宜博 / h3livestream

Records the 2026-09-02 real-H3 verification of T1-T10 (11 videos / 10
garments, the barefoot reference-video leak finding, the two bugs
found and fixed) and cross-references that T11-T17 (dev-plan-2.md)
landed the same day from a separate session, so the docs don't read
as stale the next time someone opens them.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011Deddn3sM3oQnmvqWSMb4y

…ce refs, 11 generated videos

Real (non-test) content for T1-T10's asset libraries, entered through
the actual admin CRUD flow (admin-characters/garments/dance-refs.html
endpoints) rather than test fixtures: 1 default character, 10 garment
variants, 3 reference dance videos. Batch-generated through T9's
/api/admin/batch-gen against the real H3 Console — 11 dance videos
landed in library.json (one combo generated twice from a dedupe race
before the edited:true fix), all with characterId/garmentId/
refDanceVideoId traceability.

Quality note found while spot-checking: 3 of the 11 clips show bare
feet instead of the character reference's boots, and all three trace
back to the same dance-ref ("轻柔手势舞", sourced from the S1-spike's
control clip) — the reference video's own footwear leaks through
despite the prompt's explicit "weak_reference" instruction. Worth
avoiding barefoot reference footage when curating real dance-refs.

This doesn't yet hit dev-plan.md's phase-1 exit bar of >=30 videos
(that requires 30 garments given the M.7 per-character-garment dedupe
cap; this seeds 10) but is enough real content for dance.html's
wardrobe to have something to show.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011Deddn3sM3oQnmvqWSMb4y

H3's 4 GPUs run one sharded model (H3-CONSOLE.md /v1/status), not 4
independent workers, so a batch of N jobs is served ~serially. The
poll loop's old fixed 200-iteration budget (~1000s at the 5s default
interval) didn't account for queue depth: submitting a 9-job batch hit
this live and 9/11 jobs got marked "generation timed out" locally
while still sitting healthy in H3's queue (queue_position confirmed
via GET /v1/videos/:id). Also treats H3's "cancelled" status as
terminal, matching "error"/"failed".

Recovered the wrongly-failed jobs by resetting them to "generating" in
generation-jobs.json (their h3JobId was already set, so
resumeDressupGenerations() picked them back up without resubmitting to
H3) rather than resubmitting and burning the queue further.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011Deddn3sM3oQnmvqWSMb4y

ingest.js's syncLibrary() only carries a library entry forward as-is
when edited is true; otherwise its periodic auto-sync (every 5 min)
rebuilds the record from the raw H3 job and silently drops
channel/characterId/garmentId/refDanceVideoId. That broke the M.7
dedupe cache and would have let dance clips leak into the regular
live-channel pool once channel got wiped. Caught while running a real
batch generation: the second sync cycle stripped a clip's traceability
fields within 5 minutes of it landing.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011Deddn3sM3oQnmvqWSMb4y

Backend (server.js):
- T11: dance.html gets its own schedule/queue state machine, persisted to
  dance-state.json (gitignored, same as live-state.json). Per MVCP M.8,
  the schedule stores videoId references rather than inlined clip
  snapshots like live's does, joining back to library.json at read time.
  Generalized pickBestIndexByStatus/extracted pickLeastPlayedClip so both
  channels' schedulers share the same selection logic instead of forking
  it; live's random-fill pool now excludes channel:"dance" clips (and
  vice versa) so the two channels' filler picks stay independent, which
  wasn't true before this (T8's dance clips could have leaked into
  live.html's filler).
- T12: GET /api/dance/wardrobe — public, unauthenticated list of active
  garments.
- T15: POST /api/dance/pick {garmentId, nickname?} — free path, always
  uses the configured default character, reuses submitGeneration (T8/T9)
  whole including its M.7 dedupe cache, then hands the result to
  danceQueue. In-flight generations are reconciled against generationJobs
  on each schedule tick rather than via a promise chain, so a pick
  request survives a server restart the same way T8's own jobs do.
- T17: separate danceViewers map and POST /api/dance/queue/vote/:id, so
  viewer counts and votes don't mix with live.html's.
- SCHEDULE_TICK_MS env var makes both channels' tick interval
  configurable so tests don't wait out a real 2s.

