Install a free AI image generator on your PC and create your first image in 10 minutes
Launch ComfyUI - a local node-based editor for generating images from text prompts, free and without sending data to the cloud
- 1.Download ComfyUI as a portable archive for Windows. Extract it to any folder on your drive, for example C:\AI\ComfyUI.
The archive is about 1 GB - download over a stable connection. Your antivirus may flag .bat files - this is a false positive.
ComfyUI ↗ - 2.Run run_cpu.bat (if you don't have an NVIDIA GPU) or run_nvidia_gpu.bat (if you have an NVIDIA card with 6+ GB VRAM). The console will show a line like 'To see the GUI go to: http://127.0.0.1:8188'.
Do not close the console window while ComfyUI is running. If port 8188 is busy - launch the file with the --port 8189 parameter via the command line.
ComfyUI ↗ - 3.Open http://127.0.0.1:8188 in your browser - the node-based interface will open with a preset default workflow. If the canvas is empty, click the 'Load Default' button on the right side of the menu.
Works in Chrome, Edge, and Firefox. Dark mode can be enabled via Settings (the gear icon in the menu panel).
ComfyUI ↗ - 4.Download the free SD 1.5 model (.safetensors file, ~2 GB) and place it in the ComfyUI\models\checkpoints\ folder. This is the «brain» that generates images. After placing the file, refresh the model list by clicking the circular arrows button on the Load Checkpoint node.
If you have an NVIDIA GPU with 8+ GB VRAM, you can download an SDXL model for better quality, but it weighs about 6.5 GB.
Civitai ↗ - 5.In the 'Load Checkpoint' node, select your downloaded model from the dropdown list. In the 'CLIP Text Encode (Prompt)' node, paste your image description in English into the text field - this text determines what the AI will draw.
English produces noticeably better results than Russian - the models are trained primarily on English datasets.
ComfyUI ↗Promptprofessional photo of a [subject], [environment], [lighting condition], shot on Canon EOS R5, 85mm lens, shallow depth of field, ultra detailed, 8k
What this prompt doesThis prompt creates a photorealistic portrait. Replace [subject] with your subject (e.g., a young woman with red hair), [environment] with the setting (e.g., in a sunlit coffee shop), and [lighting condition] with the lighting (e.g., soft morning light). The more detailed the description - the more predictable the result. - 6.In the bottom 'CLIP Text Encode (Prompt)' node, enter what you DON'T want to see in the image - this removes artifacts. Then click the 'Queue Prompt' button on the right. The image will appear in the 'Save Image' node within 15-60 seconds depending on your PC's power.
If generation takes too long or crashes with a memory error - reduce the width and height in the 'Empty Latent Image' node to 512x512. Saved files are in the ComfyUI\output\ folder.
ComfyUI ↗Promptblurry, low quality, distorted hands, extra fingers, watermark, text, signature, jpeg artifacts
What this prompt doesThis is a negative prompt - a list of things the AI should avoid during generation. You can keep it as is or remove items if they interfere with your desired result.
- freeComfyUIFully free open-source software, with no limits on the number of generations.
- freeSD 1.5 / SDXL ModelsMost models on Civitai are free to download; an account is only needed to download files.
- freeElectricity and PC resourcesGeneration runs on your hardware: the more powerful the GPU, the faster. It also works on CPU, but slower.
Why today
AI services change fast - interfaces and free limits may differ from what's described.