Create · pictures · updated 22 October 2026

Image

Describe a picture, or attach one and say what to change. A review panel opens first, so you can pick the model, size and shape before anything is charged.

See the recipes Models for /image
  • 53image models
  • 1–32credits a picture at default settings
  • 20reference pictures at most (model dependent)
  • 4pictures per run at most

Entry I

Start here

Run /image with a prompt. Leave model empty for Gemini 2.5 Flash (5 credits), or type to search the list. Attach up to 10 pictures with image1 to image10; models that take more add the rest from the panel's Pictures button.

The panel shows only the settings the chosen model really has: resolution, aspect ratio, quality, background and how many pictures to make. Credits are taken only when you press Generate. The button shows the price for the settings you picked, and your free 25 credits a day are spent before paid ones. A run that fails is refunded.

Entry II

Writing the prompt

Write a scene, not a tag list

Image models follow full sentences better than strings of keywords. Say what is there, what it is doing, and where.

Most important first

Lead with the subject and its action, then the style, then the setting and light. Some models weigh early words more.

Name the medium and the light

"Watercolour on rough paper", "shot on a 35mm lens, soft window light", "flat vector icon". Specific beats "high quality" or "ultra detailed".

Put words in quotes

For a poster or a sign, write the exact text in quotes and say where it goes: a shop sign that reads "Crumb & Co" above the door. Spell unusual words carefully.

Say what you want, not what you do not

"Sharp focus throughout" works better than "no blur". Several models have no separate negative prompt.

Shape and size live in the panel

Pick the aspect ratio and resolution in the panel instead of writing "16:9" in the prompt; the panel only offers values the model accepts.

Say what it is for

An app icon, a book cover, a product photo for a shop. The purpose sets the level of polish and the layout.

Entry III

Editing a picture

Attach the picture and describe the change. Say what to change and what to keep: "Replace the sky with a sunset. Keep the people, their clothes and the framing exactly the same." Repeat what to keep on each new edit, or details drift.

With several pictures, refer to them by order: "Put the jacket from the second picture on the person in the first." Every image result also has a ✏️ Edit button that opens a fresh panel with that result as the reference; the panel then lists only models that accept a reference picture.

Entry IV

Choosing a model

The model list is marked: 🟢 recommended, 🟡 fine for a particular job, 🔴 not recommended (the wiki names what to use instead). The recommended ones today are Hy Image 3.5 Preview (uncensored), GPT Image 2.5 Flare, GPT Image 2.5 Sunburst, Nano Banana 2.1, Seedream 4.5 (uncensored), Qwen Image 3, MAI-Image-2.6, MAI-Image-2.6 Flash, FLUX.3 Image and FLUX 3 Image Klein. Seedream 4.5 (uncensored) is a cheap, reliable all-rounder at 5 credits.

A transparent background is offered only by models that support it: GPT Image 1 Mini, GPT Image 2.5 Flare, GPT Image 2.5 Sunburst, GPT Image 1, GPT-5 Image Mini, GPT-5 Image, Riverflow V2.5 Fast and Riverflow V2.5 Pro. Each model's references, sizes and price are in /wiki.

Entry V

Enhance

Enhance has a small chat model (Claude Haiku 5.5, GPT-6 Luna, MiMo v2.6 Flash or DeepSeek V4.1 Flash 1017) rewrite your prompt for the model you picked, for +2 credits. You see both versions and choose Using: Yours or Using: Enhanced, and you can edit the enhanced one before generating. It helps most with a short idea. When your prompt is already detailed, the panel may say so and Enhance is not worth it.

Entry VI

Every option

What you can type after /image, read from the command itself. You can run it again after 10 seconds.

OptionWhat it doesTakes
prompt needed Image prompt text
model Model (type to search; default: Gemini 2.5 Flash) texttype to search
hidden Only you can see this (default: false — post in channel) true / false
image1 … image10 Optional reference image (guides style / edit). Add more from the panel. file

Entry VII

Recipes

Requests you can copy into Discord and change to fit.

A product shot

Copy this

/image prompt: A ceramic coffee mug on a pale oak table, morning window light from the left, soft shadow, shallow depth of field, clean editorial product photo for a shop page

Why it works. Subject first, then light and lens, then what it is for. Pick 1:1 or 4:3 in the panel.

A poster with real words

Copy this

/image prompt: A retro travel poster of a lighthouse on a cliff at dusk, bold flat shapes, teal and orange, the title "NORTH POINT" in large white capitals across the top and "Summer Ferries" small at the bottom

Why it works. Quoted words with a place for each. Use a recommended model with good text rendering and a tall aspect ratio.

Change one thing in your photo

Copy this

/image prompt: Make it a snowy winter evening with warm light in the windows. Keep the house, the car and the camera angle exactly as they are. image1: (your photo)

Why it works. An edit says the change and the keep-list in two short sentences.

Combine two pictures

Copy this

/image prompt: Dress the person in the first picture in the coat from the second picture. Keep their face, pose and the background. image1: (person) image2: (coat)

Why it works. Refer to references by order, and say how they combine.

Entry VIII

Tips

  1. I

    Start with one picture at a cheap setting, then raise quality or resolution once the idea is right.

  2. II

    Short idea? Try Enhance. Long, specific prompt? Generate it as written.

  3. III

    Set hidden:true to keep the panel and the result to yourself.

  4. IV

    Change one thing per edit. Several small edits drift less than one big one.

Entry IX

Limits

  • Each model takes its own number of reference pictures, up to 20; the command says so straight away if you attach too many.

  • Up to 2 image runs at once, and 10 seconds between /image commands.

  • A run that takes longer than 3 minutes is stopped and refunded.

  • At most 3 generations a minute, 200 a day and 1000 a month per person, across all commands.

  • The Acceptable Use Policy applies to every model, including the ones marked uncensored.

Entry X

Questions

Why did the price change in the panel?

Quality, resolution, the number of pictures and, on some models, each reference picture change the price. The Generate button always shows the current total.

Why is a setting missing?

The panel shows only what the chosen model accepts. Switch model to get, for example, a transparent background or more pictures per run.

Can I get several versions at once?

Models that support it offer "How many images?" up to 4. Each picture is charged.

The edit changed things I wanted kept

List what to keep in the prompt, and edit one thing at a time. If it still drifts, try the same edit on another recommended model.

/videoVideo /story-bookStory Book /wikiWiki

Where the advice comes from

Prompting advice is from these free public guides, matched to what ALMA lets you set. Prices and limits are read from the bot itself, as of 22 October 2026.

Try it

Type /image in any server with ALMA, or in a direct message to the bot.

All guides

Something here wrong, or missing? Tell us.