# Play Images

The images are the pictures your video shows while the narration plays. On Play you pick who draws them, which Library image prompts they come from and how many each makes, and optionally an establishing image that keeps characters and style consistent. You get a set of images, drawn in the video's frame shape, played in order.

**Where to find it:** Play → **Images** row → **Change**.

## Choose where the images come from

The Images row starts with a **Source** switch:

| Source | What happens | Cost |
| --- | --- | --- |
| **Generate** (default) | Draws images from the ticked image prompts, one image-model call each. | Per image, by the provider |
| **Provide** | Uses your own PNG, JPEG or WebP files (**Slideshow images**) in the order you pick them. | None |
| **Off** | No images. This turns the video off too: the **Video and style** row says "Video is Off because Images is Off." | None |

### Use your own images

1. Set **Source** to **Provide**.
2. Pick the files under **Slideshow images**. Each is numbered in the order it plays.
3. Remove a file with its remove button, or reattach one whose upload was lost.

A video can have at most 60 of your images, and the same file cannot be picked twice. Wait for every upload to finish before pressing Start.

## Pick the image provider and model

| Option | What it does | Default |
| --- | --- | --- |
| **Provider** | Who draws the images, the thumbnail and the shorts' pictures: a keyed service billed per image (fal.ai, Replicate, OpenAI, Google), or a command-line tool you are signed in to (Codex CLI). A greyed provider says why it cannot be used; fix it under Settings. | Your Settings default |
| **Model** | The provider's image model. Models differ in quality, speed and price per image; the estimate in the right rail shows what this run costs. The refresh button asks for the current list; **Custom ID** lets you type one. | Your Settings default |
| **Effort** | How long an image agent, such as the Codex CLI, reasons before it draws. Higher effort can follow the prompt more closely but takes longer and uses more tokens. Shown only for models that offer levels. | **Model default** |

Changing the provider clears the model.

### Drawing with the Codex CLI

With **Codex CLI** as the image provider, the **Model** list holds **Codex default** and the Codex CLI's own models. **Codex default** leaves the model and effort to the CLI's own defaults. Any other model runs with the **Effort** you pick. An image agent that reviews and redraws its picture can work for several minutes at a high effort. Images drawn this way cost nothing extra on your plan; the estimate shows the approximate API price beside **$0 on your plan**. See [AI CLIs](AI-CLIs).

## Pick the image prompts

1. Under **Image prompts**, tick the Library prompts that should draw the video's images.
2. Give each ticked prompt a **Number**: how many images it makes, 1 to 20.
3. Check the total under the list, for example **8 of 60 images**.

| Rule | Value |
| --- | --- |
| Images per prompt | 1 to 20 |
| Images per run, hand-set | at most 60 |
| Images per run, with **More images for long videos** | at most 240 |
| Order | The images play in the order you ticked the prompts |

Each image is one image-model call. A prompt that has been deleted since this draft picked it shows **Missing template · untick to remove**. With no image prompts at all, the row says "No image prompts saved. Write one on Prompts." See [Prompts](Prompts).

### Photorealistic prompts

An image prompt in the Library has a **Draws photorealistic pictures** tick. Turn it on when the prompt's style looks like real photos or film. Prepare upload then answers Yes to YouTube's AI use question for videos and shorts drawn with it. Leave it off for painterly or illustrated styles. It is set on the prompt, not on Play: open Library → Prompts → the prompt. See [Prompts](Prompts) and [Publishing to YouTube](Publishing-to-YouTube).

