AI RPGs With Images (2026): Which Ones Actually Show You Anything, and What a Picture Costs

In an AI RPG the words are usually included and the picture is the thing you pay for. That is the single most useful fact about the visual layer in this category, and no pricing page states it plainly.

It is easy to miss because “has images” reads like a yes-or-no feature. It is not. A platform that hands you one character portrait at creation and a platform that paints every turn of a year-long campaign both tick the same box, and they are not remotely the same product. One of them is decoration you will forget by the second session. The other is the reason you come back.

So this is not a list of platforms with images. It is four grades of visual layer, what each one actually costs, and the test that tells you which grade you are looking at inside the first ten minutes.

Arcanum is independent and unpaid. Several platforms named here gave us access we did not pay for: aiga_, Craft, Fable Forge, MythEngyn, NOPOTIONS, Tabled, Tidefall, AI Game Master, and Voyage through its parent company. None had any say in what follows. All product names are used for identification only; Arcanum is not affiliated with any platform named here.

The Four Kinds of Visual Layer

Grade the thing in front of you before you judge it. Almost every argument about whether an AI RPG “has good images” is two people describing different grades.

Grade 0 — nothing. The game is text and the pictures are in your head. Taverna lives inside a Discord channel, so even tactical positioning is abstracted into prose. Infinity DM leans on a clean interface and writing rather than generated battlemaps. This is not a failure state. It is the oldest form the genre has, and it is the form interactive fiction was built in.

Grade 1 — a face in a menu. You make a character, the platform generates a portrait, and that is the visual layer. AI Realm does this at creation. NOPOTIONS folds a portrait into character creation alongside race, class, attributes and bonds. MythEngyn is the purest case and the most deliberate: there is no character artwork while you talk, and you open a menu to see a face. Fable Forge sits just above the line, because its characters use looping idle animations rather than a static portrait — a small thing that changes the feeling of a conversation more than a sharper image would.

Grade 2 — scene art on demand. The game generates a picture of what is happening, usually per scene, usually metered. AI Dungeon is the reference implementation. FableAI pairs generated portraits and scene art with narration and a soundtrack. AI Game Master narrates with generated scene art alongside the prose. MacerAI generates maps and visual assets to take that job off a human game master. This is the grade most people mean by “AI RPG with images”, and it is also the grade where the money lives.

Grade 3 — an integrated visual system. The picture is not an illustration of the game, it is part of it. Friends & Fables runs hex battlemaps with line-of-sight, which means the image is game state you can read. Craft stages a scene the way a visual novel does — character sprites above the text box, expressions shifting as people speak, actual cinematics. Tabled paints every single turn, with your characters in it, so the campaign accumulates into an illustrated chronicle. Tidefall builds a place out of thousands of hand-drawn pieces and animated backgrounds and puts you inside it.

Deep Realms is worth naming as the case that shows why the grades matter: it generates images and it is still, by its own design, Grade 0 in spirit. No hex maps, no consistent portraits, no scene visualisation — it is a text engine with image generation bolted to the side, and it would rather give you a Python console.

The Picture Is What You Are Actually Buying

Here is the part that survives every redesign of every pricing page: when an AI RPG meters something, the picture is very often the thing it meters.

The clearest statement of it comes from the biggest platform in the category. AI Dungeon’s help centre says what credits are for, and the list is short: “Credits can be spent in AI Dungeon Adventures creating new, one-of-a-kind, AI-generated images to compliment stories, or to get more temporary context on certain models.” Two things. Pictures, and how much of your own campaign the model can still see. Not turns. Not play. Its pricing page leads with the same framing — the three benefits named at the top are the strongest text models, original AI images, and larger context. Your tier then decides the exchange rate: “Your membership tier determines how many Credits each image costs across different models,” and on higher tiers some generators stop costing credits at all, “as the price to run them is offset by the amount you pay every month.” The ladder from free through $14.99, $29.99, $49.99 and $99.99 a month is, among other things, a ladder of how cheap your pictures get. All verified on AI Dungeon’s own help centre and pricing page, 13 September 2026.

aiga_ makes the same structure explicit enough to be a design statement. Text play is included with every paid plan — creating worlds, starting games, taking story turns, sending chat messages, none of it spends anything. Credits exist for generated images, video clips and other media. Its cheapest paid tier, at $8.99 a month, buys unmetered text play and nothing else; if you never generate an image, that tier is the entire product. The tiers above it are priced in media: roughly thirty images at $11.99, roughly a hundred and forty at $23.99, roughly five hundred at $59.99. Its free tier, meanwhile, is five text-only turns a day. Verified on aiga.io, 8 September 2026.

Voyage draws the line in the same place from the opposite direction. Core play is unlimited and never touches your usage budget; the budget covers the premium extras, and custom image generation is on that list alongside voice narration and the premium engines. At the top of its Shadow plans the image generators themselves become the unit of value — three unlimited on the cheapest, six on the next, ten on the most expensive.

