---
title: Batches, loops and logic
display: 'Batches, *loops* and logic'
nav_title: Batches and loops
summary:
  - "Run one working pipeline over a list with Split Text and Loop, then filter, gather and branch the results."
  - "Build and check the single-item version first: a loop repeats mistakes as faithfully as good work."
  - "Cost is linear: ten items cost about ten times one, and the badge shows it before you run."
description: Run one pipeline over a list with Split Text and Loop, keep part of a list with Slice Items, gather results with Group Items and branch with If-Else.
section: workflows
order: 60
type: guide
level: Scale
roles: [all]
time: { read: 9, do: 20 }
needs: ["A workflow that works on one item", "Credits for the number of items you plan to run"]
last_verified: 2026-11-11
models_checked: ["GPT Image 2.5 Sunburst"]
related: [nodes-reference, prompts-and-llm-nodes, approval-and-review-steps, debug-a-workflow]
articles: [batches-loops-and-logic, approval-and-review-steps]
thumb: workflows/batches-loops-and-logic/thumb.webp
video: tutorials/batches-loops-and-logic.mp4
video_caption: "Batches, loops and logic"
chapters:
  - { at: 0, title: "The four sheets to make", image: "/media/images/workflows/batches-loops-and-logic/ch-01-the-four-sheets-to-make.webp" }
  - { at: 20.5, title: "Open the workflow", image: "/media/images/workflows/batches-loops-and-logic/ch-02-open-the-workflow.webp" }
  - { at: 36, title: "Add Output B", image: "/media/images/workflows/batches-loops-and-logic/ch-03-add-output-b.webp" }
  - { at: 57.5, title: "Add Output A", image: "/media/images/workflows/batches-loops-and-logic/ch-04-add-output-a.webp" }
  - { at: 69.5, title: "Add Output D", image: "/media/images/workflows/batches-loops-and-logic/ch-05-add-output-d.webp" }
  - { at: 101.5, title: "Run everything", image: "/media/images/workflows/batches-loops-and-logic/ch-06-run-everything.webp" }
  - { at: 128, title: "Duplicate it", image: "/media/images/workflows/batches-loops-and-logic/ch-07-duplicate-it.webp" }
  - { at: 139, title: "Remove the character", image: "/media/images/workflows/batches-loops-and-logic/ch-08-remove-the-character.webp" }
  - { at: 167.5, title: "Add a Loop", image: "/media/images/workflows/batches-loops-and-logic/ch-09-add-a-loop.webp" }
  - { at: 221, title: "Recap", image: "/media/images/workflows/batches-loops-and-logic/ch-10-recap.webp" }
steps:
  - title: "Open the workflow and make room"
    text: "Open **Anatomy reference sheets** from the **Project** tab, click **Switch to Edit Mode**, zoom out and drag the framing rule under the character."
  - title: "Add Output B, the pencil sketch"
    text: "Press Option+I for an **Image Generator** on **GPT Image 2.5 Sunburst** and name it `4. Output B - Pencil`. Wire **3. Final Prompt B** into **Prompt**, then the character and Output C into its reference input."
    prompt: |-
      Redraw this character illustration as a traditional graphite pencil sketch on textured off-white paper.

      Keep the exact same pose, proportions, silhouette, outfit, camera angle and framing as the source.

      PRESERVE EVERY ELEMENT OF THE CHARACTER: capes, cloaks, hoods, belts, straps, armor, gloves, boots, bags, weapons, wings, tails, jewelry and hair must all be drawn, with the same shape, length and placement as in the reference. Do not remove or simplify any part of the design.

      Hand-drawn look: confident construction lines, light underdrawing still visible, expressive contour linework, no cross-hatching, a few darker accent strokes for depth, subtle eraser highlights. Monochrome graphite only — no color, no digital gradients, no background scenery, just clean paper.

      Style: professional character design study sketch from an artist's sketchbook, figure drawing reference, no text, no watermark.

