Make unlimited AI images and videos for one flat monthly price.

We support over 30,000 compatible models and workflows from https://modeltrends.ai: a model that shows a ShortCode on its ModelTrends page runs here. You can also upload your own LoRAs from sites like Civitai and Hugging Face. With the exception of a few edge cases, most safetensors files below 25GB are compatible.

There are also many preloaded models: At the time of this update, Grok 4.5 is the default LLM, KREA 2 is the default image model and MiniMax H3 is the default video model. Also preloaded are Wan 2.2, Hunyuan, Flux 2, Anima, Ernie, SDXL, and others.

Graydient is a cloud GPU platform unlike any other.

For non-technical users, Graydient is already the engine behind PirateDiffusion (Telegram), the Discord bot Vector, the pro web studio SpyGlass, the easy chat creator Prism, PollyGPT (chatbots on the web and in Telegram), the Unified Editor, Ideogram4 Composer and MasterSelects. Pick the app that fits your device and your plan powers all of them.

For developers, the API runs the same models and about 400 ComfyAPI workflows as plain API calls: your code sends the prompt and the workflow name, Graydient's GPUs run the graph and return the result. There is no node editor to host. Search for example use cases of the Graydient API on GitHub to get started.

We no longer offer LoRA training, as consistent character workflows provide better results without the need for training. You can use multi-reference video, audio, and image workflows with Qwen and Minimax H3 to create consistent characters easily on our apps like Prism (web) or Vector (Discord) and even over Telegram, check the app's guides for more info.

Most AI platforms sell you tokens, credits, or points that don't map to anything real. They turn your dollar into 1000 magic points to make it feel like you have a ton of currency, only to hide the overcharging in hard to reconcile points usage. The ones advertising "unlimited" bury the limits in the terms then park you in the slow lane permanently in order to then sell you "boosts". We think that's a lousy way to treat customers.

So in 2023 we started calling data centers with spare capacity and built the alternative: the first genuinely unlimited image plan, launched as a fan forum. A year in, it became Graydient. Today it covers unlimited video too.

We think it's the best value in generative AI. No credits. No boosts. No fine print.

Not at this time, but we do offer a no-questions asked 7-day money back guarantee, so feel free to give it a shot and we'll gladly send you a refund if it's not a good fit.

On the Images and Plus plan, you can batch up to 8 images at a time. When you submit a prompt, the system lends you a full GPU for up to 3 minutes at a time, so it's more about "whatever you can get away with" than us limiting your abilities when workflows are involved (meaning there are various processing stops along the way). In a turbo workflow, you can create very high resolution images in a batch. On fidelity focused workflows like "krea 2 detailed", it's 1-2 at a time but you're getting an (upscaled) 4K image back.

Video is challenging because it requires more time. This is why we recommend the Plus plan for short videos and the Max plan for longer videos.

Graydient Max gives you up to six minutes of GPU time, and batching up to 16 images at a time, and creating 4K images at native resolution without upscaling.

It comes down to GPU run time. Plus lets you process a video at 3 minutes at a time, and Max at 6 minutes at a time.

480p works best on Plus. 1080 is still possible on Plus with LTX 2.5 at low lengths, but the videos will be shorter. If you want longer cohesive videos, we recommend Max.

The main technical difference between plus and max is GPU runtime. Max allows the workflow to run at more than twice the time, allowing up to 1440p and 480 videos are almost a minute long in one shot.

Yes, everything runs on our website or apps you already have, like Discord and Telegram, or right in your web browser without any installs required. Start with the easy chat creator (Prism), the simplest way in for beginners.

Images, Upscaled photos, and other workflows complete within 30 seconds. Videos in sixty seconds. Convert a photograph into a moving video, it only takes about 150 seconds. No workflow on Graydient runs longer than 2 minutes, that's our target. It's fun and fast.

Our mission is to provide you with a service that is faster than a top-of-the-line $10,000 computer, leased to you at a tiny fraction of the cost.

