feat(short_drama): turn a novel into vertical drama episodes on TRTMC - #1932

Add a separate 226-frame experimental bundle profile with synchronized
video/audio latent guides, two complete 49-forward windows and cumulative
decoding. Preserve the original single-window profile and residency options.

Reject child profile limits before first-window inference, retain each stereo
channel prefix, and serialize the complete request across the AdaLN cache
transition. Add an independent native 98-forward oracle with synthetic weights.

Validation: CPU family regression and window fixture, six host executables,
native probe syntax, family ownership and formatting checks passed. The new
GPU oracle and full trained continuation remain unqualified on this revision.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Build the two changed continuation plans from a pinned original image bundle.
Verify its complete digest, original profile, checkpoint, tokenizer and BF16
provenance before loading weights; stream and rehash the six unchanged plans.
Record reused plan digests and preserve the original bundle.

Share exact runtime metadata with the existing full builder. Runtime and graph
math are unchanged. Validate both residency presets, source mismatches,
streaming mutation rejection and the existing full build wiring with 59 CPU
cases; ownership and formatting checks pass. Full trained execution remains
unqualified until the target build and video validation complete.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Allow an explicit second-window prompt on the bounded 226-frame Ref2VA profile. Encode both prompts with the original references, validate child conditioning before first-window inference, and anchor both windows after the longest text prefix as in the fixed upstream implementation.

Keep empty and identical prompt behavior, original sampling, media clocks, and residency policies. Add shorter/longer text trajectory cases without changing numerical criteria. Host validation passes; GPU trajectories and trained dialogue quality remain to be verified.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Apply repository-pinned clang-format 22.1.8 to the source-quality diagnostics. Only whitespace and comment wrapping change; preprocessed C++ token streams and all numerical criteria are unchanged.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Recognize checkpoint snapshot identities within the owning family instead of importing the internal provenance collector. Preserve source-bundle and checkpoint matching criteria and compare the local identity against the existing behavior. Keep all direct-staging assertions in the existing bundle-writer test module to preserve the closed shared-tree contract.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Select one of eight precomposed BF16 video/audio output-head banks in a
strongly typed TensorRT graph, retaining explicit operand and output
rounding around FP32 accumulation.

Add a synthetic GPU oracle for all eight selectors, three dynamic row
profiles and the trained hidden width. CPU request validation and scope
preflight pass; complete student graph, VSA and trained-model qualification
remain required.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Bind both dynamic modalities and the PDD selector before inferring output
dimensions in the GPU component oracle. Retain shape diagnostics for each
case so a binding failure is distinguishable from numerical disagreement.

Keep the original graph, 72 cases and BF16 comparison limits unchanged.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Add segment-pure 128-token transport, independent reference-video key
budgets and the learned compression boundary. Implement the operation
through TVM-FFI with generic and optional Blackwell backends, and compose
it with the existing owner-local H3 backbone and eight PDD output heads.

CPU regressions pass 157 cases and seven pinned upstream geometry cases.
CUDA compilation, trained numerical validation and native student runtime
integration remain required; the earlier output-head mismatch is unwaived.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Select the released eight-forward student recipe explicitly, preserving original H3 scheduling and fixed conditioning. Bind native 128-token VSA geometry and the trained head selector for each window.

Validate the closed checkpoint contract, native schedules and transport against pinned source math. Add a real TensorRT VSA component oracle; target GPU compilation and trained-model qualification remain required.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Admit the pinned student contract and exhaustive parameter index, rebuild both student partitions, and reuse only explicitly compatible common plans from a verified source bundle.

Write native serialized host buffers directly into the bundle to avoid the second plan-sized Python bytes allocation. Preserve the original bytes API and original continuation defaults. Host checks cover partition ownership, metadata, source rejection, short writes and unchanged common sections; trained model and target memory qualification remain pending.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Distinguish the synthetic text segment from the complete Qwen image-token prefix. Preserve the existing workload and numerical criteria; the component case does not qualify a full six-reference request.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

The local block-output ITensor overwrote the caller's serialization stream,
so a completed TensorRT build tried to invoke ITensor.write(). Keep the
block tensor separate from the optional destination to restore both the
default bytes API and native host-buffer streaming.

Add four real TensorRT deserialization round trips for bytes/streamed output
and features/native-timesteps inputs. CPU validation: 56 passed and six GPU
cases deselected. Target GPU execution remains pending; no model shapes,
precision boundaries or numerical thresholds change.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

The output-head marking loop also reused the serialization destination's
name. Keep those ITensors separate so the complete student builder passes
the actual writer through to native serialization. This was caught during
pre-launch review; no DiT build was submitted with the earlier fix.

