# One short pour: check the stream, rim and rising liquid level

Create a single small pour from a jug into a glass with visible liquid-level continuity. The original illustration and checks describe an untested visual method, not a fluid simulation or measured volume transfer.

## Package and evidence

The first-frame SVG is original artwork authored for this task. The PNG is the raster upload derivative prepared during integration. Neither image is a model output. The illustration has no teaching text or arrows baked into the input.

Files: `first-frame.svg`, `first-frame.png`, `brief.en.json`, `brief.zh.json`, `checks.csv`, `README.en.md`, `README.zh.md`. The CSV result fields are intentionally blank; fill them from a clip you actually inspect.

## Prepare

1. Choose an input showing the jug spout, glass opening and initial liquid level without overlap. The supplied glass begins partly filled, leaving headroom.
2. Use a plain background behind the stream and glass rim. Strong reflections can hide whether liquid enters the opening or crosses the wall.
3. Keep the receiving glass on a clear tabletop and the camera fixed. Ask the jug to tilt alone, without adding a hand pickup or camera move.
4. Define a modest rise below the rim and an upright jug ending. Do not use this practice to prove exact milliliters, pouring speed or conservation measurements.

## Copyable prompt

```text
Start with the uploaded jug and partly filled glass. The glass stays fixed on the tabletop with its rim and base unchanged. The jug tilts slightly once and releases a narrow continuous stream from its spout into the open center of the glass. The stream crosses the opening inside the rim, never through the side wall. As liquid enters, the receiving level rises gradually and remains below the rim, with only a small surface ripple. The jug then returns toward upright, the stream stops, and the glass holds its new level. Keep one jug, one glass, fixed camera and lighting; no overflow, no splash burst, no second pour.
```

## Run and inspect

1. **Confirm the entry route.** Trace a clear route from the spout to the glass opening in the source. If the rim is hidden behind the jug, adjust the input before asking for pouring.
2. **Separate the three states.** Brief the action as tilt, short pour, then stop. Leave the glass stationary in all three states so a lifting glass cannot disguise an incorrect stream intersection.
3. **Sample the receiving level.** Compare before flow, mid-flow and after stop against the same part of the glass. The level should rise during entry and remain at the new height; screen reflections alone do not count as more liquid.
4. **Review the stopping transition.** Inspect the moment the jug returns toward upright. Check that the stream ends rather than continuing from a dry spout, and that the glass wall and rim stay intact.

## Acceptance

- There is only one jug and one receiving glass.
- The glass base stays in contact with the same tabletop location.
- The stream originates at the visible jug spout.
- The stream passes into the glass opening inside the rim.
- The receiving level rises progressively during entry rather than falling or teleporting.
- Liquid remains contained below the rim without passing through glass walls.
- After the jug returns toward upright, flow stops and the higher level settles.

## Change one variable per retry

- **The stream passes through the glass wall:** Change only the entry target to the open center inside the visible rim; keep the jug motion restrained.
- **The liquid level jumps or falls while pouring:** Reduce only the requested poured amount and compare the same three sampling moments.
- **The glass rises to meet the stream:** Change only the fixed-role sentence to name the base-to-table contact; keep the spout-to-opening route.
- **Flow continues after the jug straightens:** Change only the ending to an explicit stopped stream before the final hold; remove continuous-pour language.

## Ending

End with the jug near upright, no active stream and a settled level below the rim. Preserve samples from each flow state and record visual continuity rather than an invented volume measurement.

## Intent evidence and boundaries

Historical parent queries below are literal entries in `docs/research/pseo-keyword-gap-2026-10-05/evidence/bing-live-keywords.csv`. This package claims no exact long-tail demand measurement or search volume. The owner plan at `docs/research/pseo-complete-plan-2026-10-09/keyword-owner-work.csv` assigns broad prompt-template intent to `/templates` and distinguishes complete task presets from prompt collections.

- `minimax h3 video generation prompt guide liquid fluid animation eyelids`
- `minimax h3 image to video`
- `minimax h3提示词模板`

A single short jug-to-glass pour with stream-to-rim intersection, nondecreasing receiving liquid level, stationary glass and a stopped stream in the ending.

- `/blog/product-shape-consistency-video`: This supplied jug/glass scene tests fluid entry and level continuity around a fixed rim; general product shape checks do not cover the pour-state sequence.
- `/blog/product-video-shot-list`: This is one bounded pour with fluid acceptance checks, not a beverage shot list or multi-shot commercial.

Keep the original source, your submitted prompt, the provider settings actually available to you and your observed checks together. This package does not prescribe hidden provider tokens, exact duration, cost or supported endpoint parameters.