## Establishing image

The establishing image is made first and never shown in the video. Every other image, the shorts' images and (if ticked) the thumbnail are drawn with it as a reference, so characters, style and palette stay the same across the video.

| Option | What it does | Default |
| --- | --- | --- |
| **Establishing image** | **Off**, **From a prompt** (costs one more image) or **Upload** (your own PNG or JPEG, **Establishing image file**). | Off |
| **Establishing prompt** | The Library image prompt that draws it, keywords filled like the others. Describe the characters, setting and style every other image should share. | None |
| **Draw the thumbnail from it too** | Draws the thumbnail with the establishing image as a reference too, so it matches the video. Untick it for a thumbnail drawn on its own. | On |

Changing the establishing image later marks the images drawn from it as outdated. When you upload the establishing image, the style preview in the right rail is drawn on it, so you judge the Look and captions on your own picture.

### Set up an establishing image

1. Under **Establishing image**, pick **From a prompt** or **Upload**.
2. For **From a prompt**, pick the **Establishing prompt**. For **Upload**, pick the file.
3. Leave **Draw the thumbnail from it too** ticked if the thumbnail should match.

## The channel's cast

The channel you pick in the **Channel** row brings its cast: characters, creatures, places and objects, each with reference pictures. When an image's brief mentions a cast member by name or alias (as a whole word, ignoring case), that image is drawn with the member's pictures as references, after the establishing image when it is on. The establishing image and the thumbnail also take the members the title mentions. At most four members go with one image.

Models that take several input pictures (Codex, OpenAI, Google, and fal.ai's FLUX.2 and Nano Banana 2 edit models) use the pictures directly. A model that takes none gets the members described in words instead. The cast is used even when **Use the channel's brand kit** is off. See [Cast Library](Cast-Library) and [Channels](Channels).

## More images for long videos

A long video with a fixed handful of images repeats them for an hour. **More images for long videos** adds images as the narration gets longer.

| Option | What it does | Default / range |
| --- | --- | --- |
| **More images for long videos** | Adds images, planned from the expected length when the run starts. The extra images are shared among the ticked prompts in order, up to 240 in all. Each is one more image call. | Off |
| **How the rate is given** | **Every N minutes**: one image every N minutes of narration (0.25 to 60). **N per hour**: N images per hour (1 to 240). Switching keeps the same rate. | One image every 2 minutes |

### Turn it on

1. Tick at least one image prompt.
2. Switch on **More images for long videos**.
3. Pick **Every N minutes** or **N per hour** and type the rate.
4. Read the line under it, for example: "For about 60 minutes (9,000 words expected, set on Review): 30 images, 26 more than the prompts' 4."

How the count is worked out:

- The length is the pasted article's own word count, or else **Expected article words per video** in the right rail, at 150 spoken words a minute.
- Each ticked prompt keeps its own **Number** as a minimum. The extra images are handed to the ticked prompts one at a time, in the order they were ticked.
- When the prompts' own Numbers already cover the length, nothing is added.
- The count stops at 240. After that the slideshow repeats its images, as it does whenever the narration outlasts them.
- The count is planned once, when the run starts. A narration that turns out longer is still covered by the slideshow cycling its images.

Turning it on also switches **Motion** (in the **Video and style** row) from **Zoom in and out** to **Mix of both**, so a long video pans and zooms by turns. A **Pan across** or **Still** you picked stays, and you can pick Zoom again.

## Scenes from the article

Without it, every image of a prompt is drawn from the same text, so three images of one prompt tend to show the same picture three times. **Scenes from the article** gives every image its own moment.

| Option | What it does | Default |
| --- | --- | --- |
| **Scenes from the article** | Before the images are drawn, the project's AI model reads the article once and writes a short scene for each image. Scene 1 comes from the article's opening and the last from its ending, in the order the images are shown. Each image is drawn from its prompt with its scene added. One more AI call per run. | Off |

Where the scene goes:

- A prompt with `{{Scene}}` gets the scene there, for example a line `Scene: {{Scene}}` under the prompt's first sentence. `{{Scene}}` is never a field to fill in: Slopify writes it.
- A prompt without it gets the scene as a line of its own after its first paragraph, so any prompt works unchanged.
- With the switch off, a line holding `{{Scene}}` is left out and the prompt is drawn as it always was.
- A **thumbnail** prompt with `{{Scene}}` gets a scene from a call of its own (one more AI call, in the Thumbnail stage, so changing the thumbnail never remakes the images): the article's most striking moment, the one that would make someone click, taken from anywhere in the article. With **Thumbnails** set to 3, each thumbnail gets a different moment, so the A/B test compares ideas as well as compositions. A thumbnail prompt without `{{Scene}}` is drawn as it always was.

