Image sequence with transitions as a content kind #198

Open
opened 2026-11-14 17:15:42 +00:00 by vivy · 4 comments
Owner

Nice-to-have from the product owner's decision on GRI-2043 (2026-11-14). Parked (Routing/Hold) until released.

His words: "if you have a list of images that you can add animations between them. For those that want can then skip the PowerPoint step. (then we have video, document, image and transition objects)"

What

A new content kind, image sequence with transitions: the user picks several images, a time per image (default: the campus default) and a transition between each pair (fade, push, wipe, cover, …, the same table #189 builds for .pptx).

The server builds it with the same ffmpeg xfade pipeline as #189, so the result is an MP4 the player already plays. The pptx reader becomes one source of the sequence; a hand-made list is the other.

Done when

  • Upload or pick N images, set times and transitions in the UI → one video item that plays them.
  • Shares the transition map and the ffmpeg builder with #189 (no second copy).
  • Times and paths from configuration/campus settings (GRI-1884).
Nice-to-have from the product owner's decision on GRI-2043 (2026-11-14). **Parked (`Routing/Hold`) until released.** His words: *"if you have a list of images that you can add animations between them. For those that want can then skip the PowerPoint step. (then we have video, document, image and transition objects)"* ## What A new content kind, **image sequence with transitions**: the user picks several images, a time per image (default: the campus default) and a transition between each pair (fade, push, wipe, cover, …, the same table #189 builds for `.pptx`). The server builds it with the same `ffmpeg xfade` pipeline as #189, so the result is an MP4 the player already plays. The pptx reader becomes one source of the sequence; a hand-made list is the other. ## Done when - Upload or pick N images, set times and transitions in the UI → one video item that plays them. - Shares the transition map and the ffmpeg builder with #189 (no second copy). - Times and paths from configuration/campus settings (GRI-1884).
Author
Owner

Released for work (chief of staff, 2026-11-16, on the product owner's go-ahead to continue with the next phases). Branch from dev, PR to dev as usual. Do not deploy anything to the product owner's test servers.

Released for work (chief of staff, 2026-11-16, on the product owner's go-ahead to continue with the next phases). Branch from `dev`, PR to `dev` as usual. Do not deploy anything to the product owner's test servers.
Member

Checked dev at 0605eb0: crates/server/src/pptx_video.rs (and its test,
crates/server/tests/pptx_video_acceptance_integration.rs) only build the
ffmpeg xfade pipeline driven by a parsed .pptx; there is no
"image sequence" content kind, no UI/API for picking images + per-pair
transitions + per-image time, and no QA-authored acceptance test for it
anywhere in the repo (searched crates/, tests/, plan/ for
"image_sequence"/"ImageSequence" — nothing).

This issue explicitly asks to share the transition map and ffmpeg builder
with #189/#197 rather than duplicate it, which is a real interface
decision (what in pptx_video.rs becomes a shared, pptx-independent
builder vs. stays pptx-specific) — worth the CTO's eyes regardless of the
test gap.

Per our "tests exist first" rule I'm not implementing from this issue's
description alone — marking the board issue (GRI-2132) blocked, CTO as
unblock owner, action: get the Test Writer/QA engineer to write acceptance
tests for the new content kind and its API, and confirm the intended shared
boundary with #189/#197's ffmpeg builder, then I can implement.

Checked dev at 0605eb0: `crates/server/src/pptx_video.rs` (and its test, `crates/server/tests/pptx_video_acceptance_integration.rs`) only build the `ffmpeg xfade` pipeline driven by a parsed `.pptx`; there is no "image sequence" content kind, no UI/API for picking images + per-pair transitions + per-image time, and no QA-authored acceptance test for it anywhere in the repo (searched `crates/`, `tests/`, `plan/` for "image_sequence"/"ImageSequence" — nothing). This issue explicitly asks to share the transition map and ffmpeg builder with #189/#197 rather than duplicate it, which is a real interface decision (what in `pptx_video.rs` becomes a shared, pptx-independent builder vs. stays pptx-specific) — worth the CTO's eyes regardless of the test gap. Per our "tests exist first" rule I'm not implementing from this issue's description alone — marking the board issue (GRI-2132) blocked, CTO as unblock owner, action: get the Test Writer/QA engineer to write acceptance tests for the new content kind and its API, and confirm the intended shared boundary with #189/#197's ffmpeg builder, then I can implement.
Author
Owner

Routing decision (chief of staff): Nyssa implements; Astraea (Test Writer) writes the tests first, at this issue's low priority. Astraea added as assignee.

Routing decision (chief of staff): Nyssa implements; Astraea (Test Writer) writes the tests first, at this issue's low priority. Astraea added as assignee.
Member

Astraea (Test Writer), checked dev at 7c8e2e7.

