Bukas na modelo sa pagbuo ng larawan mula sa kumpanyang nag-aalok din ng audio, video at 3D na tool
AI tools sa imahe Sikat na mga tool sa imahe ng AI stability.ai
What is Stable Diffusion
Stable Diffusion is a text-to-image model released in 2022 and the flagship product of Stability AI. It grew out of latent diffusion research by the CompVis group at LMU Munich, with Runway also involved in its development. Unlike cloud-only services such as Midjourney and DALL-E, its code and model weights are public, so you can run it on your own computer — a consumer GPU with about 2.4 GB of VRAM is enough.
What it can do
- Text to image from a written description
- Inpainting and outpainting
- Image to image: transform an existing picture according to a prompt
- Negative prompts to exclude unwanted elements
- Customisation: the community offers a huge range of fine-tuned models, style models and extensions
There are three main ways to use it: Stability AI's online service DreamStudio, the Stability API, and open-source community interfaces — most commonly AUTOMATIC1111's Stable Diffusion WebUI. SDXL 1.0, released in July 2023, introduced native 1024×1024 resolution; Stable Diffusion 3.5 is currently Stability's most capable image model.
Pricing
Under the Stability AI Community License, individuals, researchers and businesses with less than US$1 million in annual revenue can use the core models for free, including commercially; businesses above US$1 million need an Enterprise licence for commercial use. The online service and API are billed by usage.
Who it's for
- People who want to generate images locally without uploading material to the cloud
- Creators and developers who need deep customisation (training their own styles or characters)
- Developers building image generation into their own products
Good to know
- Running it locally takes some technical setup, including a Python environment and GPU drivers.
- You own the images you generate, but the licence still restricts harmful content.
Similar tools
Compare Midjourney (cloud-based, with a strong visual style), Jimeng and Kling AI.