      Output the image in a 9:16 vertical portrait aspect ratio.

      If the source image is cropped (bust, half-body, or cut off at any edge), extend it into a COMPLETE FULL-BODY figure: reconstruct the missing torso, hips, legs and feet in a natural continuation of the pose, and frame the whole figure head-to-toe, centered in the 9:16 vertical frame with a small margin of empty space above the head and below the feet. Keep the same character design, proportions, style and camera direction. If the source already shows the full body, keep the pose and framing, just fit it into the 9:16 vertical frame.
  - title: "Add Output A, the gray sculpt"
    text: "Press Option+I again for `4. Output A - Gray Sculpt`. Wire **3. Final Prompt A** into **Prompt** and Output B into its reference input."
    prompt: |-
      Convert this character illustration into a neutral gray digital sculpture — an artist's anatomy/form reference maquette, like a ZBrush sculpt rendered with a matte gray matcap material.

      Keep the exact same pose, proportions, silhouette, camera angle and framing as the source.

      PRESERVE EVERY ELEMENT OF THE CHARACTER. Do not delete, simplify or omit anything that exists in the source image: capes, cloaks, coats, hoods, scarves, skirts, belts, straps, harnesses, armor pieces, pauldrons, gloves, boots, bags, weapons, tools, wings, tails, horns, jewelry, hair strands and any other accessory or equipment must all be present and sculpted as part of the statue, with the same shape, size, length, flow and placement as in the reference. Sculpt hair and flowing fabric (cape, cloak, coat tails) as solid gray sculpted volumes with the same drape and direction of movement. If an element is unclear in the source, sculpt a plausible version of it rather than removing it. Nothing may be replaced by bare body.

      Under the outfit, the body reads as a simple, modest, opaque full-coverage form-fitting bodysuit (long sleeves, full legs, high neckline) sculpted as part of the statue, so the underlying muscle structure and body forms read clearly through smooth, seamless surfacing. Fully clothed, non-revealing, no skin exposure beyond head and hands, no suggestive posing — a museum-style study figurine.

      Neutralize only SURFACE APPEARANCE, never geometry: remove all color, hair color, fabric texture, patterns, prints, logos and background detail, and render every element — body, outfit, cape, hair, gear — in one uniform mid-gray clay material with soft studio lighting, gentle specular matcap shading and a subtle rim light, on a plain neutral dark gray backdrop.

      Style: professional anatomy study reference, sculpting maquette, clean topology and clearly defined muscle groups and form volumes, no text, no watermark.

      Output the image in a 9:16 vertical portrait aspect ratio.

      If the source image is cropped (bust, half-body, or cut off at any edge), extend it into a COMPLETE FULL-BODY figure: reconstruct the missing torso, hips, legs and feet in a natural continuation of the pose, and frame the whole figure head-to-toe, centered in the 9:16 vertical frame with a small margin of empty space above the head and below the feet. Keep the same character design, proportions, style and camera direction. If the source already shows the full body, keep the pose and framing, just fit it into the 9:16 vertical frame.
  - title: "Add Output D, the textured sculpt"
    text: "Add `4. Output D - Textured`. Wire **3. Final Prompt D** into **Prompt**, then the character and Output A into its reference input."
    prompt: |-
      Convert this character illustration into a 3D digital sculpture — a fully textured render of the character as a collectible statue / game-ready model, like a ZBrush sculpt with the original color textures and materials applied.

      Keep the exact same pose, proportions, silhouette, camera angle and framing as the source.

      KEEP THE ORIGINAL COLORS AND TEXTURES. Match the character's palette exactly as seen in the reference: skin tone, hair color, eye color, and every color of the outfit, cape, armor and accessories. Reproduce the material qualities of each part — matte or worn fabric with visible weave, brushed or scratched metal with real specular highlights, glossy leather, rough stone, translucent gems, glowing emissive parts — so each surface reads as its real material, not as clay.