Validation: 78 CPU checks passed, six GPU cases deselected. All three
streaming builder entry points retain their output parameter. Model math,
shapes, precision and numerical thresholds remain unchanged.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Allow an explicit denoiser_workspace_bytes budget while retaining the
8-GiB default and the separate AdaLN budget. Record both limits in the build
receipt. The full VSA transport graph can require a tactic larger than the
previous limit; selecting more scratch does not change its input profiles,
weights, precision, sampling or correctness gates.

Validation: 96 CPU checks passed and six GPU cases deselected. Coverage
includes default and explicit per-plan budgets, invalid values rejected
before weight loading, and unchanged student/runtime contracts. Full GPU
bundle construction remains pending.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

The native Ref2VA pipeline checked every recipe against the original H3
49/98-forward counts after denoising. Valid OmniRef PDD outputs therefore
failed before video and audio decoding.

Validate the existing recipe contract: original H3 requires 49/98 forwards,
and OmniRef PDD requires 8/16. Reject unknown recipes, unsupported window
profiles, and all other counts. Preserve the denoising schedule and model
precision, shapes, weights, and numerical gates.

Add CPU regressions for both recipes and window profiles, including wrong
counts and invalid configuration. The focused test reproduces the stale
guard failure before this fix; the full pipeline host regression passes
after it with both residency modes.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Replay saved operands through the unchanged native FFI function and retain
routing, compression, and fine-attention observations. Compare the original
gated result against its original reference and numeric limits.

A separate zero-gate call exposes fine attention for diagnosis only. It
cannot qualify the original failing case or the trained model. Preserve all
original input and reference files.

Validation: Ruff and Python compilation passed. Real GPU localization
remains pending. No production graph or kernel changes are introduced.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

The optimized 128-token attention discarded the softmax probability residual
before PV. Retain a second BF16 limb in the unused half of the QK tensor-memory
tile and accumulate both products into the existing FP32 output. Wait for
both limbs before the PV handoff; keep the 64-token path unchanged.

A family-owned CPU experiment on saved operands isolates probability storage.
Its sampled result supports the change but does not qualify CUDA numerical
agreement, performance, trained-model quality or the application. Preserve
the existing reference and tolerances for GPU validation.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

The released cubic exponential approximation can move BF16 attention across
a rounding boundary, which a cancelling gated sum exposes. Use the existing
hardware exponential path for the optimized 128-token operation. Preserve
the 64-token path, probability residual, workspace and BF16 I/O boundaries.

Add a family-owned CPU counterfactual on saved boundaries. It reproduces
both fused outputs bitwise and attributes the remaining discrepancies to
fine-attention rounding. The cubic simulation explains the observed error
pattern; it is diagnostic evidence, not GPU or trained-model qualification.
Keep the reference, operands, seeds and numeric limits unchanged.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Accumulate a third BF16 probability limb without increasing the workspace,
TMEM or shared-memory allocation. Reuse the epilogue buffer before final PV,
wait for the previous TMA read, and hand it back after the MMA completion.
Keep the hardware exponential and the original BF16 I/O boundaries.

Extend the family CPU diagnostic to project the remaining cancelling queries.
Allow the GPU oracle to collect all numeric results before exiting nonzero;
its default remains fail-fast and every original comparison still must pass.
Reference, comparison, case/seed data and tolerances are unchanged. CUDA
numerics, synchronization and performance still require target validation.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

The unchanged 56-head component profile includes 708 physical tiles and declares 285624192 bytes of BYOK scratch, exceeding the historical 256 MiB builder pool. Size the oracle builder pool from the declared operation bound while retaining its previous minimum.

Keep every operand, reference, case, seed, comparison and tolerance unchanged. This fixes a build-resource error and does not resolve existing numerical failures. The 22 family VSA CPU tests pass.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Add an offline diagnostic for immutable saved failing coordinates. Compare FP64 attention with approximate FP32 scaling and softmax operation orders, retaining direct nearest-even BF16 rounding across FP32 midpoints.

Keep native fine values explicitly conditional until directly recorded. The diagnostic does not replace the GPU oracle or change inputs, masks, precision boundaries or tolerances. All 29 focused VSA and rounding CPU tests pass.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Record call geometry and host entry/enqueue boundaries only when the explicit trace environment switch is enabled. Preserve all device computations, original asynchronous behavior and correctness gates. Document that enqueues and optional launch-blocking diagnostic times are not benchmark evidence.

