Vai al contenuto Term LoRA (Low-Rank Adaptation) — Definition, How It Works & Examples LoRA (Low-Rank Adaptation) is a fine-tuning technique that teaches a large AI image or language model a new concept — a face, style or object — by training a small set of extra low-rank weight matrices while the multi-billion-parameter base model stays frozen. The result is a tiny 10–200 MB adapter instead of a full multi-gigabyte model. Term Diffusion Model — Definition, How It Works & Examples A diffusion model is a type of generative AI that creates images, video or audio by learning to reverse a noising process: during training it repeatedly adds random noise to real data, then learns to remove that noise step by step. To generate something new it starts from pure noise and denoises it into a clean result. Diffusion is the technology behind Stable Diffusion, DALL·E, Midjourney, FLUX and most 2026 image and video generators. Term Transformer — Definition, How It Works & Why It Powers Modern AI A Transformer is a neural-network architecture that reads a whole sequence — words, image patches or video frames — all at once using a mechanism called self-attention, which lets every element weigh how much it depends on every other. Introduced by Google researchers in 2017, it is the foundation of nearly all modern AI, from ChatGPT, Claude and Gemini to image and video generators. Term Prompt (AI) — Definition, How It Works & How to Write One A prompt is the text instruction you give a generative AI model to tell it what to produce — an image, a video, a chat answer or code. It is the single most important control you have: the model reads your words, turns them into numbers, and generates an output shaped by exactly what you described. A clear, specific prompt is the difference between a generic result and a great one. Term What Is a Sampler in Diffusion? Euler, DDIM and DPM++ Compared A sampler — also called a scheduler or solver — is the algorithm a diffusion model uses to turn random noise into an image, removing a little noise on each step. It sets the speed–quality trade-off: a first-order sampler like Euler wants 20–30 steps, while UniPC or LCM can land a usable image in 5–10. Term CFG (Classifier-Free Guidance) — Definition, How It Works & Best Scale CFG (Classifier-Free Guidance) is the setting that controls how strictly an AI image model obeys your prompt. On each denoising step the model makes two predictions — one guided by your text, one ignoring it — and the CFG scale decides how far to push the result toward the guided one. Low values give looser, more creative images; high values follow the prompt harder but can look over-saturated. Term What Is a Seed in AI Image Generation? Meaning & How It Works A seed is the whole number that decides the initial noise an AI image model starts from. Keep the seed and 5 other things identical — prompt, model, sampler, step count and CFG — and you get the same image pixel for pixel; change only the seed and the same prompt gives a completely different picture. Term Negative Prompt — Definition, How It Works & Examples A negative prompt is a second text field that tells an AI image model what to leave out — words like "blurry, extra fingers, watermark, low quality" that steer the picture away from those traits. Where the normal prompt says what you want, the negative prompt says what you do not, and the model pushes the result away from it during generation. Term ControlNet — Definition, How It Works & Control Types ControlNet is a neural network that adds structural control to an AI image model: you feed it a reference — an edge outline, a depth map, or a pose skeleton — and the image it generates follows that structure while the /glossary/prompt fills in style, colour and detail. It is how you keep a specific composition, pose or layout instead of leaving everything to chance. Term IP-Adapter — Definition, How It Works & Use With ControlNet IP-Adapter (Image Prompt Adapter) is a lightweight add-on that lets an AI image model take a reference image as a prompt: you give it a photo of a face, a style or an object, and the model generates new pictures that keep that look. It brings "an image is worth a thousand words" to /glossary/diffusion models without retraining them. Term Stable Diffusion — Definition, How It Works & Versions Stable Diffusion is an open-source text-to-image AI model that turns a text prompt into a picture by starting from random noise and denoising it step by step inside a compressed latent space. Released by Stability AI in August 2022 with open weights, it is the model that put high-quality AI image generation on consumer GPUs and spawned tools like /glossary/lora, /glossary/controlnet and /glossary/ip-adapter. Term Midjourney — Definition, How It Works & Versions Midjourney is a closed, subscription-based text-to-image AI service that turns a written /glossary/prompt into a highly polished, stylised image. Built on a proprietary /glossary/diffusion model by the independent lab Midjourney, Inc., it launched in open beta in July 2022 and became known for its opinionated, art-directed default look — beautiful results with very little prompt effort, in exchange for less low-level control than open tools. Term DALL·E — is it still available, and what replaced it Short answer: DALL·E is gone. OpenAI switched off the official DALL·E GPT inside ChatGPT on 30 August 2026, and the dall-e-2 and dall-e-3 API endpoints had already stopped responding on 12 May 2026. DALL·E was OpenAI's family of text-to-image models that turned a written /glossary/prompt into an original picture; the name is a play on the robot WALL-E and the painter Salvador Dalí. Below: what it was, how versions 1, 2 and 3 differed, and what to use instead. 