      PRESERVE EVERY ELEMENT OF THE CHARACTER. Do not delete, simplify or omit anything that exists in the source image: capes, cloaks, coats, hoods, scarves, skirts, belts, straps, harnesses, armor pieces, pauldrons, gloves, boots, bags, weapons, tools, wings, tails, horns, jewelry, hair strands and any other accessory or equipment must all be present and sculpted, with the same shape, size, length, flow and placement as in the reference. Sculpt hair and flowing fabric as solid volumes with the same drape and direction of movement. If an element is unclear in the source, sculpt a plausible version of it rather than removing it.

      The figure is fully clothed and modest — no skin exposure beyond what the outfit already leaves visible, no suggestive posing. A display statue.

      Render it as a clean 3D render: crisp sculpted geometry, high-detail surface work, soft neutral studio lighting with a subtle rim light, soft contact shadows, on a plain neutral dark gray backdrop. No painted-illustration look, no 2D shading — it must read as a physical sculpted and painted figure.

      Style: professional character turnaround render, collectible figure product shot, no text, no watermark.

      Output the image in a 9:16 vertical portrait aspect ratio.

      If the source image is cropped (bust, half-body, or cut off at any edge), extend it into a COMPLETE FULL-BODY figure: reconstruct the missing torso, hips, legs and feet in a natural continuation of the pose, and frame the whole figure head-to-toe, centered in the 9:16 vertical frame with a small margin of empty space above the head and below the feet. Keep the same character design, proportions, style and camera direction. If the source already shows the full body, keep the pose and framing, just fit it into the 9:16 vertical frame.
  - title: "Check the estimate and run everything"
    text: "Press Cmd+0, read the estimate next to **Run**, then press Cmd+Shift+Enter to run every node; it starts with no confirmation. The results stay on the nodes."
  - title: "Duplicate the workflow for a cast"
    text: "Open the **...** menu at the top right, choose **Duplicate Workflow**, and click **Switch to Edit Mode** in the copy."
  - title: "Replace the character with two inputs"
    text: "Delete **1. Character Illustration**, press Shift+I twice, and pick a character in each with **Select from Library**: the ranger and Queen Pip."
  - title: "Group the inputs into one list"
    text: "Click **+**, search `Group Items` and add it. Wire **Image 1** and **Image 2** into it."
  - title: "Add a Loop"
    text: "Search `Loop`, add it, wire **Group Items 1** into its **Input List** and click **Add Loop End**."
  - title: "Wire the loop"
    text: "Wire the Loop's item into the references of Output C, Output B and Output D, and Output D's result into **Loop 1 (End)**. Test one item first with **Slice Items**; the estimate doubles for two characters."
updated: 2026-11-11
canonical: https://learn.scenario.com/workflows/batches-loops-and-logic
---

Last updated: 2026-11-11

# Batches, loops and logic

> Run one pipeline over a list with Split Text and Loop, keep part of a list with Slice Items, gather results with Group Items and branch with If-Else.

In short:

- Run one working pipeline over a list with Split Text and Loop, then filter, gather and branch the results.
- Build and check the single-item version first: a loop repeats mistakes as faithfully as good work.
- Cost is linear: ten items cost about ten times one, and the badge shows it before you run.


Four reference sheets of one character are only useful if they match. In this lesson you chain the four generators of the **Anatomy reference sheets** workflow, so each sheet is made from the one before, run the whole graph with one shortcut, and then run it for a whole cast with **Group Items** and a **Loop**. The model on every generator is **GPT Image 2.5 Sunburst**.

