Repository navigation
feat(map-inset): render and composite a route inset per regular LF - #109
Merged
Merged
Conversation
Move the keycut pin from v0.1.0 to v0.2.0. The day-master cut (extract_and_concat_segments) now calls keycut's extract_and_concat_ranges_with_receipt and checks the receipt it returns with our own load_receipt. If the receipt does not validate, the function logs an error naming the failing field and returns None, because an output-to-source time map built from a bad receipt would be wrong. The signature and return type are unchanged. keycut 0.1.2 changed the re-encode fallback to seek on the input side (-ss S -i master -ss 0 -t D). montage_builder.reencode_range_argv mirrors that argv, so it follows; the lead-seam cut gives the same frames without decoding from the start of the file. keycut 0.2.0 also probes piece and output durations for the receipt, so the clean-run command count in the resume test goes from 4 to 7. Add a real-ffmpeg test that runs keycut on a synthetic master, loads the live receipt, maps output time back to source time, and compares the receipt's keys with the frozen assembly_receipt_v1.json fixture.
# Conflicts: # gomoveshift/pipeline/montage_builder.py # pyproject.toml
What: when a drive config selects map_inset="route", each regular long-form video gets a route-only map inset rendered from an offline bundle and composited inside the existing OST encode. The LF cut now keeps its keycut receipt beside the concat (opt-in sidecar), the inset timeline is built from it shifted by the real montage lead, the inset options and route are folded into the LF OST fingerprint, and the inset receipt is recorded in drive state. Refusals happen before the day master is encoded and list every problem in one error: no bundle or a bundle that fails verification, either trim not above zero, an unparsable privacy zone, no usable day GPS trace, or photo/clip inserts on the day. A per-LF renderer refusal stops the run, naming the LF. Why: an LF must never silently ship without the map it was asked for, nor with a partial overlay. With the option off the path is unchanged. The CLI flags and config fields follow in the next commit; until then this is unreachable from a real config.
Stills are spliced into the burned LF, so the map stays on its own frames and the inserted stills carry none.
# Conflicts: # gomoveshift/pipeline/day_pipeline.py
joepetjr
added
the
needs-operator
Contributor
Author
|
Unclassified (unclassified) and ready to merge -- all checks green. Offending paths:
Human merge (all 3 ready hard-denied PRs in Back-Road-Creative/framegrade): This PR alone: |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
day_pipeline: whenconfig.drive.map_inset == "route", each regular long-form video gets a route-only map inset.map_inset.render_route_insetrenders it, andost_applier.apply_ost(map_inset=...)composites it in the existing OST encode, placed at the receipt'sinset_rect.extract_and_concat_segments(receipt_sidecar=True)saves the keycut receipt validated in chore(deps): pin keycut v0.2.0 and validate its live assembly receipt #108 beside the concat as<concat>.receipt.json. It writes to a temporary name and renames, and removes the file before each cut and on every failure. The inset usesEditTimeline.from_receipt(..., inserts=[Insert(0, probed montage lead)]). A concat that has no receipt is cut again, and if the receipt is still missing the render stops.route_claims.map_inset_inputsrefuses before the day master is encoded and lists every problem: no bundle, or a bundle that failsverify_bundle; a start or end trim that is not above 0 m (the output is public, so an untrimmed route would show where the drive began or ended); an unparsable privacy zone; no usable GPS for the day. Route points are the day's GPS in master-video seconds, split at GPS gaps.state["regular_lf_map_inset"], and the clip is purged with the other per-video work files.Why
Plan unit Q5a-3: the drive wiring for the opt-in map inset. #106 added the overlay and #107 the renderer. No config can turn this on yet; the
--map-insetflags come in the follow-up PR, so this one changes nothing for existing renders.Tests
tests/test_regular_lf_map_inset.py. It covers: off behaves like a config written before the flag existed, for both the apply_ost kwargs and the fingerprint; every pre-encode refusal fires before the encode is called; the clip is placed at the receipt rect; the timeline is shifted by the montage lead; route points are in master seconds; the receipt file is written and removed; a missing receipt triggers a second cut; a per-video refusal stops the render; inserts keep the inset; and a real-ffmpeg composite whose output pix_fmt isyuv420p10le.tools/check_public_tree.pypasses.