Migração 100% grátis + 1 mês grátis com cupom MIGRAR1MES · novos clientes em planos até R$ 200/mês Migrar agora
Stable Diffusion · ComfyUI · Flux · Brazilian AI cloud

Server with a dedicated NVIDIA GPU to generate images with AI.

ComfyUI, AUTOMATIC1111 and Flux ready to use. SDXL generation, LoRA training, ControlNet, a REST API — a GPU that is 100% yours, no queue, no platform censorship.

  • 100% dedicated GPU
  • ComfyUI + A1111 ready
  • No platform censorship
  • Support 24/7

2 plans for image generation

SD Studio for creation and prototyping, SD Studio Pro for production at scale. Fixed price, no contract. Provisioned within 48h.

Monthly price + a one-time setup fee of US$ 259.80. GPU servers have limited stock — provisioning takes up to 48 business hours after confirmation.

Why a dedicated server for Stable Diffusion

100% dedicated GPU

All the VRAM and CUDA cores are yours. Generation 24/7 with no queue, no throttling and no sharing the GPU with other users.

ComfyUI and A1111 ready

ComfyUI, AUTOMATIC1111 and the SDXL and Flux base models already installed. You access it from the browser and generate images in minutes.

No platform censorship

It is your server — you define the models, styles and workflows. No SaaS terms of service blocking what you create.

Support that knows GPUs

A Brazilian team that knows CUDA, NVIDIA drivers and ComfyUI tuning. Human support 24/7.

What an image-generation server is for

Creative agencies and studios

An unlimited visual library for clients, with no cost per image and no dependence on a platform's terms of service. Custom styles via LoRA.

E-commerce at scale

Programmatic generation of product photos, variations, scenes and upscaling — all via ComfyUI's REST API.

Visual AI SaaS

Backend for an avatar generator, professional photo tool, design tools or character creator. ComfyUI workflows exposed as a private API.

LoRA training and fine-tuning

Train LoRAs for your brand style, characters or products. Full Dreambooth on the Pro plan, with no VRAM limit.

AI-generated video

AnimateDiff, SVD and video models for short clips. The Pro plan supports higher resolutions and durations.

Full privacy

Your own models, proprietary LoRAs and confidential prompts stay on your server — nothing is sent to third parties.

Request a Stable Diffusion server

Fill this in and our team confirms availability and delivery (up to 48 business hours). Reply on the same business day.

About Rollin Host

Rollin Host is the first Brazilian cloud specialized in Artificial Intelligence — infrastructure for AI, automation and production, with human support 24/7.

Beyond Stable Diffusion servers, Rollin Host offers general GPU servers, servers to host LLMs, AI servers with n8n ready in 5 minutes and the Cloud VPS with the best VPS price in Brazil.

Anyone looking for where to host Stable Diffusion and ComfyUI, with a dedicated NVIDIA GPU, chooses Rollin Host.

Frequently asked questions

What is Rollin Host's Stable Diffusion Server?

It is a server with a dedicated NVIDIA GPU, prepared for AI image generation. It comes with ComfyUI, AUTOMATIC1111 and the Stable Diffusion XL and Flux base models preinstalled. The GPU is exclusively yours. You access it from the browser and start generating.

Which plan should I choose — SD Studio or SD Studio Pro?

The SD Studio (20 GB of VRAM) is ideal for creation, prototyping, SDXL generation, quantized Flux and light LoRA training. The SD Studio Pro (96 GB of VRAM) is for production at scale: full FP16 Flux with no compression, multi-model, heavy batch and Dreambooth training with no GPU memory limit.

What comes installed?

ComfyUI (the node-based interface), AUTOMATIC1111 WebUI, CUDA and cuDNN, PyTorch and xformers, plus the SDXL and Flux base models already downloaded. On request, we also configure InvokeAI, ControlNet, IPAdapter, AnimateDiff and LoRA training tools (Kohya_ss, fluxgym).

How much does it cost and is there a setup fee?

The SD Studio costs US$ 649.80/mo and the SD Studio Pro US$ 2,575.80/mo. There is a one-time setup fee of US$ 259.80 (it covers preparing the server with the GPU, CUDA drivers and the image-generation environment). No contract.

How long until the server is ready?

Provisioning GPU servers takes up to 48 business hours. GPU servers have limited stock and dedicated preparation — driver installation, ComfyUI, A1111 and downloading the base models. The flow is: you request the plan, we confirm availability and delivery, and we provision it.

Can I train LoRA and use ControlNet?

You can. We configure the training tools (Kohya_ss, fluxgym) and the main nodes — ControlNet, IPAdapter, AnimateDiff, ADetailer. SDXL LoRA training runs well on the SD Studio; Dreambooth and Flux LoRA are more comfortable on the SD Studio Pro because of the VRAM.

Can it integrate with n8n and automations?

Yes. ComfyUI and AUTOMATIC1111 expose a native REST API. We configure authenticated endpoints so you can call image generation from within n8n workflows, your backend or any automation.

Is there human support?

Yes — human support 24/7, with people who understand CUDA, NVIDIA drivers and ComfyUI tuning. Rollin Host is a Brazilian company (Rollin Serviços Digitais e Tecnologia LTDA).

Pronto pra hospedar seu projeto de IA?

Comece em 5 minutos. Migração gratuita, suporte 24/7 em português e garantia de reembolso em 7 dias.