:::note As of November 2026
The models, settings and screens in this article are the ones Scenario offered in November 2026. New models and updates arrive often, so check the app for newer options before you choose.
:::

:::steps Quick version
1. Open the workflow from the **Project** tab and click **Switch to Edit Mode**.
2. Add three Image Generators (`Option+I`): **Output B - Pencil**, **Output A - Gray Sculpt** and **Output D - Textured**. Wire each one's prompt, and use the sheet before it as a reference.
3. Fit the graph (`Cmd+0`), check the estimate next to **Run**, and run every node (`Cmd+Shift+Enter`).
4. For a cast, open the menu and choose **Duplicate Workflow**. In the copy, replace the character with two Image nodes, gather them with **Group Items** and wire the list into a **Loop**.
5. Wire the Loop's item into the generators that took the character, and the last sheet into **Loop End**. Test one item first with **Slice Items**.
:::

## Before you start

- **The workflow from the earlier lessons.** This lesson picks up **Anatomy reference sheets** where [Prompts that repeat](https://learn.scenario.com/workflows/prompts-and-llm-nodes.md) left it: one character image (**1. Character Illustration**), a framing rule (**1. Full Body + 9:16 (empty = keep framing)**), four style prompts, four Prompt Builders (**3. Final Prompt A** to **D**) and one generator, **4. Output C - Anatomy**, from [Build your first workflow](https://learn.scenario.com/workflows/build-your-first-workflow.md).
- **Where it lives.** **Workflows** › **Workflows** in the top bar, **Project** tab. A workflow opens in run-only mode; click **Switch to Edit Mode** to change it.
- **What a run costs.** The badge next to **Run** shows the estimate for one full run before you start. A loop multiplies it by the number of items.

| Sheet | Generator | Prompt | References |
|---|---|---|---|
| Anatomy sketch | **4. Output C - Anatomy** | **3. Final Prompt C** | The character |
| Pencil sketch | **4. Output B - Pencil** | **3. Final Prompt B** | The character, then Output C |
| Gray sculpt | **4. Output A - Gray Sculpt** | **3. Final Prompt A** | Output B |
| Textured sculpt | **4. Output D - Textured** | **3. Final Prompt D** | Output A, then the character |

## Step 1: Open the workflow and make room

Open **Anatomy reference sheets** from the **Project** tab and click **Switch to Edit Mode**. Press minus (or click the zoom-out button in the right toolbar) to zoom out, then drag the framing rule down, under the character, so the column of nodes is clear.

![The workflow in Edit Mode: the character, the framing rule, four style prompts, four Prompt Builders and one generator, Output C - Anatomy](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/9a9f782283c810729cf84a03c0bdb0ccef03d2eed255dc27cd6330ea9dcab30e.webp)

## Step 2: Add Output B, the pencil sketch

Press `Option+I` for a new **Image Generator**. It lands on the canvas with the model already set to **GPT Image 2.5 Sunburst**, the same as the first generator. Drag it next to **3. Final Prompt B** and double-click its title to name it `4. Output B - Pencil`.

Clicking a generator opens its settings panel on the left:

- **Model:** the image model this node runs.
- **Prompt:** filled by a wire from a Prompt Builder, or typed in.
- **Reference Images:** the images the model works from.
- **Settings:** **Image Count**, **Dimensions**, **Quality** and **Background**. Leave them on the defaults shown (1 and **Auto**).
- **Run This Node:** runs this generator alone.

![The settings panel of a new Image Generator: Model GPT Image 2.5 Sunburst, Prompt, Reference Images and Settings](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/8d7d3bfba55cbec0c494a55638a203c46320f0b3f34046c71544517b3008d19e.webp)

Now wire it:

1. Drag from the output of **3. Final Prompt B** to the generator's **Prompt** input.
2. Wire the character into its reference input, then wire the result of **4. Output C - Anatomy** into the same input. One port takes both images.

Wires follow the colors of the ports: blue for text, green for images. A text wire never connects to an image port.

