Merrimo User Guide › AI generation and art

Generate artwork for panels on Mac

In Merrimo on your Mac, you can generate illustration artwork directly inside comic panels, type prompts on canvas nodes, and reroll iterations with diffusion AI engines.

Generate artwork inside a panel

Render comic illustrations on your render machines:

Render comic illustrations with the online service you set up in Settings › Providers:

1. Select any empty panel on the canvas.
2. Press Return to open the in-canvas prompt bar, or open the Generation section in the Inspector (⌥⌘I).
3. Type your scene description (for example, Rain-slicked alleyway at midnight, neon signs reflecting in puddles).
4. Click away from the panel (in the prompt bar, Return starts a new line and Esc closes it), or click Render in the Inspector. A live progress shimmer appears over the panel as the image renders.
5. Once generated, the high-resolution artwork is embedded directly into the panel frame.

A prompt of only a few words typed in the prompt bar is enhanced before it renders: the AI fills in the lighting, setting and camera, keeping your subject, and the richer prompt replaces yours in the panel. This needs an AI model set up in Settings > Providers; without one, your prompt renders as you typed it. Turn it off, or change how short a prompt has to be, in Settings > Canvas. To enhance a longer prompt, click the sparkles button, the only button in the panel's floating toolbar while the prompt bar is open.

While the Style row in the Inspector is on, the style you picked is added when the panel renders, so an enhanced prompt leaves the style out rather than describe a different one.

To start from a picture you already have, drop it on the open prompt bar or on the Prompt box in the Inspector: drag an image from elsewhere in Merrimo (the image in the Inspector, a photo in the Media tab, a character's picture), or an image file from Finder. The AI looks at the picture and writes a prompt that would recreate it, which replaces the panel's prompt; render it as usual, or edit it first. Unlike Draft with AI, it doesn't start a render by itself. While the panel's Style row is on, the prompt leaves out the picture's style, so the panel renders in the style you picked; turn the Style row off to have the prompt describe the picture's style too. This needs an AI model that can read images, set up in Settings > Providers.

Draft with AI, Refine with AI and the sparkles button write the prompt in the format the engine reads best: comma-separated tags for an SDXL engine, plain prose for every other. To choose yourself, hold the pointer over the Prompt box in the Inspector, click the Prompt Format button under the Enhance button, and choose Text or Tags; choose Automatic to follow the engine again. The choice is kept for the story, one for panels and one for characters, locations and the rest of the cast and world. An SDXL engine always writes tags, so Text is dimmed while one is selected.

Some engines can't run until something is set up first. A cloud model's engine circle is gray while AI is off or your OpenRouter connection has no API key. Hover over it to see why, or click it and choose Open Settings to add the key in Settings > Providers.

An engine circle with an orange dot can't run on any of your render machines yet, for example because an app needs a node or a model file none of them has. Select it, and the row under the engine circles names what's missing.

Reroll and iterate on variations

Explore different interpretations and seeds quickly:

1. Select the rendered panel and press ⌘R (Reroll).
2. Merrimo renders the panel again exactly as its shown image was made, with a new seed: the same prompt and negative prompt, engine, size and References, whatever the Inspector is set to now. The previous iteration stays in History.
3. Use the variation controls in the Inspector to control how closely the new render adheres to the current composition.

Artwork rendered with ComfyUI keeps the graph that made it inside the image file, as ComfyUI's own saves do. Drag a panel's or character's picture out of Merrimo to save it as a file, then drop the file on ComfyUI's canvas to see how it was made.

Generate again with the inspector closed

When the inspector is closed, a tray floats under the picture of the character, location, prop or other world entry you're looking at under Cast & World, and under the panel you've selected on the canvas. It has what you need to try again without opening the inspector or asking the Agent:

• The prompt runs along the top, two lines of it; a longer prompt ends in …. Click it to change it in place: the field grows upward to show the whole text, up to 12 lines, then scrolls. Press Return or click away to keep the change (Esc drops it). The change is saved, just as if you'd typed it in the inspector, and ⌘Z undoes it.
• The style button, named for the story's style, turns it on or off for this picture: click it once to push it in or out. On, it's filled with the accent color; off, it's empty but keeps the style's name, so you can see what one click brings back.
• Generate renders straight away. A character, location or prop gets a new take, which waits in the strip beside the current one. While it renders, the button reads Generating….