A prompt using `{{Slopify_Scene}}`, `{{Slopify_Setting}}` or `{{Slopify_Mood}}` gets its scene even with the switch off, and the switch shows it as **On for** that prompt. `{{Slopify_Setting}}` adds the era and place of each moment, so the picture stays period-correct, and `{{Slopify_Mood}}` its feeling, for light and colour; both are written by the same call. See [Keywords Slopify fills](Keywords-Slopify-Fills#per-picture).

The style, composition and everything else in the prompt stay as written; only the scene changes from image to image. The scenes are written once the article exists, and each image waits for them. Switching it on or off in **Edit project** remakes the images.

## Looks looked up: {{Appearance}}

An image model draws a named figure from whatever it half remembers, so the same character can look different in every image. Put `{{Appearance}}` in an image, thumbnail, establishing or shorts image prompt, for example a line `Looks: {{Appearance}}`, and Slopify looks the looks up instead.

- Once the article exists, one call to the project's AI model **searches the web** for how the video's subject and every named character in the article look: build, face, clothing, colours and the marks that make them recognisable. Where depictions disagree, it describes the most iconic one.
- With **Scenes from the article** on, each picture gets the looks of the figures its scene names (up to three besides the subject), so a scene of a place or of someone else isn't redrawn with the subject in it; a scene that names no one leaves the looks line out. A thumbnail always gets the subject's look. Without scenes, every picture gets the subject's look.
- There is no switch: a prompt with `{{Appearance}}` turns it on, and prompts without it are drawn as they always were. `{{Appearance}}` is never a field to fill in.
- It costs one more AI call per run, with web search. The pictures that use it wait for it.
- `{{Slopify_Appearance}}` is the new name of `{{Appearance}}` and means exactly the same.

## How long each image stays on screen

**Seconds per image** (1 to 600, default 15) is in the **Video and style** row, beside **Cuts**, **Zoom** and **Motion**. When cuts follow the narration, it is the target length and a cut waits for a sentence end. When images run out, they start again. See [Play Video and Style](Play-Video-and-Style).

## The thumbnail and the shorts

The thumbnail and the shorts' images use the same image provider and model. If the video's images are not generated but the thumbnail is, the thumbnail gets its own **Provider**, **Model** and **Effort** pickers in the **Outputs** row. See [Play Outputs](Play-Outputs) and [Shorts](Shorts).

## What the row summary says

Folded, the Images row lists each ticked prompt with its Number, the provider and model, and "establishing image" when one is on, for example `Scene × 6, Portrait × 2 · fal.ai · flux-pro · establishing image`. With Provide it says how many of your images there are.

## Tips

- Keep one style across the video with an establishing image, instead of repeating the style in every image prompt.
- Picture reviews can check each image for malformed hands, stray text and a match with its brief. They need Claude Code or Codex as the reviewer. See [Reviews and Checkpoints](Reviews-and-Checkpoints).
- After the run, you can add, remove or regenerate single images on the project page. See [Editing a Project](Editing-a-Project).

## Related pages

- [Play Overview](Play-Overview)
- [Prompts](Prompts)
- [Keywords Slopify fills](Keywords-Slopify-Fills)
- [Cast Library](Cast-Library)
- [Channels](Channels)
- [Models](Models)
- [Costs and Run Cost](Costs-and-Run-Cost)
- [Play Video and Style](Play-Video-and-Style)
- [Play Outputs](Play-Outputs)