The state of AI image and video generation in 2026

We catalog AI image generators and AI video generators for a living, which puts a number in front of us that most coverage of this market never reaches: how many of these tools actually exist, what they cost, and how many have already died. As of June 2026 we track 258 AI image and video generation tools — 169 on the image side, 114 on the video side, and 25 that do both. This report is what that dataset says about the state of AI image and video generation in 2026, by scale, by price, and by survival.

The frontier-model headlines move fast — a new image model one week, a longer-clip video model the next — but the shape of the market underneath them changes more slowly, and it is measurable. We are not neutral observers here; we maintain this catalog by hand, and we will be specific about what it shows and where its edges are. For the products behind the numbers, our buyer's guides to the best AI image generators and the best AI video generators are the running picks this report sits on top of.

Key findings

• We track 258 AI image and video generation tools as of June 2026: 169 tagged image, 114 tagged video, and 25 that do both. The image side is still roughly half again as large as the video side.
• Only about 1 in 10 is genuinely free. Of 238 active tools, just 24 (10%) are free to use without limits. About half (49%) require payment for any real use; the other half offer a free tier you will most likely outgrow.
• Image and video have converged on the same pricing shape. Both sit near 41% freemium and 42% paid — video has not ended up meaningfully pricier by pricing model, even though it burns far more compute per generation.
• 20 of the 258 are already dead or acquired, and image generation is the harder-hit side — 18 of those 20. The most prominent casualty is OpenAI's Sora, shut down in March 2026 about six months after launch. AI image creation is one of the most crowded wings of our AI graveyard.
• The buyers are incumbents, not AI-native startups. The acquirers in this set are Adobe, Google, Autodesk, Freepik, and Jasper — established platforms bolting generation onto an existing product.

How we built this report

The dataset is every tool in our catalog tagged AI Art & Image Creation or Video Creation — 258 products after de-duplicating the 25 that carry both tags. For each one we read its pricing model and its lifecycle status (active, deceased, or acquired), plus, for the tools that ended, a cause and an acquirer where we could establish them. The snapshot was taken in June 2026 and is reproducible from the same script that generated these numbers.

Four caveats, because the method shapes the figures:

1. This is our catalog, not a census of every AI image or video tool that has ever existed. It is a large, hand-maintained sample skewed toward products notable enough to list. Read it as a structured cross-section, not the whole population.
2. "Pricing model" describes how a vendor gates access, not the dollar amount. Freemium means a permanent free tier with paid upgrades; Free means usable without paying; Free Trial means time-limited with no lasting free tier; Paid means no free path at all. We record the model, not the monthly price, because list prices on these tools change almost monthly.
3. We deliberately leave editorial ratings out of this report. Our star scores are curated, not survey data, so presenting them as a market-quality metric would overstate them. The counts, pricing models, and lifecycle marks in this report are catalog facts; the ratings are not, so they stay out.
4. The churn figures are floors, not finals. Quiet deaths surface on a lag — we often confirm one months later — so the dead-or-acquired count for recent years rises over time. Twenty is the number we can stand behind today, not a ceiling.

The field is enormous — and still lopsided toward image

The first thing the data settles is size. Two hundred fifty-eight tools is a lot of products doing two tasks, and it reflects how low the barrier to entry has stayed: a capable team can wrap a frontier image or video model and ship a usable product in weeks. The 169-to-114 split between image and video is the second thing worth noting. Image generation has had a two-year head start — the post-2022 diffusion wave — and the catalog still carries half again as many image tools as video tools.

That gap is closing, though, and faster on the video side. Video is where the frontier moved in 2025 and 2026, and the count is catching up as longer-clip, higher-fidelity models pull a new wave of products into the category. If the image side is the mature, crowded market, the video side is the one still adding entrants at speed.

What AI image and video generation actually costs in 2026

Here is the finding most buyers get wrong. The popular impression is that AI image and video generation is essentially free — type a prompt, get a result. The catalog says otherwise. Of 238 active tools, only 24 (10%) are free to use without limits. The genuinely free options are concentrated in a handful of model-backed names — DALL·E through ChatGPT, Stable Diffusion, ComfyUI — not the broad field.