The prompt, as run: **3. Final Prompt B** sends the pencil prompt, a blank line, then the framing rule.

```prompt
Redraw this character illustration as a traditional graphite pencil sketch on textured off-white paper.

Keep the exact same pose, proportions, silhouette, outfit, camera angle and framing as the source.

PRESERVE EVERY ELEMENT OF THE CHARACTER: capes, cloaks, hoods, belts, straps, armor, gloves, boots, bags, weapons, wings, tails, jewelry and hair must all be drawn, with the same shape, length and placement as in the reference. Do not remove or simplify any part of the design.

Hand-drawn look: confident construction lines, light underdrawing still visible, expressive contour linework, no cross-hatching, a few darker accent strokes for depth, subtle eraser highlights. Monochrome graphite only — no color, no digital gradients, no background scenery, just clean paper.

Style: professional character design study sketch from an artist's sketchbook, figure drawing reference, no text, no watermark.

Output the image in a 9:16 vertical portrait aspect ratio.

If the source image is cropped (bust, half-body, or cut off at any edge), extend it into a COMPLETE FULL-BODY figure: reconstruct the missing torso, hips, legs and feet in a natural continuation of the pose, and frame the whole figure head-to-toe, centered in the 9:16 vertical frame with a small margin of empty space above the head and below the feet. Keep the same character design, proportions, style and camera direction. If the source already shows the full body, keep the pose and framing, just fit it into the 9:16 vertical frame.
```

## Step 3: Add Output A and Output D

Press `Option+I` again for **4. Output A - Gray Sculpt**. Its prompt is **3. Final Prompt A**, and its only reference is **Output B**.

The prompt, as run:

```prompt
Convert this character illustration into a neutral gray digital sculpture — an artist's anatomy/form reference maquette, like a ZBrush sculpt rendered with a matte gray matcap material.

Keep the exact same pose, proportions, silhouette, camera angle and framing as the source.

PRESERVE EVERY ELEMENT OF THE CHARACTER. Do not delete, simplify or omit anything that exists in the source image: capes, cloaks, coats, hoods, scarves, skirts, belts, straps, harnesses, armor pieces, pauldrons, gloves, boots, bags, weapons, tools, wings, tails, horns, jewelry, hair strands and any other accessory or equipment must all be present and sculpted as part of the statue, with the same shape, size, length, flow and placement as in the reference. Sculpt hair and flowing fabric (cape, cloak, coat tails) as solid gray sculpted volumes with the same drape and direction of movement. If an element is unclear in the source, sculpt a plausible version of it rather than removing it. Nothing may be replaced by bare body.

Under the outfit, the body reads as a simple, modest, opaque full-coverage form-fitting bodysuit (long sleeves, full legs, high neckline) sculpted as part of the statue, so the underlying muscle structure and body forms read clearly through smooth, seamless surfacing. Fully clothed, non-revealing, no skin exposure beyond head and hands, no suggestive posing — a museum-style study figurine.

Neutralize only SURFACE APPEARANCE, never geometry: remove all color, hair color, fabric texture, patterns, prints, logos and background detail, and render every element — body, outfit, cape, hair, gear — in one uniform mid-gray clay material with soft studio lighting, gentle specular matcap shading and a subtle rim light, on a plain neutral dark gray backdrop.

Style: professional anatomy study reference, sculpting maquette, clean topology and clearly defined muscle groups and form volumes, no text, no watermark.

Output the image in a 9:16 vertical portrait aspect ratio.

If the source image is cropped (bust, half-body, or cut off at any edge), extend it into a COMPLETE FULL-BODY figure: reconstruct the missing torso, hips, legs and feet in a natural continuation of the pose, and frame the whole figure head-to-toe, centered in the 9:16 vertical frame with a small margin of empty space above the head and below the feet. Keep the same character design, proportions, style and camera direction. If the source already shows the full body, keep the pose and framing, just fit it into the 9:16 vertical frame.
```

Then add **4. Output D - Textured**. Its prompt is **3. Final Prompt D**, and it takes two references: **Output A** for the shape, and the character for her real colors.

