> ## Documentation Index
> Fetch the complete documentation index at: https://docs.memo.solutions/llms.txt
> Use this file to discover all available pages before exploring further.

# Workflows

> Reusable generation recipes — save a working canvas setup once, run it again on any brand, and explore official starters.

A Workflow is a saved, reusable generation recipe: the scenes, prompts, and reference slots of a canvas setup, packaged so it can be run again with new inputs. Build something that works once — a product-shot pipeline, a character sheet, a format converter — and stop rebuilding it by hand.

Find them under **Workflows** in the sidebar.

The **List / Grid** switch beside **Explore / My workflows** chooses preview cards or compact rows. Your browser remembers this preference; search and category filters work in both layouts. Rows retain the available edit and management actions.

Select anywhere on a Workflow card to open it. Cards share the landscape layout used by Studio projects and Collections, with the description directly beneath the title.

## Explore starters

The **Explore** tab ships official, ready-to-run starters, each with example output so you can see what it produces before spending credits:

* **Character Sheet** — turn a single character reference into a consistent multi-view sheet (chained scenes keep the character identical across poses).
* **Aspect Ratio Converter** — take an image and re-frame it cleanly across aspect ratios instead of cropping.

More starters land over time. Running a starter copies it onto your canvas wired up and ready — drop in your references, press Execute.

## Running a Workflow

1. Open a Workflow and fill its reference slots (the images or assets it needs).
2. Review the credit estimate — every Workflow prices itself before running.
3. Execute. Each scene generates in order, and outputs land on your canvas like any other generation.

### Choosing your outputs

You don't have to run everything a Workflow makes. The run panel lists every output by name — untick the ones you don't need and the run only generates (and only charges for) what's selected. Each row shows its own price and the estimate updates as you pick. One rule keeps quality honest: if a sheet is built from another output (the Character Sheet's sheets are all built from the Turnaround), that reference stays locked in and is delivered with your run — you pay for it, you get it.

### Choosing a model

Some Workflows offer a **Model** picker in the run panel — a short, curated list of models the recipe has been tested on. The Character Sheet starter, for example, runs on **Nano Banana 2** by default or **Seedream 5.0 Pro** for a premium finish at a higher credit cost. Pick a model and the per-output prices and the estimate update instantly, so the trade-off is always visible before you generate. Workflows without a picker always run on their tuned default.

If a model can't produce one of a Workflow's outputs (the Aspect Ratio Converter's 4:5 frame doesn't exist on GPT Image 2, for instance), that output is simply disabled while the model is selected — shown, not hidden, and never silently produced at the wrong ratio. You're only charged for what actually runs.

## Runs — your results, kept

Every run from a Workflow's page is saved automatically. Nothing is lost when you navigate away:

* **Runs history** — each Workflow's page keeps a **Runs** section: every past run, newest first, with the inputs it used. Come back anytime; your results are waiting.
* **Per-result actions** — click any result to view it full-size, download it, delete it, or **save it to Assets** so it becomes a reusable brand reference.
* **Save as character** — on character workflows (like the Character Sheet starter), one click saves a whole run as a single character collection in **Assets → Characters**: the turnaround becomes the cover, the other sheets ride along.
* **Download all** — one click zips a run's outputs, named by scene (e.g. `poster-9-16.png`), ready to hand off.
* **Studio → Generations** — the same results also appear in the **Workflow runs** project alongside the rest of your generations.

## Saving your own

Build the setup on the Studio canvas, then save it as a Workflow. It gets its own slug, shows up in your Workflows list, and anyone in the workspace can run it.

If **My workflows** is empty, its first-run state links straight to **Open Studio**. The Workflow illustration is interactive: hover it for the same responsive line-motion language used across Memo's other empty libraries.

## Editing a Workflow

Everything lives on the **Workflows** page: browse **Explore** and **My workflows** as a gallery or a compact list (the view toggle sits beside the tabs). Clicking a Workflow opens it ready to run; **Edit** lives in its card's ⋯ menu, and as an **Edit** button on the Workflow's own page. Official Memo starters can't be edited — run them as-is, or switch to their **Canvas** view and copy the board into your own Studio: that copy is fully yours to change and re-save as your own Workflow.

The editor has three tabs:

* **Setup** — rename the Workflow, describe what it makes, and set **Rules**: firm constraints (one per line, e.g. "Never redraw the logo") appended to every output's prompt.
* **Inputs** — the images someone provides when they run it. Rename each input, swap its default image, and give it a **role** that tells the model how faithfully to treat it:
  * **Preserve exactly** — reproduce the image as provided; for logos, products, packaging. Never redrawn or restyled.
  * **Style reference** — copy the image's look, mood, and palette, never its subject or content.
  * **Subject** — place it as the subject of the scene, kept clearly recognizable.
* **Outputs** — one card per generated image or video, each with an example thumbnail when available. Edit titles and prompts freely; format settings (aspect ratio, quality, duration) are frozen from the source brief — restructure by re-saving from an edited brief.

An output's **Uses** row shows what it receives. Besides the Workflow's inputs, **output of…** chains an *earlier output's generated result* in as an extra input — the way the Character Sheet's pose sheets build on its Turnaround. Chained outputs always run after the output they build on; when a Workflow chains, the Outputs tab opens with a **How it runs** map showing the steps at a glance.

Not sure what the model will actually be told? Every output card has a **Final prompt** toggle that shows the exact composed prompt — your prose, the Rules block, and each input's role instruction — updating live as you type.

## Workflows vs. Skills

* **Skills** are written playbooks — instructions and know-how the agent reads and applies while working with you.
* **Workflows** are executable recipes — a fixed sequence of generation scenes with defined inputs that runs the same way every time.

Use a Skill to teach the agent *how* to think about your brand; use a Workflow to make a *repeatable deliverable* one click.

## From an agent

Connected agents (see [Connect Your Agent](/get-started/connect-your-agent)) can list your Workflows and run them: `list_workflows` shows each recipe and its reference slots, `run_workflow` prices a run, and only an explicit confirmation spends credits.