The broad field is split almost evenly between two gates. About 41% are freemium (a real free tier, paid upgrades for resolution, speed, commercial rights, or watermark removal) and about 42% are paid outright, with another 7% offering only a time-limited trial. Put the paid and trial-only tools together and 49% of active AI image and video tools require payment for any meaningful use. The other half let you start free — and then meter exactly the thing you came for.

The split between image and video is the surprise: there essentially isn't one. Image tools run about 42% freemium and 40% paid; video tools about 47% freemium and 40% paid. Video generation costs a vendor far more per output — seconds of generated video are orders of magnitude more expensive to compute than a still — yet the headline pricing model has converged with image. What differs is underneath the label: a video tool's "free tier" is usually a few watermarked seconds, while an image tool's free tier might be dozens of full images. The gate is in the same place; the bars are spaced differently.

Image and video are merging into one creative stack

Twenty-five of the tools we track carry both tags — they generate stills and clips in the same product. That is the multimodal creative suite taking shape: Picsart, Freepik, Runway, Krea, Kling, Hailuo, and the rest building one workspace where an image becomes a video becomes an edit. Nearly all of them are freemium, which fits the strategy — get a creator in on free image generation, then upsell the compute-hungry video.

This is the structural trend behind the 2026 product roadmaps. The standalone single-trick generator is being squeezed from both sides: by the frontier labs whose models do more out of the box, and by these converging suites that fold image, video, and editing into one subscription. A tool that only makes square images is now competing with products that make images, animate them, and cut them together.

The other side of the boom: churn

A market this crowded clears out the bottom, and ours is doing it. Twenty of the 258 tools we track in this space — about 1 in 13 — are already dead or acquired, and the weight falls on the image side: 18 of the 20. That lines up exactly with our AI graveyard report, where AI image creation was one of the three deadliest categories overall. Easy to launch, brutally hard to defend once the base models absorb your one feature.

The most prominent ending is a shutdown, not an acquisition. OpenAI shut down Sora — its flagship text-to-video generator — on March 24, 2026, about six months after launch, citing the operational cost of running video generation at scale and fading usage. When the most-hyped video model of its moment can be switched off inside a year, durability is not just a small-vendor problem — and the reason it gives, compute cost, is the same economics that meters every free tier in the section above.

Below that headline, the endings split the familiar way. Some were quiet deaths — image tools like Meero, Booth.AI, and Stadio that simply let their domains lapse rather than post a shutdown notice. The rest were acquisitions, and the acquirers are the tell: Clipdrop went to Jasper, Magnific to Freepik, Galileo to Google, Wonder Dynamics to Autodesk, and Rephrase to Adobe. Almost none of the buyers are AI-native startups. They are incumbents — design platforms, stock-media companies, the model labs themselves — buying a capability or a team to fold into a product that already has customers. The independent generator gets absorbed; the standalone tool you were using gets sunset.

For anyone choosing a tool to build a workflow around, that is the durability question in one line: the most crowded, most-hyped corner of AI is also the one most actively being thinned and bought.

Who leads right now

By the tools our audience saves and returns to most — a directional signal, not a benchmark — the image side is led by Midjourney, Recraft, Leonardo AI, FLUX, Ideogram, Krea, and the free-via-model options DALL·E and Stable Diffusion. The video side clusters around Descript, HeyGen, Synthesia, Runway, Kling, and Captions, splitting into two jobs: cinematic generation (Runway, Kling) and avatar or talking-head production (HeyGen, Synthesia).

The open-versus-closed divide is worth watching. Open-weight models — Stable Diffusion, FLUX, and the ComfyUI ecosystem around them — anchor most of the genuinely free options and a large share of the tools built on top. The closed frontier models drive the paid and freemium tiers. Much of the 258-tool field is a layer of interface, workflow, and niche-tuning built on a much smaller set of underlying models. For the full picks and how they compare, the buyer's guides to AI image generators and AI video generators go tool by tool.