The prompt, as run:

```prompt
Convert this character illustration into a 3D digital sculpture — a fully textured render of the character as a collectible statue / game-ready model, like a ZBrush sculpt with the original color textures and materials applied.

Keep the exact same pose, proportions, silhouette, camera angle and framing as the source.

KEEP THE ORIGINAL COLORS AND TEXTURES. Match the character's palette exactly as seen in the reference: skin tone, hair color, eye color, and every color of the outfit, cape, armor and accessories. Reproduce the material qualities of each part — matte or worn fabric with visible weave, brushed or scratched metal with real specular highlights, glossy leather, rough stone, translucent gems, glowing emissive parts — so each surface reads as its real material, not as clay.

PRESERVE EVERY ELEMENT OF THE CHARACTER. Do not delete, simplify or omit anything that exists in the source image: capes, cloaks, coats, hoods, scarves, skirts, belts, straps, harnesses, armor pieces, pauldrons, gloves, boots, bags, weapons, tools, wings, tails, horns, jewelry, hair strands and any other accessory or equipment must all be present and sculpted, with the same shape, size, length, flow and placement as in the reference. Sculpt hair and flowing fabric as solid volumes with the same drape and direction of movement. If an element is unclear in the source, sculpt a plausible version of it rather than removing it.

The figure is fully clothed and modest — no skin exposure beyond what the outfit already leaves visible, no suggestive posing. A display statue.

Render it as a clean 3D render: crisp sculpted geometry, high-detail surface work, soft neutral studio lighting with a subtle rim light, soft contact shadows, on a plain neutral dark gray backdrop. No painted-illustration look, no 2D shading — it must read as a physical sculpted and painted figure.

Style: professional character turnaround render, collectible figure product shot, no text, no watermark.

Output the image in a 9:16 vertical portrait aspect ratio.

If the source image is cropped (bust, half-body, or cut off at any edge), extend it into a COMPLETE FULL-BODY figure: reconstruct the missing torso, hips, legs and feet in a natural continuation of the pose, and frame the whole figure head-to-toe, centered in the 9:16 vertical frame with a small margin of empty space above the head and below the feet. Keep the same character design, proportions, style and camera direction. If the source already shows the full body, keep the pose and framing, just fit it into the 9:16 vertical frame.
```

![All four generators on the canvas, each wired to its Prompt Builder, with the references chained from one output to the next](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/b7d5dc87edc289b505c9d78b678bbd64e5e0680597102e0d14e999ae305df1cc.webp)

This is the chain: each sheet is made from the one before, so no detail drifts between them. A strap that stays on the anatomy sketch stays on the pencil sketch, and so on through the gray and textured sculpts. That is what makes the four sheets match.

## Step 4: Check the estimate and run everything

Press `Cmd+0` to fit the whole graph on screen. The badge next to **Run** now shows what one full run costs.

Press `Cmd+Shift+Enter` to run every node. The run starts at once, with no confirmation, so read the badge first. Clicking **Run** in the top right does the same.

Each node shows its state while the workflow runs: **Running** on the node that is working, **Waiting** on the ones whose inputs are not ready yet. The sheets arrive in order, C, then B, A and D, because each one waits for its reference.

![The run in progress: Output C - Anatomy is Running, while Output B, A and D are Waiting for their references](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/e94f9014ff24bd519d8220682f388f4b09babdaccd173927931974ea5482ba1f.webp)

A couple of minutes later all four sheets are on their nodes: the same pose and the same gear in all four.

![The four sheets on their nodes: anatomy sketch, pencil sketch, gray sculpt and textured sculpt of the same character](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/4320947a95b7a3fae9b8dfe250934337344e0e1c385e31a65d22861f23c86a7c.webp)

The workflow saves as you work, results included, and Cmd+S (Ctrl+S on Windows) saves right away: the four sheets stay on their nodes, and the next lesson uses them.

:::check
**Accept when:** all four sheets show the same pose, proportions, cape, straps and gear; the pencil and gray sheets keep every part of the outfit; the textured sculpt has the character's own colors.
**Reject when:** a sheet dropped or added a part of the outfit, the framing changed between sheets, or one sheet looks like a different character. Fix the prompt or the references of the first sheet that drifted, then run from there.
:::

## Step 5: Duplicate the workflow for a cast

To run the same pipeline over several characters, work in a copy so the original stays as it is. Open the **...** menu at the top right and choose **Duplicate Workflow**. The copy opens as **Anatomy reference sheets (copy)** in run-only mode; click **Switch to Edit Mode**.