Validation: all 29 existing VSA layout, transport and rounding-observation CPU checks pass. CUDA compilation, device-instruction comparison and bounded native failure localization remain separate checks.
Signed-off-by: chaofengw <chaofengw@nvidia.com>

Keep the third probability limb in dedicated shared tiles rather than
borrowing the output epilogue buffers and waiting on their TMA lifetime.
Reassign two KV ring stages to the tiles within the existing shared-memory
reservation; retain all three PV products and the external scratch contract.

Include absolute host trace timestamps to distinguish a stalled call from
setup time. CPU layout and rounding contracts pass; GPU output parity and
full-request progress remain to be verified.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

A one-slot KV ring cannot prefill both K tiles before Q: the producer waits
for MMA to release the first K slot while MMA is waiting for Q. Prefill Q
first when the ring cannot hold the initial K loads. Keep the four-slot
64-token order, probability limbs and shared-memory reservation unchanged.

CPU layout and rounding contracts pass. The saved-output preservation and
original full-request stall checks remain required.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Replace generic online attention with separately scaled FP32 Q/K, explicit masked softmax and FP32 PV inside the owning TVM-FFI binding. Complete compression first and reuse its dead scratch while retaining routing indices, counts and compressed output.

Keep optimized Blackwell arithmetic available and reject old generic bundles with insufficient scratch. The existing 56-head large workspace sizes stay unchanged. Host layout checks and 32 focused CPU tests pass; integrated GPU and trained-model qualification remain required.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Singleton SGEMM changes the PV accumulation order on the complete saved interleaved case. Use two independent batched products while retaining one head of residency: dead Q holds one output, the fine tile the other, and the completed key mask holds device pointers.

Keep every arena size unchanged and assert output/input/pointer-table separation in the host workspace checks. All 32 focused CPU tests pass; complete integrated GPU gates remain required.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Use the existing FP32 warp and block reductions with explicit division for compression probabilities. Valid INT32 tile sizes directly mask empty keys, retaining the external workspace and BF16 boundaries.

A complete saved-output diagnostic isolated the sixteen remaining large-case failures to compression rounding after the old reduction. Keep fine attention and the runtime ABI unchanged. Updated complete-case GPU qualification remains pending.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Keep the compressed output at the front and routing indices at the aligned tail of caller-provided scratch. Fit up to 32 independent original query tiles between them while retaining one head of FP32 K/V, separate regional masks, and disjoint QK/PV destinations.

Reuse dead PV output and completed masks for device pointer tables. Preserve the duplicate PV reduction when only one query is active, and leave the external workspace contract and BF16 boundaries unchanged. CPU lifetime and capacity checks pass; complete GPU numeric and performance evidence remain required.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

apps/short_drama is a small web app that runs every model step through
trtmc-server on the user's own GPU: a language model plans the cast,
places and shot lists from the novel, Qwen-Image draws reference stills
to approve, MiniMax H3 (FastH3 OmniRef) films each shot with sound from
the approved stills, and Whisper times the subtitles before FFmpeg
assembles 1080x1920 episodes.

`python -m trtmc_short_drama up` picks a tier by GPU memory, builds
missing bundles with the TRTMC builder (H3 through the family recipe)
and loads only the bundles each step needs. The flow (steps, models,
parameters) is versioned JSON edited in the app, and a step can be
added as a plugin file; dialogue_polish is the example. Planning can
use an optional OpenAI-compatible API whose key stays in a mode-600
file.

The bundled sample is the public-domain text of 《阿Q正传》 by Lu Xun.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…ds it

trtmc-server reads a bundle once into GPU memory. On GB300 the GPU's
memory is a NUMA node of the host, and the file cache of a 100 GB
bundle can land there as memory no process owns: a later stage's model
then fails to load. The engine now advises the kernel to drop the
cached pages of the bundles it starts with, before the start and again
once the server is ready.

On a GB300, the cache an earlier run of the Qwen3.6-27B bundle had left
on the GPU's node fell from 52 GB to 0.4 GB once the model had loaded;
all GPU memory in use then belonged to the server.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…idea

The app now leads a project through its four stages and keeps a person
in charge of every step that costs GPU time.

