# iCouldPlay stop-motion observation: manually index poses and object continuity
Language: en
Canonical: [iCouldPlay stop-motion observation: manually index poses and object continuity]()
中文：[中文 HTML]()
English: [English HTML]()
Published: 2026-11-06T00:00:00+08:00
Modified: 2026-11-06T12:13:22+08:00
Author: Bison
Log poses, supports and landmarks in self-made stop motion, comparing ordinary playback and seeking while separating intentional holds, camera changes and unresolved jumps.
English translation of the [Chinese original](), translated on 2026-11-06. Original publication dates, attribution, code and version notes are retained.
Stop motion uses discrete poses, so a jump is not automatically playback stutter. Observe a self-made or authorised clip in iCouldPlay and index three poses by time and content. Use ordinary playback, pause and available seeking without assuming frame stepping, exact frame numbers, trajectory measurement or interpolation. The observations do not verify every image in the source production project.
## Choose a sample with three clear poses
Prepare a self-made or review-permitted short clip, for example a paper figure in left, centre and right poses, excluding dangerous actions. Record filename, version, duration, production explanation and sharing scope, retaining the original separately. Check beginning, middle and end for identical content identity. If production method is unknown, label it unconfirmed rather than inferring stop motion from pauses. Follow one figure or prop first; a short simple sample separates pose differences from viewing conditions more clearly than a complex sequence.
## Record the first normal-speed observation
Keep normal speed and one playback path, noting apparent holds, jumps or prop changes with nearby sound/wording anchors. Do not seek repeatedly first and mistake seek waiting for original pacing. Record orientation, output and actual playback conditions without changing engine, source and speed simultaneously. Uncertainty is valid. Continuing sound over a held image is an observation, not automatically dropped frames; the creator may intentionally hold a pose.
## Index three poses by time and state
For each pose, record nearby displayed time, orientation, limb/paper position, ground contact, held object and a separate background landmark. Use approximate times without claiming capture-frame precision. An invented left raised-hand, centre lowered-hand and right card-holding sequence needs those explicit states, not generic pose-one labels. A paused image may not be the transition boundary, so replay from several seconds before it. Mark occluded parts unseen rather than claiming a hidden hand vanished or moved to the other side.
## Separate figure changes from camera changes
Check whether table edges, fixed paper grids or background marks move with the figure. Shared displacement can indicate camera/composition change rather than the character sliding. A stable background supports a local object change, still potentially an intentional cut. Zoom/cropping also alters screen positions, so fix display conditions instead of comparing centimetres across fits. Separate observed displacement, camera interpretation and unknowns without declaring a loose capture support from footage alone.
## Check supports and props across poses
Track prop shape, holder and contact through the three poses, excluding similar but different objects. Ground contact must follow visible evidence; an airborne pose can be an intended jump, not a missing capture. Attribute stated support removal or compositing to its actual source and scope instead of guessing. Do not derive real velocity or acceleration from three states. Specialist review needs production evidence; this index identifies rewatch locations, not complete physical correctness.
## Replay from before anchors to test repeatability
Seek normally to several seconds before each anchor, let playback settle and replay at least once. Note whether the hold/jump recurs at the same content, without treating the first post-seek image as precise measurement. A one-off remains occasional; stable recurrence can belong to the source and does not alone prove player fault. Compare the identical original in another trusted existing tool if needed, fixing speed and target. A re-export is a different file, not a single-variable comparison.
## Leave and recover the indexed poses
Leave the player and reopen the same entry, verifying file identity and duration before finding all three poses by time and landmarks. Restate figure, prop and support at each to exclude a similar action. Record seek uncertainty and unresolved boundaries instead of inventing exact frame numbers. Keep the original clip and text index; screenshots capture observations, not an entire motion sequence. Rematch after edits or revisions because old times may fail, then restore ordinary viewing position.
## Conclude with observation, not automatic diagnosis
Keep three poses, time anchors, background references, first/repeat observations and unknown causes. Discrete poses may belong to the source, recurrence still requires identical-file path comparison, and ordinary seeking supplies no exact inter-frame measurement. State stable observations and the one point needing creator/source review. Do not change speed, clear a library or claim automatic interpolation merely to eliminate every visible jump.
## Related tutorials and sources
[Inspect held images](), [Prepare a minimal report]().
Checked on November 6, 2026 against the [official iCouldPlay App Store listing]() and [Apple public lookup](): version 1.1.1, requiring iOS / iPadOS 15.0 or later. The public 1.1.1 release notes list Kodi, OpenList / AList and S3-compatible storage, with improvements to source switching and captions after seeking. The tasks use this public scope and existing playback capabilities. Check controls in your installed version. Self-made samples, manual records and external file preparation are workflows, not newly promised automatic features. No physical-device or private-server test is claimed.