![The workflow menu: Lock Workflow, Export Workflow, Duplicate Workflow, Copy Workflow ID and the canvas actions](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/671e8815c571425e3eaa78b1e7f9fdfe12d662090740799fd3ab202150274cdf.webp)

## Step 6: Replace the character with two inputs

1. Click the **1. Character Illustration** node and press Delete. Its three reference wires go with it.
2. Press `Shift+I` twice to add two **Image** nodes. Each one offers **Drag & drop or paste**, **Select from Library** and **Upload**.
3. In each, choose **Select from Library** and pick a character: the ranger from the earlier lessons in **Image 1**, and a new one, Queen Pip, in **Image 2**.

![Two new Image nodes, Image 1 and Image 2, each with Drag & drop or paste, Select from Library and Upload](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/328b5696744118fa05f180c4c694befb80cd7179b63ae306cd9a0d41326894a4.webp)

## Step 7: Group the inputs into one list

Open the node list with **+** in the right toolbar, type `Group Items` in the search bar and add it. **Group Items** ("Bundle multiple assets") takes separate images and passes them on as one list, in slot order.

Wire **Image 1** and **Image 2** into it. They now travel as one list.

![The node list searched for Group Items, which shows under Utilities as Bundle multiple assets](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/1795da0b869dd319bc48def8b01b6561226eb4ca2d5583aa88ed05c22f6104aa.webp)

## Step 8: Add a Loop

Open the node list again and search for `Loop` ("Iterate over a list of inputs"). Add it, wire the output of **Group Items 1** into its **Input List**, and click **Add Loop End**. The Loop gets a partner node, **Loop 1 (End)**: "All results from the loop end up here."

![The node list searched for loop list, with Loop under Utilities, next to Group Items 1 holding Image 1 and Image 2](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/d87ca0cc111286d9ce5a693959e88758b280ac1d1c7395ecf8b348b3bf68f4e5.webp)

## Step 9: Wire the loop

1. Wire the Loop's item output to every place the character went before: the reference inputs of **Output C**, **Output B** and **Output D**.
2. Wire the result of **Output D - Textured** into **Loop 1 (End)**, to collect the results.

Everything between the Loop and its end now runs once per character. The Loop shows which item it is on (**Next Iteration: #1 of 2**), and its **Concurrent Runs** switch lets items run at the same time instead of one after the other.

![Group Items 1 feeding Loop 1, which shows Next Iteration #1 of 2 and the Concurrent Runs switch, with Loop 1 (End) under it](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/b763380510f9b7a1fbe1b34dae9040d9decc6968d319003b4537b3384fee9c27.webp)

The estimate next to **Run** doubles: two characters cost two runs. For a long list, test on one item first with **Slice Items** (below) before the full run.

When you are done, close the copy with the **X** at the top left. The original workflow stays as it was.

## Branch with If-Else

When two cases need different steps, **If-Else** sends each run down one branch. The framing rule in this workflow is the simple version of the same idea: leave it empty and every sheet keeps the image's own framing.

If-Else checks its cases from top to bottom and runs the first one whose conditions are met. If none is met, the run follows **Else**.

- **Inputs:** connect the nodes whose results you want to check.
- **Case outputs:** one per case, **Case 1**, **Case 2** and so on.
- **Setting up a case:** open the node's settings panel and choose a field (a result from a connected node), an operator and, when it needs one, a value. **Add Condition** requires more than one check in a case, with an AND / OR toggle; **Add A New Case** adds another branch.
- **Operators:** is empty, is not empty, contains, does not contain, = and !=.

You write no expressions in the node. Rename the source node first: its title is how the field shows in the condition. Put the most specific case first, since the first case met wins, and always connect **Else** so every run ends somewhere.

