Pocket 3 exposure, focus and audio

Part of the Osmo Pocket 3 reference. Read the overview for firmware, survey scope and evidence levels.

Mode-specific controls

Control	Video	Low-Light	Slow Motion at 1080P/240
Pro	Advanced settings visible when enabled	Disabling hides the advanced list, leaving Pro and Grid	Advanced settings visible
Focus	Single, Continuous, Product Showcase Mode	Single, Continuous; no Product Showcase entry	Single, Continuous; no Product Showcase entry
White balance	Auto, Custom; 2000–10000 K endpoints selected	White Balance row visible	White Balance row visible
Color	Normal, DLog-M 10bit, HLG	No Color row in the inspected full list	No Color row between White Balance and Channel
Color Recovery	Toggle appears with D-Log M; absent in the inspected Normal and HLG states	No row	No row
Built-in audio	Mono/Stereo; wind reduction; All/Front/Front and Back	Same rows visible	Same rows visible; visibility does not prove an encoded audio track
Monitoring	Grid, Overexposure Alert, Histogram, Timecode Display	Same rows visible	Same rows visible

At 4K/120 Slow Motion, the Color row reappeared with Normal, DLog-M 10bit and HLG. D-Log M and HLG selections were accepted. D-Log M added a Color Recovery toggle; HLG removed it. The comparison above changed both resolution and rate, so it does not isolate the cause to 240 alone. Short 1080P/240, 4K/120 D-Log M and 4K/120 HLG takes had accepted start/stop commands. Their independently inspected original playback rates, color tags and audio streams are listed in the original media reference.

In Video, enabling Histogram displays a histogram widget over the preview; disabling it removes that widget. The observed monitor-toggle actions did not have a camera write classified by the packet analysis. That is not proof that they can never affect camera traffic. Color Recovery is a Mimo display control; this survey does not establish its transform or a change to recorded pixels. Low-Light’s Grid menu offered Off, Gridlines and Grid + Diagonals; selecting each visibly changed or removed the corresponding preview overlay. The Video monitor’s mirror control also visibly reversed the preview and restored it on the next toggle. The effect on saved media was not tested.

Sharpness, noise reduction and focus breathing compensation were not found in the inspected Video settings list. Their absence from this list does not prove that the camera lacks the feature or that no body-side control exists.

Focus sequencing

Single and Continuous use 0x02/0x24 values 01 and 02. Accepted replies and cam_lens_state values B1/B2 corroborate those choices. Mimo also manages AF-C tracking parameter 0x003B through 0x02/0x8E:

Action	Observed sequence
Choose Single	Clear the tracking parameter with value 01 00, then set focus 01.
Choose Product Showcase from Single	Set tracking value 01 01; camera lens state changes to B2 without a separate Continuous write in that action.
Return to Continuous	Clear tracking, then set focus 02.

These sequences were accepted. A near/far subject test is still required to judge focus behavior; other write orders have not been shown to fail.

White balance and exposure

The existing five-byte white-balance request is confirmed:

0x02/0x2C: [mode:u8] [Kelvin / 100:u16le] [trailing:i16le]
Custom mode = 06; Auto mode = 00

Mimo selected 2000 K and 10000 K in Custom, then restored Auto. Requests were accepted and mode/custom Kelvin were corroborated by cam_image_effect. All trailing request values in this sweep were zero. The signed value at status bytes 7–8 nevertheless changed, so interpreting those bytes as the operator’s tint is not confirmed on Pocket 3. Auto’s displayed Kelvin is also distinct from a requested Custom value. Do not write a changing Auto measurement back as a manual setting.

Low-Light’s manual ISO dial offered and selected 50, 100, 200, 400, 800, 1600, 3200, 6400, 9600, 12800 and 16000. Auto exposure showed an ISO MAX endpoint of 16000. These are UI results, with additional accepted requests and exposure status corroborating ISO 9600 (0x02/0x2A value 10), ISO 16000 (value 11) and the 1/8000 shutter endpoint at 4K/30. The ISO encoding is sparse; do not derive these added values by extending the earlier power-of-two sequence. In fixed-ISO manual mode, the ISO MAX row disappeared and the EV adjustment buttons appeared dim. This is not a measured shutter or exposure calibration.

