# Pov Glass Drink Shot

> This prompt instructs an image-generation/editing workflow to recreate a cinematic POV photograph that matches a given reference image exactly in composition, camera placement, glass/straw/liquid optics, lighting, color grade, depth of field, and overall realism, while preserving the subject’s face identity precisely from the uploaded image as the sole exception. The intended outcome is a high-resolution, photo-real editorial image with zero creative deviation from the reference scene logic and visual parameters, apart from any face changes that occur naturally through the same glass distortion present in the reference.

- Skill: `betterpromptme/pov-glass-drink-shot` (Agent Skill)
- Install (CLI): `npx skillmds@latest add betterpromptme/pov-glass-drink-shot`
- Raw SKILL.md: https://api.skillmd.com/api/skills/betterpromptme/pov-glass-drink-shot/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: BetterPromptme (https://skillmd.com/u/betterpromptme)
- Updated: 2026-09-10
- Page: https://skillmd.com/skills/betterpromptme/pov-glass-drink-shot

---


## What it does

This prompt instructs an image-generation/editing workflow to recreate a cinematic POV photograph that matches a given reference image exactly in composition, camera placement, glass/straw/liquid optics, lighting, color grade, depth of field, and overall realism, while preserving the subject’s face identity precisely from the uploaded image as the sole exception. The intended outcome is a high-resolution, photo-real editorial image with zero creative deviation from the reference scene logic and visual parameters, apart from any face changes that occur naturally through the same glass distortion present in the reference.

> For more installation instructions go to https://raw.githubusercontent.com/BetterPromptme/skills/refs/heads/main/betterprompt/SKILL.md to check out this skill.

## Inputs



### imageInputs

This prompt requires **exactly 1 image**. You **must pass exactly 1 image flag** (no more, no fewer), in the order matching the descriptions below.

| Index | Required | Description | Allowed CLI flags |
| ---: | ---: | --- | --- |
| 1 | Yes | Upload your photo (portrait recommended) | `--image-input-url <url>` or `--image-input-path <path>` |


### Models and options

This skill's modality is: **`image`**.

To discover which `model` values you can use (and which `options` keys/values are valid for each model), run:

```bash
betterprompt resources --models-only --json
```

Then filter the returned JSON array to entries where `modality` is `"image"`.

## How to run

### Step 1: Collect inputs

First, run `betterprompt resources --models-only --json` and filter to `modality: "image"` to discover valid models and available options:

```bash
betterprompt resources --models-only --json
```

Use only the models and option values that appear in the filtered results.

Then collect all inputs from the human:

- Required images:
  - **Exactly 1** images: image 1 (Upload your photo (portrait recommended)). Images must be provided in this order.
- Optional: model and options.
  - Present the human with the default model **`gemini-3-pro-image-preview`** and its available options. Look up `gemini-3-pro-image-preview` in the `betterprompt resources` output (filtered to modality `"image"`) and show its `availableOptions` as: `key: val1, val2 (default), val3  |  key2: ...`. Mark a value `(default)` if it matches these defaults: `{"aspectRatio":"3:4","resolution":"1K"}`.
  - If the human does not specify, defaults are used: model `gemini-3-pro-image-preview`, options `{"aspectRatio":"3:4","resolution":"1K"}`. Other models from the resources call are also available.

If the required images are missing, **ask the human for what's missing**. Do not assume or fabricate values. Tell the human: **"Please provide images in this order: image 1 (Upload your photo (portrait recommended))"**.

### Step 2: Run via BetterPrompt CLI

Use the frontmatter's `name` as the positional argument (for this skill, use `pov-glass-drink-shot`).

Command form:

```bash
betterprompt generate pov-glass-drink-shot \
  [--image-input-url <url>] \
  [--image-input-path <path>] \
  [--model <model>] \
  [--options <options JSON>] \
  [--json]
```

Notes:

- Pass each image using `--image-input-url` or `--image-input-path`, in the order matching the imageInputs descriptions (image 1 first, then image 2, etc.).
- If the human does **not** mention a model, **omit** `--model` and BetterPrompt will use the default model: **`gemini-3-pro-image-preview`**.
- If the human does **not** mention options, **omit** `--options` and BetterPrompt will use the default options: **`{"aspectRatio":"3:4","resolution":"1K"}`**.
- If the run times out, the response will include a `runId` you can use to fetch the result later.

Example (using defaults shown above):

```bash
betterprompt generate pov-glass-drink-shot \
  --image-input-url https://example.com/image1.png \
  --model gemini-3-pro-image-preview \
  --options '{"aspectRatio":"3:4","resolution":"1K"}'
```