## The other list nodes

| Node | What it does | Use it for |
|---|---|---|
| **Group Items** | Bundles separate assets into one list, ordered by input slot | Gathering parallel branches or a few chosen images |
| **Split Text** | Splits one text into a list by **Comma**, **Space**, **Semicolon**, **Period**, **Newline** or **Custom**; shows the items as chips | A list of names, products or prompts |
| **Slice Items** | Passes on part of a list | Testing on one or two items; chunks of a long list |
| **Loop** | Runs the nodes between it and **Loop End** once per item | Any pipeline that must repeat per item |
| **Approval** | Pauses the run until someone approves or rejects | A quality gate before an expensive step |

An **Image** input node can also hold up to 10 images with **Enable Multiple Images** turned on, and passes them all on.

The classic pattern from text: a Text node with `cat, dog, bird`, **Split Text** on **Comma**, a **Loop**, then a Prompt Builder (`A photo of a [current item]`) and an Image Generator inside it. Three items in, three images out.

![A Text node with "teapot, lantern, armchair" feeds Split Text, which feeds a Loop containing a Prompt Builder and an Image Generator. The Loop outputs a collection of three images. An optional Slice Items node keeps only two items for a test run](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/86f365e67488970f21cd81a343d6d51bb3c30aa4a681ced7e2803a82302bc6eb.svg "One list in, one result per item out. Cost scales with the list.")

### Slice Items modes

| Mode | What passes |
|---|---|
| **Default** | **Number of Items** items starting at **From Item**. From Item is zero-based (0 is the first item); a negative value counts from the end (-1 is the last item) |
| **Randomize** | A random sample of **Number of Items**. Lock the seed to get the same sample every run |
| **Choose** | The run pauses and a person picks exactly **Number of Items** items, with your **Message** as the instruction. Inside a Loop, each iteration asks for its own pick |

With a list of 10: Default, From 0, Count 3 gives items 0, 1 and 2; Default, From -1, Count 1 gives the last item. If the range runs past the end, you get fewer items, with no error.

## Keep batches under control

- **Test small.** Run one to five items first, and check them with the same criteria you will use on the full batch.
- **Keep the loop body lean.** Anything that does not change per item stays outside the Loop and is wired in, so it runs once.
- **Put the expensive step after a filter.** Generate cheap stills for every item, keep the good ones with Slice Items in **Choose** mode or an **Approval** node, then upscale or animate only those.
- **Chunk long lists.** Process a long collection in chunks of 10 to 50 with Slice Items.
- **Leave a note.** A sticky note inside the loop explains the logic to the next person.

## Common problems

| What you see | What to do |
|---|---|
| A sheet doesn't match the others | Check its references: each sheet must take the one before it, and Output D also takes the character |
| The run started before you checked the cost | `Cmd+Shift+Enter` runs at once. Read the badge next to **Run** first, and stop any nodes still waiting if needed |
| The Loop does nothing | Its **Input List** is not connected, or the nodes that should repeat are not between the Loop and **Loop End** |
| Only one character came out | The Loop's item is not wired into every node that took the character before, or **Output D** is not wired into **Loop End** |
| The estimate is higher than planned | Every item multiplies the run. Test with **Slice Items** on one item, then raise the count |
| Split Text gives one item | The delimiter does not match the text. Check for semicolons, new lines or extra spaces |
| If-Else always takes **Else** | Check the case order and that each condition reads the field from the node you expect, with the right operator |
| Slice Items returns the wrong items | In **Default** mode the first item is 0, not 1 |

## What's next

The workflow now makes four matching sheets for any number of characters. In [Image and Video Studio](https://learn.scenario.com/workflows/compose-with-image-and-video-studio.md) you tidy the canvas into named, colored groups and combine the four sheets into one reference sheet. For every node and port, see the [nodes reference](https://learn.scenario.com/workflows/nodes-reference.md).

Source: https://learn.scenario.com/workflows/batches-loops-and-logic