Frontend:
- dance.html (T13/T14/T16): reuses live.html's visual language via
  live.css, adds dance.css for the 9:16 boxed player and wardrobe grid.
  Left column is a 衣柜/传衣服/传人 tab set (only 衣柜 is wired up —
  the other two need accounts/credits, dev-plan-2.md 第三/四期); right
  column shows the queue with vote buttons instead of live.html's chat.
  Age-confirmation gate (5.1节) blocks the page on first visit via a
  cookie, non-dismissable by clicking the backdrop.

Tests: dance-channel.test.js spins up a mock H3 Console (same approach
as dressup-generate.test.js) to exercise wardrobe listing, pick
rejection/happy-path/dedupe, vote, and independent viewer counts
end to end.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Breaks down dev-plan.md's phase-2 scope (dressup-dance-room.md 9节
第2条: dance.html + 免费衣柜模式 + 去重缓存, no paid modes/accounts/
credits yet) into T11-T17 with acceptance criteria, work estimates,
and a dependency graph, same format as dev-plan.md's T1-T10.

Notes that dev-plan.md's own phase-1 exit criterion (batch-generate
>=30 seed videos with real assets) is still outstanding and is an
operational step, not blocking this plan but blocking the wardrobe
actually having content once T11-T17 land.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011Deddn3sM3oQnmvqWSMb4y

Fetches characters/garments/dance-refs alongside the library on load
and joins each clip's characterId/garmentId/refDanceVideoId back to
their names, so an operator can see "谁的脸 + 谁的衣服 + 哪条参考视频"
per row (M.7). No composedImageId column — the one-step pipeline
never produces an intermediate composed image.

Extends the dressup-generate.test.js happy-path test to confirm the
join resolves real names, not just matching ids.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

POST /api/admin/batch-gen takes {characterId, garmentIds[],
refDanceVideoIds?[]}, expands to one job per garment (cross-producted
with refDanceVideoIds when given, else one random active dance-ref per
garment like the single-generate default), and submits each through
the same submitGeneration() helper the single POST /api/admin/generate
route now uses — so the M.7 dedupe cache (existing characterId+
garmentId combo) is honored per combo, not just per batch.
GET /api/admin/batch-gen/status?batchId= returns that batch's
generationJobs records for progress polling.

admin-batch-gen.html: pick one character, multiple garments, and
optionally pin specific reference dances; submits the batch, polls
status every 3s into a progress table, and lets failed rows be
retried individually via POST /api/admin/generate. Linked from the
other admin pages' nav.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

generateDanceVideo() takes a queued generationJobs record through
prompt build -> forward character/garment/dance-ref uploads to H3
Console's /v1/assets -> POST /v1/videos -> poll -> push a traced clip
into library.json (characterId/garmentId/refDanceVideoId). Mirrors
generateForPrompt's submit/poll/persist shape but keyed off the new
generationJobs store so a restart resumes any in-flight job by its
h3JobId, same as resumeGenerations() does for the live queue.

POST /api/admin/generate accepts characterId/garmentId (+ optional
refDanceVideoId, else a random active dance-ref is picked) and skips
straight to the cached clip if that character x garment combo was
already generated (M.7 dedupe). GET /api/admin/generate/:id polls job
status.

Tests spin up a mock H3 Console over HTTP so the full submit/poll/land
path runs offline. Also fixes two existing test fixtures
(dressup-crud, admin-pages-flow) that spawn server.js without copying
prompt-template.js — worked until T8 added the first require() of it
into server.js, at which point the spawned server silently failed to
boot.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

buildStage2Prompt() renders MVCP M.15's confirmed one-step six-section
template and enforces the two dynamic risks M.15.1 flagged: caps
appearanceDescriptor/garmentDescriptor at ~40 words each, and
truncates motionDescriptor if it alone would push detailed_description
past H6's 500-word cap (the only placeholder that lands in that
section).

Also fixes the footwear-drift the S1 spike found (model self-selected
shoes when the prompt didn't constrain them) — pins footwear to
<Picture 1> in both subject_definitions and retention_analysis, and
says <Picture 2> doesn't affect it. Propagated the same fix into the
M.15 template in the MVCP doc so the doc and the code don't drift.