Our hardware is extraordinary: We use the latest AMD and Intel Chips for full PCI lane efficiency, 250GB+ of RAM, large solid state NVME drives.

Unlike our competitors, we aren't leasing random people's basement computers while they're playing Counterstrike or mining Bitcoins. All of our hardware is strictly dedicated to A.I.

Graydient owns or leases A100s, A6000s, RTX 5090s, and upgrades regularly. We do not partition the GPUs, give you the full might of each server. We also support DMD 2 (Distribution Matching Distillation 2), various 4-step turbo loras, and LCM which can deliver 1024x1024 SDXL images under 8 seconds.

Our service is fast, we guarantee it.

Yes, your content is protected with 2FA password logins and is not published or shared in a feed unless you create a public "space" or share on groups. For the safety of the community, there are random anonymous checks for illegal activities.

None at all. We're a small family company, and our moms raised us right. You won't find any fine print. Our service is truly unlimited, and if you fill your drive you can simply delete images or change plans. We don't nickel and dime our customers. You can easily download your images, too.

We stand by our products and truth in our advertising. If you're not satisfied for whatever reason, we'll promptly refund you asap. Refunds typically take 3-4 business days within the US and longer overseas (we're at the mercy of banks).

It's not applicable, because Graydient is the most convenient way to access Open Source A.I. models, optimized for API developers, your web browser and Telegram. We are the software part of the ecosystem, not the models part.

Graydient does not create models or scrape the web to train images. Additionally, any photos that enter the Graydient platform are not sold or trained for other uses. We take our customer's privacy very seriously.

Persons who do not want to appear in A.I. models should opt out with Stability AI, the creators of the Stable Diffusion model (or) similar creators.

Yes, but understand the limits: It doesn't include video or API access, and high resolutions may not be possible. Newer AI models may not be compatible or time out. If you're primarily creating with SDXL or older models, you can still buy Images Basic here: Monthly Images Basic Only - $15.99 or Yearly Offer, 2 months free - $159.99.

Attention previous/returning customers: If you had a legacy plan that's no longer sold, you can still resume it, but it may not support new features. Manage your account

If your old billing agreement has expired, it may say 'Customer not found'. To fix this, sign up as a new customer with the same email (without logging in) and you'll be able to resume your old projects.

Card didn't go through? Tips:

• Make sure you are in the 'full' browser and not the Telegram browser. We recommend Google Chrome or similar. (Please update your browser to the latest version)
• Write your full name & postal code exactly as they appear on your credit card statement, or if it's debit try a credit type
• See if your app has 3 dots on the corner, choose "use browser"
• Update the clock/timezone on your device to automatically set
• Try disabling adblock & browser extensions
• Try a different web browser in private mode
• Sync the world clock time on your computer
• Disconnect from VPN and load the page again

Have a question? Email us

Graydient co-founders Yanier and Thomas built websites like Destructoid, Upcomer, and Classic dot com. We've worked as professional web developers for over 20 years. This is not our first rodeo, and take great pride and care of our community. We built Graydient with our own money, because we couldn't quite find anything like it. We hope to grow this into something wonderful for the Generative AI community at large. We also employ really lovely moderators, podcasters, and sponsor model makers occasionally.

No, it is a personal API key to generate as a single user. It is suitable for app development, but not for reselling. It is rate-limited and will auto expire if too many requests are sent. Please limit your frequency to 10 seconds per call. Batching is discouraged.

To resell our API or to batch renders all day long, please rent a dedicated VIDEO MAX GPU server, offered at $499.99. This means you have your own dedicated machine (no queue waiting) and can run longer jobs or limit what your customers can do and how long their queue slots run. Renders are $0.02 each including video. You can stack multiple MAX GPUs to increase your total capacity as you grow. Contract terms are easy: 30 days notice, no long term complexity. Please contact us to rent a dedicated GPU.