- Each stage (plan, references, shots, episodes) shows what is done and
  what to do next.
- Plan checks run on every change: people the chapters name often but
  nobody plays, someone acting in a shot without being in it, a speaker
  outside the shot, a line too long for its shot, unknown ids, shots
  longer than one clip, episodes over their limit. Each line is marked
  as quoted from the novel or adapted.
- Three actions run with the plan step's language model: review (notes
  per shot on missing events, continuity and over-full shots), revise
  (rewrite one shot, character, location or prop from a note and keep
  the rest of the plan) and story (write a short story from a one-line
  idea, so a project can start without a novel). A plugin step becomes
  such an action with model_from.
- A re-plan, revision, save or restore first keeps the plan it replaces;
  earlier versions can be restored.
- Only characters, locations and props that appear in a shot get a
  still, and each still lists its shots.
- Every take keeps the prompt and stills it was rendered from. A redrawn
  still or a changed prompt marks the take outdated, and only outdated
  shots need rendering again.
- Shots start with one sample. A batch first says how much GPU time it
  takes, from the times measured per still and take on this machine. A
  strip shows every shot of the episode.
- A cast member who is heard but not seen, as on a phone, keeps the line:
  the video prompt gives it as an off-screen voice with that character's
  voice, instead of moving it to someone in the shot.

On a GB300 with Qwen3.6-27B, the review of the bundled four-episode plan
took 178 s and found, among others, a line given to the wrong speaker;
revising one shot took 12 s.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

`up` no longer builds the 27B language model unless asked with
`--llm local`. Planning, stories, reviews and revisions use the chat
API chosen in Settings, which presets two free tiers (the NVIDIA API
catalog and OpenRouter) and still takes any OpenAI-compatible URL.

A chat API that answers 429 or 5xx is asked again three times,
honoring Retry-After, and a failure carries the API's own reason with
what to check. Without any language model the plan stage points to
Settings instead of failing on the first step.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

A model server that exited during its load left the step with only "see <log>", so the page
could not say why nothing happened. The error now carries the last line of the log, for
example a missing serving module, and the engine reports what it holds, what it is loading and
since when, and the end of its log, for the page to show.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…t management

- NVIDIA look: NVIDIA Sans, NVIDIA green, light and dark themes, and the header of NVIDIA's
  documentation sites.
- A Projects page with each project's progress, rename and delete. New project asks for the
  story source, the visual style, the length and a title, and estimates the shots and GPU time.
- A project shows five steps and one card with the next thing to do. While a step runs, the
  card shows its phase (the model load included), the time so far and the usual time; a failed
  step shows its reason with Try again and the server log.
- An activity panel: the loaded models, running and queued steps of every project, recent
  errors and the end of the trtmc-server log.
- A style chosen with the project becomes the plan's style and shapes the looks; shot buttons
  wait for the pictures they use; the page token is kept across restarts.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…e last one ends

TRTMC's MiniMax-H3 starts every clip with a loud burst of about 0.15 s that the
official samples do not have. Clips now start 0.25 s in, fade their audio in and
out, and the episode gets one loudness.

Within an episode, a shot in the same place as the shot before it gets that
shot's last frame as an extra reference, so positions, clothes and light carry
over. A new take of a shot marks the next one outdated. The shot list asks for
one main action per shot, one ambience per scene and no sound effect in the
first second.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…ysis and hardware tiers

- The page switches between Chinese and English (i18n.js; every page text has a
  translation, checked by a test). A project has a language: the story, the plan,
  the lines and the reviews are written in it, and a line is spoken in its own
  language. Every textarea grows with its text and opens in a large editor.
- Video analysis (the analyze action, run with the speech model): measures each
  rendered shot's opening burst, clipping, level, spoken line, black or frozen
  frames and the picture and level jump at every cut. With a vision model set in
  Settings it also sends the frames, up to two reference pictures and the
  previous shot's last frame, and keeps the model's findings on content,
  on-screen text, continuity, consistency and reference pictures. The page lists
  the findings with an editable suggestion and an action; applying them queues
  plan revisions, new takes or redrawn pictures.
- Three hardware tiers by GPU memory: A (140 GB, H3), B (80 GB: RTX PRO 6000,
  H100, DGX Spark) and C (24 GB), with what each still lacks in TRTMC. GPUs that
  share system memory (DGX Spark, Jetson Thor) read it from /proc/meminfo.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…hree tiers

Also fixes two Chinese strings that repeated 约 and says plainly that an empty
vision model means measuring only.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…s missing

On an RTX PRO 6000 or a 24 GB card the shots step would only fail later; up now
says which TRTMC pieces the tier still needs. --allow-planned starts it anyway
for stories, plans and pictures.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…d with the last frame

The FastH3 OmniRef student often holds one of the reference pictures still for
0.5 to 1 s and then cuts into the shot (the original 50-step H3 path does not);
that held picture made every cut jump. A clip now starts after such a cut when
it comes within 1.2 s: nearly still frames, then a change of 40 grey levels or
more. Fast motion keeps the 0.25 s lead-in. Each take records where it starts.

