, with the aim of avoiding a structure in which REWYND’s server-side AI processing cost rises in proportion to usage.

Usage →Provider compute costCloud-first AISwitchLog (local-first)The added compute runs on the user’s device
A diagram of the structure, not a measurement. Real figures will be published once measured.

Better models make SwitchLog better.

Better models that can run on a device are not only a threat. Models are swappable parts, and when a better one arrives SwitchLog’s local processing improves with it. REWYND’s edge is not owning a model; it is what gets decided on top of one.

  • Processing architecture
  • Creator profile
  • Workflow data
  • Feedback loop

What REWYND decides

  1. 01Which analysis runs, and in what order
  2. 02Which parts need deeper analysis
  3. 03How earlier analysis is reused
  4. 04How the Creator Profile is applied
  5. 05How a user’s corrections shape the next job

The bottleneck isn’t just cutting.
It’s repeating the same decisions.

  • Which scenes to keep
  • Where to cut
  • How to write captions
  • When to zoom in
  • Which sound effects and assets to use

Creator Profile stores those repeated decisions and reuses them.

Enterprise personalization

Building

Teach SwitchLog how your team edits.

Every channel edits differently. We turn that difference into a profile.

Given the same footage, channels keep different scenes, drop different ones, and differ in cut timing, captions, sound effects, music and how they use assets. For an individual creator, fast and easy editing comes first. Creator Editing Profile is the next step, for channels and teams.

One channel. One editing profile.

Creator Editing Profile is under development. Where a profile is stored and how it is fitted are not settled yet; the priority is that raw footage does not leave the device. The product today ships with per-genre profiles.

  • Raw footage
  • Edited result
  • User decisions

Local analysis

Raw footage and the edited result are compared on the device

Editing pattern extraction

What gets kept, trimmed and dropped

The core asset

Creator Editing Profile

A channel’s editing habits become one profile

Future local editing

New footage is edited that way, also on the device

Designed to stay on the device

Creator Profile roadmap

The profile gets more automatic over time.

Creator Profile is not a finished technology. It starts with people building it together, and moves toward the system building and correcting it on its own.

  1. Phase 1Building

    Human-assisted

    REWYND designs the profile directly, from raw footage, final edits, assets and editor feedback.

  2. Phase 2Research

    Profile automation

    The system extracts patterns from the difference between raw and edited footage: keep or drop, cuts, captions, effects, asset use, rhythm.

  3. Phase 3Research

    Feedback learning

    When a person corrects SwitchLog’s draft, the difference before and after updates the profile. The next job needs fewer corrections.

Less setup. Less correction. Better fit over time.

This does not retrain a foundation model. It updates a profile that holds that channel’s editing preferences.

Data flywheel

A profile that keeps getting better.

Not one automatic edit. A profile that keeps improving.

  1. 01Creator partnership
  2. 02Raw + edited video + assets
  3. 03Creator profile
  4. 04Auto edit
  5. 05Human correction
  6. 06Feedback data
  7. 07Better profile
  8. 08More usage

Each profile improves inside that customer’s own editing workflow. Data use is permission-based, and the priority is that raw footage does not leave the device.

For editors, MCNs & media teams

Scale content without scaling headcount.

More content, without adding headcount.

SwitchLog Enterprise helps editors and creator teams automate repeatable editing patterns and build creator-specific workflows. We are looking for pilot partners now.

  1. 01More outputThe same editing team handles more videos.
  2. 02Lower labor cost per videoLess repetitive work goes into each piece of content.
  3. 03Consistent styleEach channel’s editing rules are kept in its Creator Profile, so the style carries on when a new editor joins.

Create more with the team you already have.

How Enterprise Profile works

  1. 01Existing workflowWe learn your raw footage, final edits, assets and how you edit.
  2. 02ProfileWe analyse the cuts, choices, captions, effects and patterns that repeat.
  3. 03PilotWe apply the channel’s editing profile to new raw footage.
  4. 04Feedback loopHuman corrections are fed back to improve the profile.
  • Editors
  • MCNs
  • Creator Agencies
  • Media Teams
  • YouTube Creators

Designed so footage stays on the device

Profiles are being designed to be built on the customer’s own device wherever possible. Where footage has to be shared, it is used only for that customer’s profile, and only with explicit consent.

Become a pilot partnerTalk to REWYND

Build with us.

For investorsThe primary product, customer value, cost structure, team and what we need next, on one page.Read the investor summaryFor creators, editors & MCNsWe are looking for creators, editors and MCNs to fit a profile on a real editing workflow.Become a pilot partner