The pattern repeats wherever you look for it. AI Realm advertises “29 Storytelling AI Models” and “3 Custom Image Generators” side by side on its homepage, as though a generator count were a spec like a model roster — verified 13 September 2026. When we last checked their pricing, Fable Forge granted portrait generations per day by tier and MythEngyn granted them per month, both rising with the price; FableAI’s free tier limits image generation specifically, and lets you earn more by watching ads.

Which gives you a genuinely useful move when you are comparing two subscriptions: find the number that goes down when you generate a picture, and ask what else spends it. If pictures and memory come out of the same pot — as they do on AI Dungeon — then every image you make is a small decision about how much of your campaign the model gets to remember. That is not an argument against generating images. It is an argument for knowing that you are making the trade. We take the rest of the metering question apart in what an AI RPG actually costs.

The Two Best-Looking Games in the Category Generate the Least

This is the finding that surprised us, and it has held up across a year of playing these things.

The two best-looking products in our directory are Tidefall and Craft, and neither of them is winning on image generation. Tidefall’s art is drawn. Thousands of pieces of it, attached to specific people and specific places, with animated scene backgrounds and a scored soundtrack under them — made first, and performed by the model second. Craft’s is staged: sprites, expressions that move with the dialogue, cinematics. They are not trying to look like the same thing and neither is the other’s runner-up. What they share is that the pictures existed before your session did.

Set that against the Grade 2 pattern, where a picture is conjured per scene and has never seen the rest of your story. Generated scene art is almost always the most impressive thing in the first ten minutes and the least important thing by the tenth hour, because it has no memory, no continuity and no consequences. It is a mood. And moods are exactly what this category is already oversupplied with — our whole design critique of the field is that it is generous with atmosphere and stingy with things that persist.

There is one product that gets generated images to do real work, and it is instructive that it did so by changing what the picture is for. Tabled paints every turn, with your characters in it, in a consistent hand, as the scene resolves. Because the game is asynchronous — a thoughtful move a day, or five rounds in a lively evening — the picture arrives as the reward for the round rather than as wallpaper behind it. And because every turn is painted, the campaign ends with something that exists: an illustrated chronicle the table keeps. That is not decoration. It is an artefact, and it is the only case in the directory where we would tell you the image generation is a reason to choose the platform.

What a Generated Picture Doesn’t Know

Here is the test, and you can run it in your first session.

Does the picture know what happened?

Three specific things to watch for, in the order they usually break:

Does your character stay the same person? Generate a portrait, play for twenty minutes, generate another. If the face has changed, the platform is producing pictures of a description, not of your character. This is harder to fix than it sounds, which is why aiga_ treats it as a headline feature rather than a default — it advertises image continuity that holds your hero’s face across scenes and across twelve different art styles. You only advertise that when it is difficult.

Does the picture agree with the text? In our Craft review this is the seam we could not un-see: the tactical map layer gives you troops, terrain and markers, and the narrator misplaces or ignores them. When the image and the prose describe different situations, the image is not a view of the game — it is a second, parallel game running beside it. Craft’s team names context management as its own top open problem, which is the honest version of the same observation.

Does anything in the picture persist? A scene image that is generated, shown and discarded is a screensaver. Ask whether it is attached to a place you can return to, a character who will appear again, or a log you keep.

Failing these is not automatically a strike against a platform, as long as you know which product you bought. MythEngyn fails the first two by refusing to play — no artwork while you talk, a face in a menu if you want one — and it is transparently spending the effort somewhere else, on a world that holds together fifty sessions deep. That is a legitimate trade. The platforms to be careful with are the ones that sell you the picture and then quietly let it drift.

The Other Direction: Putting a Picture In

There is a second, much less served intent hiding inside every search for AI roleplay with images, and it runs the other way: people who want to hand the game a picture rather than receive one. A photo of a face they already have in mind. A sketch. A piece of art they commissioned for a character they have played for years.

Very little in this category does it. aiga_ is the exception worth knowing: you can upload a photo or a sketch and it will carry that face through its art styles. On the do-it-yourself side, SillyTavern’s image-generation extension connects to a Stable Diffusion backend, local or hosted, which means the whole pipeline — including what it starts from — is yours to configure. Everywhere else, the answer today is that you write the character’s appearance into your character card and hope.

If this is what you came for, that is the current state of it, and it is a real gap in the category rather than something you are failing to find.

Playing With Pictures for Nothing

The free routes are narrower than the search results suggest, because free tiers are usually where the image budget gets cut first — aiga_’s free tier is text-only, FableAI’s limits generation specifically, and AI Dungeon’s free tier has no credits at all.

Three things do work without paying:

• Perchance is free, needs no account, and its AI character chat sits next to a Stable Diffusion-based image generator that some setups tie into. It is the most common answer on the internet to this exact question, and it is a fair one — provided you know you are getting a free browser toy rather than much of a game.
• A general model you already pay for. ChatGPT’s Custom GPTs can generate images mid-session — a location, an item, the face of somebody you just met — without leaving the game, which is the smoothest visual layer any general model offers. How to build that is in our ChatGPT RPG guide.
• SillyTavern costs nothing and generates images through a backend you supply. The trade is that you are assembling a cockpit rather than opening a game; our SillyTavern guide walks the setup.

