Live Camera Visual Intelligence falls back to Camera while Siri AI and Screenshot VI actually work

[Ssiswent]

Summary

I want to separate this from #101 because I think my case was misunderstood there.

This is not a case where Siri / Apple Intelligence UI merely appears while the underlying features do not work.

On my mainland China (CH/A) iPhone 17 Pro, Siri AI was enabled through GestaltEdit and the underlying features are genuinely functional:

Siri AI                               ✅ working
ChatGPT integration                   ✅ working
Screenshot / on-screen Visual Intel. ✅ working
Live Camera Visual Intelligence       ❌ not working

Only the live-camera Visual Intelligence path fails.

A more detailed investigation is documented here:

frs0n/GestaltEdit#26

Environment

• Device: Mainland China / CH/A iPhone 17 Pro
• Physical Camera Control: Yes
• iOS: 27.0 Beta 4
• Build: 24A5390f
• Siri AI / Apple Intelligence: actually working
• ChatGPT integration: working
• Screenshot / on-screen Visual Intelligence: working
• Live Camera Visual Intelligence: not working

Reproduction

Camera Control

1. Start from the Home Screen or Lock Screen.
2. Long-press the physical Camera Control button.
3. A Visual Intelligence / rainbow-style transition appears.
4. It then opens the regular Camera UI, instead of the live Visual Intelligence camera interface.

Control Center

1. Tap the Visual Intelligence control in Control Center.
2. It also opens the regular Camera instead of live Visual Intelligence.

Screenshot Visual Intelligence

On the exact same device:

1. Take a screenshot containing recognizable content.
2. Open screenshot / on-screen Visual Intelligence.
3. Visual Intelligence works normally.

So Visual Intelligence itself is not generally broken on this device.

Video

Screen recording showing the issue on the affected device:

ScreenRecording_08-17-2026.12-01-04_1.MP4

The recording demonstrates that:

• Long-pressing Camera Control triggers the Visual Intelligence-style animation, but opens the regular Camera UI.
• Launching Visual Intelligence from Control Center also falls back to the regular Camera.
• Screenshot / on-screen Visual Intelligence works on the same device.

Region-spoof A/B test

Initially, I enabled Siri AI by manually changing only:

h63QSdBCiT/z0WU6rdQv6Q
CH → LL

At that point:

Siri AI                         ✅
Screenshot Visual Intelligence ✅
Camera Visual Intelligence     ❌

To rule out incomplete region spoofing, I then used GestaltEdit's full:

Enable Siri AI (US Region)

flow, applied the changes, rebooted the device, and verified that all expected values were successfully written:

h63QSdBCiT/z0WU6rdQv6Q = LL
yK+xavymRGZ3xWc1tb8XDg = LL/A
97JDvERpVwO+GHtthIh7hA = A3256

A3256 is the US regulatory model used by GestaltEdit for iPhone 17 Pro.

After the full US-region configuration, the result remained exactly the same:

Siri AI                         ✅
ChatGPT integration             ✅
Screenshot Visual Intelligence ✅
Camera Visual Intelligence     ❌

So the issue does not appear to be caused simply by incomplete CH → US MobileGestalt region spoofing.

Additional investigation

I also checked several potentially relevant MobileGestalt keys.

The following keys are not explicitly present in CacheExtra on this native Camera Control device:

CwvKxM2cEogD3p+HYgaW0Q
// CameraButtonCapability

lyJZrSDc8J8eQ5Zb7A1Rvw
// DeviceSupportsTouchSensitiveCameraControl (iOS 27)

I did not manually add them because absence from CacheExtra does not necessarily mean the capability is false. On native Camera Control hardware, these values may be dynamically derived by MobileGestalt instead of being stored as explicit CacheExtra overrides.

The old region-bypass key:

zHeENZu+wbg7PUprwNwBWg

is also completely absent on this iOS 27 device.

Why this is different from #101

In #101, you wrote:

we cant write to feature flags and therefore only make the Siri and apple intellegence related UIs appear, not actually use them.

That does not describe my setup.

The underlying Siri AI functionality really works.

ChatGPT integration really works.

Screenshot / on-screen Visual Intelligence really works.

Only the live-camera Visual Intelligence path is blocked.

This is why I suspect there may be a separate check specifically for the camera-based Visual Intelligence path.

Current hypothesis

There may be an additional gate affecting live-camera Visual Intelligence, possibly involving:

• a Feature Flag;
• an eligibility / availability check;
• a country / region eligibility state;
• a device eligibility cache;
• a separate Camera / Visual Intelligence capability;
• or launch routing that falls back to Camera.app when the Visual Intelligence camera path is considered unavailable.

The exact root cause is still unknown.

Potentially relevant iOS 27 clue

While investigating the iOS 27 Visual Intelligence stack, I found that the Visual Intelligence app / frameworks (Tamale.app, VisualIntelligenceCore, VisualIntelligenceUI) reference GenerativeModels availability reasons including:

regionIneligible
countryLocationIneligible
deviceNotCapable
noCacheForDeviceEligibility

This makes me wonder whether the live-camera Visual Intelligence path performs an additional availability / eligibility check that screenshot Visual Intelligence does not perform in the same way.

For example, the current behavior could potentially look like:

Siri AI
    → available ✅

Screenshot Visual Intelligence
    → available ✅

Live Camera Visual Intelligence
    → additional availability / eligibility check
    → fails
    → falls back to Camera.app ❌

This is only a hypothesis, not a confirmed diagnosis.

Request

Could you please take another look at this specific case and, if possible, try to identify what blocks the live-camera Visual Intelligence path?

It would be especially helpful if you could check whether mond or your previous research has identified:

1. a Feature Flag that controls live-camera Visual Intelligence;
2. an eligibility / availability value used by Tamale / VisualIntelligenceCore;
3. an eligibilityd or GenerativeModels availability domain specifically involved in Camera VI;
4. a way to log the actual unavailable reason when live-camera Visual Intelligence is launched;
5. or a tweak / patch that could make the Visual Intelligence camera route launch instead of falling back to Camera.app.

Since Siri AI and screenshot Visual Intelligence are already fully functional, I think this may be a relatively isolated remaining gate rather than a general Apple Intelligence activation problem.

I am happy to:

• run targeted A/B tests;
• compare MobileGestalt values;
• provide logs;
• inspect specific keys;
• or test experimental changes / builds if you can point me toward a specific check.

Any help investigating this would be greatly appreciated.

[000ww]

In fact, the screenshot recognition of visual intelligence is functioning properly.
By accessing the file directory of Visual Intelligence through the 3105 app, you can view the screenshot in JPEG format.