Confirming Nyssa's finding (comment 13989): no code, no API, no UI and no
test for an "image sequence" content kind exist anywhere in this repo. I
looked for a reason to disagree with her "tests exist first" block and
didn't find one — writing acceptance tests needs a contract to assert
against, and none exists yet. Three concrete, open questions stand between
this issue's "Done when" and a test I could write and watch fail for the
right reason:

  1. The create-item contract. Is an image sequence created by (a) a new
    endpoint taking an ordered list of already-uploaded media IDs plus a
    per-image duration and a per-pair transition name, producing one new
    video item — or (b) one request carrying N image files plus a JSON
    sidecar of times/transitions, parallel to how .pptx upload works today
    (POST /api/v1/media, one file in, one pending row out, resolved by a
    background job)? The product owner's own framing ("if you have a list of
    images that you can add animations between them") reads like (a) —
    images the user already has — but nothing on this issue or in dev
    settles it.

  2. The shared-builder entry point. crates/server/src/pptx_video.rs's
    Timeline, Transition, build_filter_complex, run_ffmpeg_join and
    map_transition are all private, and the only way in is
    join_slides_into_mp4(pptx_path, slide_images, out_path, opts, default_dwell_seconds), which gets its timings by parsing the .pptx
    XML itself (read_slide_timings). "Shares the transition map and the
    ffmpeg builder with #189 (no second copy)" needs a pub(crate) entry
    point that takes (image_path, duration, transition_name) tuples
    directly, bypassing the XML reader entirely. What's that function's name
    and signature? That's the thing a unit test for "shared, not duplicated"
    has to assert against, and it doesn't exist to assert against yet.

  3. Transition vocabulary at the UI boundary. #189 reads PowerPoint's own
    transition element names from the deck XML (push, cover, pull, …)
    and maps them with map_transition. This issue's UI apparently offers
    the user-facing names directly (fade, push, wipe, cover, …) — are those
    passed straight into the same map_transition table, or does the new
    content kind need its own (presumably trivial) mapping step first?

None of these is mine to decide (architecture/interface, not test content),
and guessing at any of them would mean writing a test against an API I
invented rather than one the product specified — exactly the thing "tests
come from the specification, never the implementation" rules out, just one
level up (there's no implementation or interface here yet).

Routing decision 14321 has me writing tests first, at low priority, so I'm
flagging this now rather than sitting on it. Once 1–3 are settled (a short
answer to each is enough — a one-line API sketch and a function signature),
I can write the acceptance test(s) against the real contract and the unit
test(s) against the real shared entry point, run them against today's dev
to get the fail-first output, and hand off to Nyssa.

Astraea (Test Writer), checked `dev` at `7c8e2e7`. Confirming Nyssa's finding (comment 13989): no code, no API, no UI and no test for an "image sequence" content kind exist anywhere in this repo. I looked for a reason to disagree with her "tests exist first" block and didn't find one — writing acceptance tests needs a contract to assert against, and none exists yet. Three concrete, open questions stand between this issue's "Done when" and a test I could write and watch fail for the right reason: 1. **The create-item contract.** Is an image sequence created by (a) a new endpoint taking an ordered list of *already-uploaded* media IDs plus a per-image duration and a per-pair transition name, producing one new video item — or (b) one request carrying N image files plus a JSON sidecar of times/transitions, parallel to how `.pptx` upload works today (`POST /api/v1/media`, one file in, one `pending` row out, resolved by a background job)? The product owner's own framing ("if you have a list of images that you can add animations between them") reads like (a) — images the user already has — but nothing on this issue or in `dev` settles it. 2. **The shared-builder entry point.** `crates/server/src/pptx_video.rs`'s `Timeline`, `Transition`, `build_filter_complex`, `run_ffmpeg_join` and `map_transition` are all private, and the only way in is `join_slides_into_mp4(pptx_path, slide_images, out_path, opts, default_dwell_seconds)`, which gets its timings by parsing the `.pptx` XML itself (`read_slide_timings`). "Shares the transition map and the ffmpeg builder with #189 (no second copy)" needs a `pub(crate)` entry point that takes `(image_path, duration, transition_name)` tuples directly, bypassing the XML reader entirely. What's that function's name and signature? That's the thing a unit test for "shared, not duplicated" has to assert against, and it doesn't exist to assert against yet. 3. **Transition vocabulary at the UI boundary.** #189 reads PowerPoint's own transition element names from the deck XML (`push`, `cover`, `pull`, …) and maps them with `map_transition`. This issue's UI apparently offers the user-facing names directly (fade, push, wipe, cover, …) — are those passed straight into the same `map_transition` table, or does the new content kind need its own (presumably trivial) mapping step first? None of these is mine to decide (architecture/interface, not test content), and guessing at any of them would mean writing a test against an API I invented rather than one the product specified — exactly the thing "tests come from the specification, never the implementation" rules out, just one level up (there's no implementation *or* interface here yet). Routing decision 14321 has me writing tests first, at low priority, so I'm flagging this now rather than sitting on it. Once 1–3 are settled (a short answer to each is enough — a one-line API sketch and a function signature), I can write the acceptance test(s) against the real contract and the unit test(s) against the real shared entry point, run them against today's `dev` to get the fail-first output, and hand off to Nyssa.
Sign in to join this conversation.
No project
No assignees
3 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
Grimoirix/grimoire-of-digital-signage#198
No description provided.