Tests check the mechanical M.15.1 items that are checkable from the
rendered string (H1/H5/H6/H9) plus the truncation behavior and the
footwear fix. H13 (reference counts) is a submission-payload concern,
covered by T8's tests instead.

Table + inline edit + add-form pages matching admin.html's existing
conventions, wired to the T3-T5 CRUD API and T2's upload endpoints.
Cross-linked with admin.html and each other in the topbar.

No browser tooling available in this environment to render/click
through these, so admin-pages-flow.test.js exercises the exact API
call sequence each page's JS performs (login -> upload a real file ->
create with the returned URL -> list -> confirm the shape the table
renders), which is where a field-name mismatch would surface.

POST /api/admin/upload-video: mp4 magic-byte sniff (checks the ftyp
box, not offset-0 like images), 50MB cap, no re-encode needed (unlike
images, no EXIF/GPS concern for video). Reuses the multipart parser
and uploads/ layout from T2's image path via a shared placeInUploads
helper.

mountCrud() is a small generic GET-list/POST-create/PATCH/DELETE
router shared across the three entities (same shape, see MVCP
M.3/M.4/M.5) instead of copy-pasting the boilerplate three times.
Each entity fills in its own admin_seed-appropriate defaults on
create. characters additionally enforces the single-default-character
invariant (2 节) across create/update.

HTML admin pages for these come next.

Single shared password (.env ADMIN_PASSWORD) + in-memory HttpOnly
session cookie, no accounts/roles — MVCP M.1/M.10's real account
system supersedes this in period 3. Covers the existing
/api/admin/library/* routes (previously unauthenticated, see
architecture.md 1.2's tech-debt note) plus the new upload endpoint.

admin.html gets a fetch() interceptor that redirects to the new
admin-login.html on any 401 from /api/admin/*, rather than patching
every call site.

POST /api/admin/upload: hand-rolled multipart parser (no deps installed
in this project), real magic-byte sniffing (rejects extension-spoofed
files), 8MB cap, and EXIF/GPS stripping via a sips bmp round-trip
(macOS-native, no dependency — sips's direct jpeg->jpeg re-encode alone
does NOT strip EXIF, only the bmp intermediate does). GET /uploads/*
serves the result with a path-traversal guard.

forwardToH3Assets() is the T7/T8 hook for pushing a saved upload to H3
Console's own POST /v1/assets (found in the S1-S4 spike) — not covered
by the automated suite since it needs a live network call, same
treatment as the spike scripts.

S1: person+garment images directly into ref2va with attribute_transfer
produce correct face/outfit/motion in one call — no separate compose
step needed. S2: reference_video_urls has a real, strong effect on
output motion (A/B tested). S4: H3 Console's own POST /v1/assets
endpoint (undocumented until now) solves asset reachability without
exposing our own uploads/ publicly.

Evidence and scripts in spike/s1-s2-s4-2026-09-02/. Propagated the
one-step design into the requirements doc, MVCP design doc, the
architecture doc, and H3-CONSOLE.md; dev-plan.md's spike section
marked done and its task table updated accordingly.

Generic createStore(filePath) shared by characters.json/garments.json/
dance-refs.json, same full-read/full-write pattern as library.json.
Not yet wired into server.js routes — that's T3-T5.

Duration: fixed 10s (within H3's 4-15s hard constraint), frames=240 @ 24fps.

Prompt: follows the internal H3 prompt-writing methodology already sitting
in the repo (docs/h3skill.md, from a parallel session) — the "all-purpose
reference mode" six-section structure. Character appearance details are
auto-captioned from the Stage1 composed image (new composed_images.
appearanceDescriptor field); dance motion phrasing is admin-authored once
per reference video (new dance_refs.motionDescriptor field), since the
format requires concrete descriptors and forbids "as shown in the picture."

