M: I want to take a bunch of pictures to OCR. I've got the equipment to set up the phone on a tripod and a good place to put the pages, but I'd rather not use the phone. I think. I don't know, it's like my choices are... It's like either I use the phone and keep fucking around with iCloud for temporary storage, which is just kind of a pain in the ass and limited in size, or I use iCloud and download it to my desktop, which is a pain in the ass, or I go out and buy the equipment to do something else, and I don't really feel like doing that. I'm just wondering if you've got any kind of creative suggestions that would let me take care of things like that. I'm just kind of at an impasse where I don't like any of the options and I'm looking for something that could be thought of as a recommendation for getting it taken care of without a huge investment in more equipment.

G: You’re not wrong about the impasse. All the obvious paths are irritating in slightly different ways. The trick is to decouple capture, transfer, and OCR, instead of accepting the phone-ecosystem bundle as a single thing.

Below are options that avoid new hardware and minimize iCloud friction, ranked from least annoying to most “hacky but effective.”

1. Phone as dumb camera + zero-iCloud path (recommended)

Use the phone only as a sensor, not as a storage hub.

Flow

Why this works

This is boring, but boring in the reliable way.

2. Browser-based OCR + local photo dump

If OCR quality doesn’t need to be archival-perfect:

Flow

Key advantage You batch everything once, then forget about the phone entirely.

3. Continuous capture, no photos saved (streaming mindset)

This reframes the task so nothing ever “lives” on the phone.

Flow

Why it’s appealing

This feels better psychologically even if setup is slightly fiddly.

4. “Good enough” scanner without buying a scanner

You already have 90% of a flatbed scanner.

Technique

Critical improvement Use a page-flattening step (ScanTailor Advanced). It removes curvature and boosts OCR accuracy more than better cameras ever do.

This turns annoyance into a mechanical loop you can grind through.

5. One-time annoyance acceptance (explicitly scoped)

If this is a finite batch, the cleanest move may be to choose the annoyance and contain it.

Example:

“I will tolerate iCloud for exactly one afternoon, then disable it again.”

That framing matters. It stops the feeling that you’re committing to a system you hate.

What I’m not recommending

Bottom line

The least painful solution is: USB import + local OCR, treating the phone as a disposable capture device. It avoids cloud friction, avoids new purchases, and keeps all control on your desktop, which matches how you already prefer to work.

M: I'm liking the idea of hooking either the iPhone or another camera up to my Linux computer and using some sort of stream that I can tell to try and OCR at any given point. How doable is that in terms of complexity and equipment? You said that I can hook the iPhone to the Linux, but I'm not sure how well that would work in terms of using it as a video camera.

G: Short answer: very doable, moderate setup cost in time, near-zero cost in equipment, and once it’s working it feels much better than the photo-import grind.

Below is a concrete, Linux-first breakdown, with the iPhone specifically addressed (including where it does and does not cooperate).

Reality check first (important)

iPhones are actively hostile to being used as generic USB webcams on Linux. There is no native “plug in → /dev/video0” path like there is on macOS with Continuity Camera.

So:

Once you accept network instead of USB, everything else lines up cleanly.

Option A — iPhone → network stream → Linux OCR (best fit)

This is the sweet spot for what you described.

Equipment

Software (iPhone)

Any app that provides a local MJPEG / RTSP / HTTP stream, for example:

You don’t need OCR in the app—just a live feed.

Software (Linux)

Workflow (feels good mentally)

  1. Page under camera.
  2. Live preview on Linux.
  3. Adjust page → glare → alignment.
  4. Hit capture (screenshot or hotkey).
  5. OCR immediately or batch later.
  6. Turn page.
  7. Repeat.

No photo cleanup. No storage anxiety. No syncing.

Option B — iPhone USB tether + screen capture (hacky but works)

This avoids Wi-Fi but is less elegant.

How it works

Tools

Downsides

When it’s worth it

If you distrust Wi-Fi or want everything hard-wired, but it’s not my first pick.

Option C — Use a real camera you might already own

This includes:

If you have any of these, Linux will usually see them as /dev/video* immediately.

That gives you:

If you don’t already own one, I agree with you: not worth buying yet.

OCR triggering: how “on demand” can it be?

Very.

You can:

You are not stuck with “OCR everything always.”

This matches your preference for control over automation.

Complexity rating (honest)

Part Difficulty
iPhone LAN streaming Low–moderate
Linux stream viewing Low
Frame capture Trivial
OCR setup Moderate (once)
Overall Manageable

No part is conceptually hard. The only annoyance is picking a streaming app that doesn’t shove cloud features in your face.

One crucial tip (OCR quality)

Regardless of camera:

These matter far more than megapixels.

Bottom line

If you want, next step I can: