Diffus is a browser-based AI image generator with 70,000+ models, ControlNet and ComfyUI support. Plans start at $1 per month. See features and pricing.
Researched and fact-checked by Bartu Cavusoglu · Updated October 26, 2026
Diffus, also written GRAVITI Diffus, is a browser-based platform for generating AI images with Stable Diffusion and related models. Instead of installing software or buying a powerful graphics card, users open a web page, pick a model, write a prompt and generate. The site describes access to more than 70,000 user-contributed models, including checkpoints, LoRAs and embeddings.
Beyond image generation, Diffus lists tools for image editing and video generation, plus a set of free creator tools built around Flux 1, Flux Kontext and Nano Banana. The company lists a New York City address and offers support through Discord and a ticket system. Its target audience is professionals and creators who want fine control over their output.
The homepage describes a three-step workflow. First, the user chooses a model from thousands of specialized options to match the look they want. Second, the user writes a prompt, which the site calls an intelligent prompt system. Third, the user refines faces, composition and details with professional tools before finishing the image. All of this runs in the browser.
Generation runs on Diffus infrastructure and uses credits tied to the subscription plan. Higher plans raise limits such as maximum sampling steps, maximum inference time and the number of concurrent jobs. Plus and Pro also include free boosting time, which runs tasks on a dedicated GPU, and let users upload and use private models.
Control tools are the main focus. ControlNet supports up to four units on Basic and above, and one on Teaser. ADetailer detects faces automatically and regenerates them with more detail. Couple lets users write prompts for different regions of an image. Inpainting is listed on the homepage, and Segment Anything creates masks from prompts or clicks on Plus and Pro.
Other capabilities include Layer Diffusion for transparent backgrounds, face swapping for images and videos, and ComfyUI for custom workflows. Video generation covers text-to-video and image-to-video with Kling, Veo and Wan. Pro adds advanced video features with sound, finetuning on a user's own data, and a higher cap of 160 sampling steps per task.
Diffus, also called GRAVITI Diffus, is a browser-based AI image generation platform built around Stable Diffusion and related models. It offers access to 70,000+ user-contributed models and tools such as ControlNet, inpainting and ADetailer. It also lists video generation and image editing tools.
Diffus lists four plans: Teaser at $1 per month, Basic at $12, Plus at $24 and Pro at $48. Each of the last three shows a higher struck-through price ($18, $36 and $72). The page offers monthly, quarterly, biannual and yearly billing, with yearly labeled Save 40%.
The pricing page includes a question about a free trial, but the answer was not available in the reviewed sources. The site does offer free creator tools for Flux 1, Flux Kontext and Nano Banana. Check the site for current trial terms before subscribing.
Diffus suits freelancers, designers, developers and creators who want Stable Diffusion models and detailed control without running their own hardware. Users who train or upload their own models should consider Plus or Pro.
Yes. ComfyUI is available as an addon on Basic and Plus and is included on Pro. It is not included on the Teaser plan.
Yes, on Plus and Pro. The pricing page lists text-to-video and image-to-video with Kling, Veo and Wan, and Pro adds advanced video features with sound and extras.
Plus and Pro include uploading and using private models, with 25GB and 50GB of private model storage. Pro also includes finetuning, which trains models on your own data. All plans can upload public models.
Support depends on the plan: none on Teaser, limited Discord support on Basic, Discord support on Plus and priority Discord support on Pro. The site also lists a ticket system and a contact email.