One character portrait is also free almost everywhere. AI Realm hands you a custom AI-generated image of your character at the end of creation, on the free tier, before you have spent anything. If all you want is to see your character once, you rarely need to pay for it.

What to Pick, by What You Want the Picture For

Hubs should route rather than re-rank, so this is a routing table, not a leaderboard. For the ranked view of the platforms themselves, see best AI roleplay platforms and best AI RPG by priority.

• You want to see your character, once. Any Grade 1 platform, free. AI Realm or NOPOTIONS at creation.
• You want something to keep at the end. Tabled. The illustrated chronicle is the only visual output in the directory that outlives the campaign.
• You want to see the battlefield. Friends & Fables, whose hex battlemaps are readable game state rather than an illustration of it.
• You want to be inside a place. Tidefall, for hand-drawn art and animated backgrounds, or Craft, for visual-novel staging. Different bets, both gorgeous.
• You want a picture of every scene. AI Dungeon, aiga_ or FableAI — and read the metering first, because this is the want that costs money.
• You want to spend nothing. Perchance, SillyTavern, or a Custom GPT on a subscription you already have.
• You want the pictures to be consistent. aiga_ is the only platform we have found that treats character continuity across images as a feature rather than an accident.

One last thing worth saying plainly, because it cuts against how this category markets itself. The best-made mobile AI RPG we have played is not close to the best-rated one, and the gap is entirely production values — portraits, cinematic narration, soundtracks. None of that is among the seven things our methodology scores, and that is deliberate. A beautiful game that forgets your campaign at turn fifty is a beautiful game that forgets your campaign at turn fifty. Pictures are worth wanting. They are a bad proxy for whether the game underneath them works, and the first twenty minutes is where you should be testing the other thing.

How We Checked This

Every structural claim here comes from our own play of the platforms in the client directory, from each platform’s own documentation, or from published first-hand findings in our reviews, which are linked where used.

AI Dungeon’s credit and tier facts were verified on its help centre and pricing page on 13 September 2026, and are quoted rather than paraphrased. AI Realm’s generator count and character-creation portrait were verified on its homepage the same day. aiga_’s pricing structure was verified on aiga.io on 8 September 2026 and Voyage’s on 12 September 2026.

One limit worth stating. Where a price or an allowance was last verified more than a few weeks ago we have described the shape of the metering and said when we checked, rather than quoting a figure that may have moved — pricing in this category changes faster than anything else about it. AI Realm has shipped a version 2 since our directory entry was first written; that entry’s pricing table was re-checked against AI Realm’s pricing page on 14 September 2026.

No score in this article, and none implied. The visual layer is not one of the seven axes we measure, for the reasons above.

Frequently Asked Questions

Which AI RPG has the best images?

It depends entirely on what you want the picture to do. For hand-drawn art and animated backgrounds that put you inside a place, Tidefall. For visual-novel staging with sprites and expressions that move with the dialogue, Craft. For readable tactical maps, Friends & Fables. For a generated picture of every single turn that you keep afterwards as an illustrated chronicle, Tabled. For sheer volume of generated scene art on demand, AI Dungeon. These are five different products, not five places on one ranking.

Do AI RPGs charge extra for images?

Usually, yes — and in this category it is more often the picture that is metered than the play. AI Dungeon’s credits are spent on generated images and on extra context, not on turns. aiga_ includes all text play on every paid plan and spends credits only on images, video and other media. Voyage keeps core play unlimited and puts custom image generation in the premium budget. When you compare two subscriptions, find the number that goes down when you generate a picture and check what else spends it.

Is there a free AI RPG with images?

Yes, but fewer than the search results imply, because free tiers are usually where the image budget is cut first. Perchance is free, needs no account, and pairs its character chat with a Stable Diffusion-based generator. SillyTavern is free and generates images through a backend you supply. A Custom GPT on a ChatGPT subscription you already pay for can generate images mid-session. And a single character portrait at creation is free on several platforms, including AI Realm.

Can I upload my own picture to an AI RPG?

Rarely. aiga_ is the clearest exception — it will take a photo or a sketch and carry that face across scenes and across its art styles. SillyTavern gives you full control of the image pipeline because you supply the backend. On most platforms the answer is that you describe your character in the card or the creation flow and accept what comes back.

Why do generated character portraits keep changing?

Because most platforms generate a picture from a description each time rather than from your character, so the model is drawing a fresh interpretation rather than the same person. Holding a face steady across scenes is genuinely hard, which is why aiga_ treats image continuity as a named feature instead of an assumption. If consistency matters to you, test it before you subscribe: generate a portrait, play for twenty minutes, generate another, and compare.

Do images make an AI RPG better?

They make it more appealing in the first ten minutes and they have almost no bearing on whether it is still good in the tenth hour. Generated scene art has no memory, no continuity and no consequences — it is a mood. The exception is when the picture does a job: readable game state on a battlemap, or an artefact you keep at the end of a campaign. Our scoring deliberately excludes presentation, because a beautiful game that loses your campaign at turn fifty is still a game that loses your campaign at turn fifty.