On the canvas the tray floats along the bottom edge, clear of the minimap in the corner. Open the inspector and the tray goes away, since the inspector has all of it.

Other controls you can add to the tray work like their counterparts in the inspector:

• Model, Shot and Aspect show what's chosen. Click one and the inspector's picker opens over it: type to narrow it down, or pick from the grid of sample pictures. At the bottom of the window it opens upward.
• Shuffle decides the seed. On, every take gets a new seed. Off, each take reuses the last one, so only what you change varies. The seed's number is in the inspector.
• Each reference is its own button, with its picture and name: click it to send it or leave it out. A picture that's left out is dimmed. After the first three, the rest are in a + menu at the end.

The tray's buttons stay live while a take renders; what you change applies to the next take.

Choose what's in the tray

Like a toolbar, the tray holds what you put in it. Control-click the tray and choose Customize Tray…. A palette opens just above the tray: drag items from it down into the tray, drag items out of the tray onto the palette, or drag the default set back in. Click Done when you're finished. You can add the model, the shot, the aspect, Shuffle and Flexible Space, and on world entries and panels the references, plus Avoid on world entries. Take everything else out and the tray is a single Generate button.

Characters, locations, props, the rest of the world and canvas panels each keep their own tray, so your cast can render alike while panels show different controls. You can make the same choices in Settings > Canvas, under Generate Tray. Your trays are kept on this Mac.

Make a short video from a panel

Animate turns a panel's image into a few seconds of video. A panel or loose image with no image yet gets a video made from your description alone. Each video is added to the story's Videos, so you can browse its frames and use them as images.

1. Select the panel or image on the canvas, and open the Image group in the Inspector (click its header if it is closed).
2. Click Animate (next to Restyle).
3. Choose an engine: LTX-2.5 runs on your ComfyUI, and the cloud video models (such as Veo 3.1) run through your OpenRouter connection.
4. Describe what happens in the video, and choose its Length (3, 5 or 8 seconds). Each panel keeps its own description, so it's still there when you reopen the story or open it on another device.
5. Optionally, with LTX-2.5 and an image, turn on Enhance prompt: a small model rewrites your prompt from it and the image before the video starts. After the video is made, open Prompt used (last video) under the prompt to read what it was made from; the job in the Jobs window shows the same prompt.
6. Click Animate. The note beside the button shows how far along the video is, and Cancel stops it. Videos take a minute or more, and you can keep working while they render.
7. When it's done, the note says which video was added, and the video plays at the top of the Image group, in place of the image. It also appears in the panel's History, one step after the image it was made from.

Whenever you select a panel or image you animated, its newest video plays at the top of the Image group, on a loop with the sound off (turn the sound on in the player's controls). Click the photo button at the top right to see the image again, and the play button in the same place to go back to the video. The branch button beside the photo button shows History: click a video there to play it. When the panel has more than one video, the film button lists them: choose one to play it. To export the video, drag the share button at the top left onto the desktop, a Finder window or another app, or click it to save a copy where you choose. Starting a new image for the panel shows the image again, so you see the new one arrive. All of the story's videos are also in the canvas sidebar's Media mode: open the menu at the top of the sidebar, click Videos and choose the video (below the chapters, under Imported Videos, when the story has generated takes) to browse its frames.

If your render machine stops answering while an LTX-2.5 video renders (the network drops, or it's asleep), the note says Waiting for the render machine to answer… and StoryForge keeps asking, then collects the video once it answers. When a video still doesn't make it to Videos (it failed while the render machine kept working, or you quit StoryForge before it was done), the panel's Animate mode says a take may have finished on the render machine. Click Bring In Finished Takes: StoryForge asks the render machine for each one and adds the finished videos to Videos. The note under the button says what it found; a take still rendering stays there to bring in later.

To look for a panel's finished takes on your render machines even when StoryForge kept no record of them (another Mac made them, or they're from an earlier version), click Find Finished Takes in the panel's Animate mode. StoryForge reads each render machine's recent history and adds the panel's finished videos to Videos; a video already in Videos isn't added twice. Videos made before StoryForge named each file after its panel don't say which panel they're of: they're listed under the button with their prompt and when they finished, so click Bring In next to the ones of this panel. A render machine forgets its history when it restarts, so look before you restart it.