Frozen inputs A–F
BYTE-IDENTICALA is an uploaded 1000×558 JPEG. B–F are 1392×752 generated PNGs (aspect 1.851, not 16:9). Sequence A → B → C → D → E → F. F is an open debris void around a central light, not a corridor.
Terran's original/manual workflow established spatially contiguous travel, continuous locomotion, threshold/opening traversal, anti-dissolve / anti-morph prompting, motion-conditioned endpoint preparation, and human review of generated clips. This Camotion implementation and the agentic pipeline are not Terran's code.
This is not Camotion retuning, not Director replanning, and not destination regeneration. The exact six forest stills from the live runtime session were frozen. One 01.8 shooting frame per canonical. Shared A′–F′ across adjacent legs. One locomotion prompt. Seedance 2.5. Hard concat only.
Five 6s legs hard-joined after a same-size H.264 transcode. No crossfades, titles, grading, or speed ramps. A→B is letterboxed to 1306×706 so mixed Seedance adaptive sizes can play. Human review of this movie was positive.
A is an uploaded 1000×558 JPEG. B–F are 1392×752 generated PNGs (aspect 1.851, not 16:9). Sequence A → B → C → D → E → F. F is an open debris void around a central light, not a corridor.
Minimum conservative radial-forward CameraMotionPlan v1 per canonical. The same B′ is A→B end and B→C start, and so on through F′. Current Camotion has no yaw/strafe; F′ is still radial-forward toward the light.
Local evidence keeps exactly five clips. Replicate Recent predictions showed eight succeeded jobs. Cause: overlapping --execute processes after auto-review retries, not prompt edits or quality reruns of kept files. A→B / B→C prediction ids were lost when the first process died. Kept C→D / D→E / E→F ids: 0w3bchvse5rmt0d0fj7aa95gb4, qh6y0dg1fnrmy0d0fj99m8hkfw, 3yvf8sk8zdrmt0d0fjav5hrspc.
Stream-copy of mixed resolutions (A→B 1284×716 vs later 1306×706) froze around 8s. The review movie was replaced in place with filter concat + scale/pad. No extra Seedance calls. No seventh local movie kept.
A→B is the strongest locomotion. B→C is the same tunnel, deeper. C→D pulls to the crystal early and does not hold C′ at t=0. D→E shows the next space through an opening before the previous tube is gone. E→F replaces the corridor with a void/warp. Camotion helps corridor stills and hurts F, where radial blur on an open void looks like a stargate cue. F is not spatially earned from A. Full assessment: REPORT.md.