Flagged as needing a pass through h3skill.md's own 22-item mechanical
acceptance checklist before production use, not claimed as fully compliant.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SdtXTsDRz5QWMjvt1HBXLz

H3_GEN_PARAMS.frames 120 -> 240 (at 24fps, 5s -> 10s), still within H3's
4-15s range (H2 in docs/h3skill.md). Updated the prompt-optimization system
prompt's duration reference and widened its word budget (40-70 -> 60-100)
to match — still one continuous shot, just more room to let the camera move
and the action unfold. Verified against the real GLM endpoint.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0144MiJ2cLjMd9jWaeEpVfUJ

Previously left as inline "勘误" patch notes; now the baseline doc
(docs/done) states the current 480x832 default outright, and the
dance-room docs describe live.html/dance.html as two independently
configured H3_GEN_PARAMS that happen to agree, rather than flagging an
erratum.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SdtXTsDRz5QWMjvt1HBXLz

Data model, front/back-end view inventory, button-level control design, and
per-user-type process flows for the not-yet-built feature. Also corrects the
requirements doc's stale "live.html stays landscape" claim now that
H3_GEN_PARAMS was changed to portrait there too (commit 1dc7084).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SdtXTsDRz5QWMjvt1HBXLz

The previous system prompt asked GLM for a free-form "shot-by-shot" expansion,
which reliably produced 4-5 fake cuts with timestamps running out to 8-10s —
impossible for a clip fixed at ~5s (H3_GEN_PARAMS.frames / FPS). Rewrote it
against docs/h3skill.md (MiniMax's own H3 prompt-writing methodology, added
alongside this commit): every chat prompt has no reference images, so it's
always T2VA mode, and at ~5s it's one continuous shot, never a montage.

The new prompt asks for H3's actual three-field T2VA structure
(integrated_multimodal_description / overall_soundscape / non_diegetic_music,
fixed order and labels), a single dominant camera movement from the official
vocabulary (type + amplitude + speed), concrete lighting (no hollow buzzwords
like "8K"/"cinematic" alone), and — previously missing entirely — explicit
sound design. Verified against the real GLM endpoint on several sample
prompts: single continuous shot throughout, correct field structure, no more
impossible multi-shot timestamps.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0144MiJ2cLjMd9jWaeEpVfUJ

Documents the new outfit-swap dance live room (dance.html) requirements
alongside the existing text-prompt live channel baseline, plus the
combined system architecture covering both.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SdtXTsDRz5QWMjvt1HBXLz

Adds a prompt-optimization step ahead of H3 generation: a viewer's short
chat line ("一只猫跳舞") gets expanded into a detailed shot-by-shot English
prompt (camera work, lighting, action, setting) via one GLM Coding Plan call
(glm-5.3-flash, /api/coding/ endpoint — the coding-plan-specific path, not
the pay-per-token one) before being handed to H3. The expanded prompt becomes
the clip's fullPrompt; the original short text stays as the display prompt.
Persisted on the queue entry so a restart mid-generation doesn't re-call GLM.
Best-effort throughout — no API key, a timeout, or any GLM error just falls
back to using the original text unchanged, never blocking generation.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0144MiJ2cLjMd9jWaeEpVfUJ

Swaps the H3_GEN_PARAMS default so future chat-triggered generations come out
mobile-friendly (9:16) instead of 832x480 landscape. Existing library clips
keep whatever aspect ratio they were generated at — the player already uses
object-fit: contain, so mixed aspect ratios render fine either way.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0144MiJ2cLjMd9jWaeEpVfUJ

Adds playCount to each library clip, incremented every time it airs
(markPlayed). The idle-filler pick in pickNextClip now randomizes only among
the least-played clips in the eligible pool — everyone at 0 plays first, then
everyone tied at 1, and so on — instead of picking uniformly at random across
the whole library, which let some clips rack up far more airtime than others
by chance alone.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0144MiJ2cLjMd9jWaeEpVfUJ

The video autoplays muted underneath a full-screen splash — logo, play icon,
slogan and a credit line, matching infiniteslop.ai's own splash. Clicking it
is the user gesture that lets the browser allow unmuted playback, so desktop
viewers land with sound instead of the silent-until-you-find-the-mute-button
experience from the previous change. Mobile stays muted after entry, same as
infiniteslop.ai, since unmuted autoplay-adjacent playback is unreliable there.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0144MiJ2cLjMd9jWaeEpVfUJ

Drops the muted attribute and tries an unmuted autoplay first; if the browser
blocks unmuted autoplay (no prior user gesture on the page), falls back to
muted playback and syncs the mute button label to match. Once a viewer
unmutes once, later clip switches keep playing with sound.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0144MiJ2cLjMd9jWaeEpVfUJ

Reworks live.css/live.html to match infiniteslop.ai's visual language: Tahoma
bold type, pure-black canvas, hot-pink accent, and hand-drawn white cards with
a thick black border + hard offset shadow for every button/bubble/prompt card.
Left queue now groups into 接下来播放/生成中/排队中 sections, the center gets a
white "now playing" card (time + requester + prompt) plus a pulsing LIVE pill,
chat bubbles are colored per-nickname via a pastel hash, and the top bar gained
a chat-close toggle alongside mute and the watching count.