A shot that continues the shot before it now gets that shot's last frame as
Picture 1, and the prompt says the shot opens exactly on it, so a held opening
picture becomes a continuation instead of a jump.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

… join neighbours only

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…ge cannot hear

The Whisper bundle is built for Chinese; an English project's analysis now
measures the speech but reports no line findings, and says why.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Signed-off-by: chaofengw <chaofengw@nvidia.com>

… unplanned speech

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…-off chose

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…ames none

A flow saved while the app planned through a chat API has no model on its plan
step; starting the app with --llm local then failed every story and plan with
"model None is not served".

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…kes again

The FastH3 OmniRef student can hold the previous shot's last frame for 1.4 s, longer than the
1.2 s the opening check looked at, and can hold a portrait and then a prop in turn. A clip now
starts after every still picture that ends in a cut within 2.5 s; a picture is still when no
frame changes by 3 grey levels or more (held pictures measured at most 2.05, inserts and moving
shots at least 4.47 on the clips rendered so far).

Assembling an episode cuts takes made under an older rule again from their raw clips and moves
their line timing with the new start. Subtitles follow the clips' real lengths, which can now be
shorter than planned. The level-jump finding no longer claims that assembly removes the jump.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Assembling ran after the subtitles step, so lines were timed and heard on the old cut. Both steps
now cut takes of an older rule again first. A take cut again keeps its moved line timing but drops
what was heard on the old cut, and the analysis listens to the new clip.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

… no false clipping

trtmc-server takes one user message with one picture, so a vision model served by TRTMC rejected
the analysis. The shot's frames, its reference pictures and the previous shot's last frame now go
as one picture (frames on top, pictures below, described in the prompt), and a request with
pictures is one user message that starts with the system text. Thinking models get 12,000 tokens
as plans do, and the prompt asks for the project language in its own line: Qwen3.5-9B wrote
English findings for a Chinese project before.

Every clip was reported as clipping: FFmpeg's mono downmix adds the two channels at 0.71 each,
so H3 audio that peaks at 0.95 measured 1.34. The analysis and the speech recognizer's input now
take the mean of the channels.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…--vision local)

The analysis could only show frames to a vision model of the chat API in Settings, and Settings
hold one API for both. Tier A now names Qwen3.5-9B built for pictures as its vision model; with
--vision local the app builds and serves it, and the analysis loads it next to speech recognition
so the server is not restarted shot by shot. A vision model named in Settings comes first. The
local server gets no thinking switch with pictures, which trtmc-server rejects.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…ut tokens

TRTMC's Qwen3.5 vision path takes 2,048 tokens of picture and text together, and the analysis
of shots with long plans was rejected as invalid. The picture is now 1024 wide with 360-high
pictures (about 600 tokens), and each plan field, look and the style are cut to bounded lengths.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Given the recognized speech, Qwen3.5-9B reported fully heard lines as garbled and filed them under
on-screen text. Sound and speech are measured already, so the vision prompt drops what was heard,
the measurements and the planned sound, and says that only the picture is judged.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

Measured on GB300 and RTX PRO 6000: with thinking on (trtmc-server cannot turn it off for pictures)
Qwen3.5-9B takes 2 to 4 minutes a shot and ran out of 12,000 tokens on two of three shots of one
episode; with thinking off (tried on a patched server copy) it answers in about 35 s but reads the
reference pictures as frames of the clip.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

H3 had people say lines of its own in shots without one (the analysis heard "我今天要做的" in a
shot of the owner's episode). Such shots now say that no one speaks and lips stay closed, as
voiceovers already do. Their existing takes show as outdated, since the prompt changed.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…nglish

With the plan in the project's language, Chinese prose reached H3 outside the dialogue tags and H3
read it aloud: after the analysis loop revised a shot of the owner's episode into Chinese, the new
take spoke the shot summary, and the cartoon project's lines came with extra words. The shots
step now asks the language model for the English of every description written in Chinese,
Japanese or Korean, before the video model loads, and keeps it per project; prompts, the outdated
check and the prompt preview use those translations. Names and lines stay as written.

Signed-off-by: chaofengw <chaofengw@nvidia.com>

…the prompt wording

A take was marked outdated whenever its saved prompt differed from the prompt the shot would get
now. The prompt text mixes the plan with how the app writes prompts, so an app update (the
no-one-speaks sentence, descriptions put into English) marked takes outdated that no edit touched.

Each take and still now keeps the parts of the plan its prompt was written from. A take is outdated
when that plan and the plan now give different prompts, both written the way prompts are written
now. Takes rendered before this keep the old check, accepting the descriptions as written and a
line-less shot without the no-one-speaks sentence.

Signed-off-by: chaofengw <chaofengw@nvidia.com>