What this means if you're choosing an AI image or video tool in 2026

The dataset turns into three practical rules for buyers.

Budget for paid — free is the exception. With only 10% of tools genuinely free and half requiring payment for real use, assume the tool that actually fits your workflow has a bill attached. The free tiers are real, but they are designed to meter the exact capability you will need at volume — resolution, length, commercial rights, no watermark.

Weight the category's churn. Image generation specifically is the most thinned-out wing of this market. A new entrant can still be excellent, but its category's track record says scrutinize durability hardest — can the team survive the base models absorbing its feature, and does it have a moat beyond the model it wraps? Our editorial review methodology leans on exactly these signals.

Prefer the converging suites for anything long-term. If you are building a repeatable workflow rather than making one image, the multimodal suites and the model-backed leaders are the safer durability bet than a single-trick generator — both because they are harder to displace and because they are the ones doing the acquiring, not the ones getting bought.

Frequently asked questions

How many AI image and video generation tools are there in 2026? In our catalog, 258 as of June 2026 — 169 tagged image generation, 114 tagged video generation, and 25 that do both. That is the set we track closely enough to maintain pricing and lifecycle on; it is a large hand-maintained sample, not a census of every tool that exists, and it grows every month.

Are AI image and video generators free? Mostly not. Of 238 active tools, only about 10% are genuinely free to use without limits. Roughly 41% are freemium (a free tier with paid upgrades) and about 49% require payment for any real use once you include trial-only tools. The genuinely free options cluster around model-backed names like DALL·E, Stable Diffusion, and ComfyUI.

Is AI video generation more expensive than image generation? By pricing model, barely — both run about 41% freemium and 42% paid. Video costs vendors far more to compute per output, but that shows up in tighter free tiers (a few watermarked seconds) rather than a higher share of paid-only tools. The gate is in the same place; the free allowance is smaller.

Which AI image and video tools have shut down? In our data, 20 of the 258 are dead or acquired, 18 of them on the image side. The most prominent is OpenAI's Sora, shut down in March 2026. Others were quiet domain-expiry deaths (Meero, Booth.AI, Stadio) or acquisitions — Clipdrop to Jasper, Magnific to Freepik, Galileo to Google, Wonder Dynamics to Autodesk, Rephrase to Adobe. AI image creation is one of the deadliest categories in our AI graveyard.

Did OpenAI's Sora shut down? Yes. OpenAI shut down Sora, its flagship text-to-video and image-to-video generator, on March 24, 2026 — roughly six months after launch — citing the operational cost of running video generation at scale, declining usage, and a shift in strategy. It is one of the most prominent entries in our AI graveyard, and a reminder that frontier branding does not guarantee a tool's survival.

Who is acquiring AI image and video startups? Incumbents, not other startups. The acquirers in our set are Adobe, Google, Autodesk, Freepik, and Jasper — design platforms, stock-media companies, and model labs buying a capability or team to fold into a product that already has customers, rather than AI-native companies.

Are AI image and video tools merging? Yes. Twenty-five of the tools we track generate both stills and clips in one product — the multimodal creative suite, led by names like Picsart, Freepik, Runway, and Krea. The standalone single-format generator is increasingly competing against workspaces that do image, video, and editing on one subscription.

Where to go next

For the products behind these numbers, our buyer's guides to the best AI image generators and the best AI video generators go tool by tool. For the churn side of the story, the AI graveyard report covers what kills AI tools across every category, and the AI graveyard is the running record of the ones that ended. For the spending that funds — and eventually starves — this whole field, our AI capex analysis follows the money.

The honest version: a 258-tool field where only 10% is free and 1 in 13 has already ended is not a market in trouble. It is a market maturing in fast-forward — a crowded, low-barrier middle being thinned and bought while value consolidates into the frontier models and the suites built around them. The job of a buyer in 2026 is to tell which side of that line a tool is on before you build a workflow on it.

— The ToolDirectory.AI editorial team