Adds real upvoting to go with the new queue UI: POST /api/queue/vote/:id
toggles a per-viewer (cookie uid) vote on a queued prompt, and pickNextClip
now schedules the highest-voted matched/ready entry first instead of strict
FIFO — lets chat bump a favorite prompt to the front of the line.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0144MiJ2cLjMd9jWaeEpVfUJ

- "昵称 X ✎" button (next to the "聊天" title) replaced with a row directly
  above the chat input: hidden until a nickname is set, then shows
  "<nickname>  change name" — matches infiniteslop.ai's own placement/label
- Nickname prompt now triggers on focusing the chat input itself, not just
  on hitting send — so a first-time visitor sees it the moment they click in

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Previously the nickname modal only appeared implicitly on first chat send,
with no way to see or change it afterward. Adds an always-visible "昵称 X ✎"
button next to the chat panel title — same modal, pre-filled with the
current nickname, keeps the existing uid when editing.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

- Cookie-based viewer identity: first chat attempt prompts for a nickname
  (inline modal, not a blocking native prompt()), generates a client-side
  uid, both persisted in cookies for 1 year
- Right panel is now a live chat feed (username + message + status) with
  an input row below it, replacing the old bare prompt textarea
- New POST /api/chat: runs a moderation check (banned-word blocklist —
  ponytail: starter list, not a real content policy, swap in the real one)
  before a message is allowed into the existing prompt-matching/generation
  pipeline; rejected messages still show in the chat log, just flagged
- New GET /api/chat: recent chat history (capped at 300), polled every 3s
- Left queue simplified to prompt + submitter name only — dropped the
  resolution/duration/date metadata that was there before
- schedule/promptQueue now carry `requestedBy`/`nickname` through to
  /api/live/state so the left queue and "now playing" bar can show who
  asked for what
- /api/prompt kept working (now a thin wrapper over the same
  submitPromptToQueue() the chat endpoint uses) for direct API testing

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Inspected infiniteslop.ai's computed styles directly (fonts, colors, button
shapes) rather than guessing from memory:
- Pure black background, Tahoma body font (matches their retro-web feel)
- Shrikhand display font for the logo mark, pink drop-shadow for pop
  (their black stroke doesn't read on a black background, so swapped)
- Hot-pink accent (#ff90e8, their exact "Say" button color) for active
  tabs, the primary CTA button, and active-row/tab indicators
- White pill buttons with bold black text for secondary controls (tabs,
  mute toggle) — matches their unmute/chat button treatment

Layout structure (queue-left/player-center/input-right, mobile sub-tabs,
no native video controls) is unchanged — this is a re-skin, not a layout
clone, since that structure was built to this project's own requirements.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

- library.json: all 303 real entries now have casual zh/en names (prompt/promptEn)
- ingest.js: fixed a data-loss bug where the H3 console fetch was hardcoded
  to limit=200, silently dropping older jobs (and their edited names) as new
  content pushed them out of the window; sync is now a non-destructive
  merge with a much higher fetch limit
- ingest.js: replaced the prefix-only test-job regex with a length-gated
  word-boundary check after 5 infra smoke-test clips slipped past it
  ("test verify", "cancel test - ...", "final flip test") — no false
  positives against the 303 real entries
- deleted the 5 stray test clips (also removed on the H3 console side)

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Static video-gen testing tool (MiniMax native API, then fal.ai) that grew
into an AI drama live channel backed by an internal H3 Console GPU cluster:
shared virtual-playhead live playback, prompt submit/match/generate queue,
admin CRUD for the video library, and bilingual (zh/en) semantic clip names.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>