Glamour Effects

The Video monitor has a separate Glamour Effects panel, with None and a horizontally scrolling list: Smooth, Brighten, Slim, Eyes, Dark Circles, Nose, Mouth, Teeth, Lipstick, Blush and Brows. Selecting None displayed OFF; selecting an effect displayed ON and exposed its numeric slider.

The corrected slider sweeps established these UI values:

Effects	Observed values
Smooth, Brighten, Eyes, Dark Circles, Nose	0 and 99 reached; the upper endpoint was not independently established
Slim, Teeth, Lipstick, Blush, Brows	0 and 100 at the slider endpoints
Mouth	−50 and +50 at the endpoints, with 0 at the center

The Mouth control therefore needs a signed UI range. The other observed sliders placed zero at the left edge. Their wire values use different scales.

Tagged strength values

Parameter 0x0039 of 0x02/0x8E carries a 62-byte blob: a little-endian count of 15, followed by 15 four-byte entries. Each observed entry contains a little-endian u16 tag, a preserved 01 marker, and a u8 value. The marker may represent a length; its general semantics are not established. Camera GETs reorder entries by tag while Mimo SETs use UI order, so compare tagged values rather than relying on one position for each effect.

The controlled slider writes were accepted, with independent camera GETs corroborating their final values. The following numbers are decimal:

Control	Tag	Observed UI → wire pairs
Master	0	Off → 0; On → 1
Smooth	1	0 → 0; 25 → 20; 99 → 79
Brighten	2	0 → 0; 25 → 25; 99 → 99
Slim	3	0 → 50; 50 → 70; 100 → 90
Eyes	4	0 → 50; 99 → 90
Nose	5	0 → 50; 99 → 90
Mouth	6	−50 → 0; 0 → 50; +50 → 100
Teeth	7	0 → 0; 100 → 100
Lipstick	10	0 → 0; 20 → 20; 100 → 100
Blush	11	0 → 0; 100 → 100
Dark Circles	12	0 → 0; 99 → 99
Brows	14	0 → 0; 100 → 100

Tags 8, 9 and 13 had no identified UI control and remained zero. Preserve them; neither a general rounding rule nor a legal range for unknown tags follows from this sample. Restoring the controls and selecting None returned all 15 tagged values to their initial camera-reported state.

This live-camera Glamour test scene did not contain a face, so its facial processing and effect on saved pixels remain unverified. The recorded Glamour-on take had Smooth at zero. Its immediate ancillary glamour SET returned DF, while recording itself succeeded. The separate local editor tests do not establish an equivalent camera effect or command.

Audio DSP: preserve the Pocket 3 blob

The Pocket 3 returned a 27-byte audio DSP blob after the status byte in 0x02/0xA0 GET replies. Mimo’s accepted 0x02/0x9F SETs preserved all 27 bytes. This differs from the 26-byte form previously documented for other captures.

Mimo action/state	Observed blob byte 2
Wind reduction off → on, Front and Back selected	BC → BD
Wind on, direction All	1D
Wind on, direction Front	3D
Wind on, direction Front and Back	BD

Selecting Front also produced a preceding accepted SET changing blob byte 0 from C0 to 80. The observations do not justify replacing the entire blob, assuming one universal wind mask, or applying another model’s 18/1A and DA/3A/BA table. Read the current blob, preserve its length and unknown fields, and qualify any mutation by model and state. Acoustic direction, wind filtering and audio content require separate measurements. Stream/channel counts for the preserved files are recorded in the original media reference.

The current AudioDspBlob.blob(fromGetReply:) implementation copies exactly 26 bytes. It would truncate this Pocket 3 response; this reference records the gap and does not claim the implementation is fixed.