AI Glossary / Sözlük
Understand the terms behind better AI creations.
Plain-language definitions for the concepts you will encounter while creating images, videos, audio, and prompts with AI.
Essential AI creation terms
Use this glossary alongside the Vynix blog guides when you want to understand a setting, compare models, or improve a prompt.
- AI model
- A trained system that turns an input such as text, an image, or audio into an output. Different models have different strengths, speeds, controls, and credit costs.
- Aspect ratio
- The width-to-height shape of an image or video, such as 1:1 for square, 9:16 for vertical, and 16:9 for widescreen. Choose it before generating so the composition fits its destination.
- Diffusion model
- A type of generative AI that begins with visual noise and repeatedly refines it into an image. It is behind many text-to-image and image editing workflows.
- Frame rate
- The number of frames shown per second in a video. Higher frame rates can make motion look smoother, while a cinematic look often uses a lower frame rate.
- Guidance scale
- A setting that controls how strongly a model follows a prompt. Higher values can increase prompt adherence but may make an image feel less natural or flexible.
- Image-to-video
- A workflow that uses a still image as the starting frame and generates motion from it. It is useful when you need to keep a particular subject, product, or composition.
- Inference steps
- The number of refinement passes a model uses while generating an image. More steps can improve detail, but they also take longer and do not always produce a better result.
- Keyframe
- A reference frame that establishes an important moment in a video. In AI video workflows, a starting image can act like a keyframe for the desired composition.
- Negative prompt
- A concise list of elements to avoid in a generated result, such as unwanted text, extra limbs, watermarks, or a visual style that does not fit the brief.
- Prompt
- The instruction you give an AI model. A useful prompt explains the subject, setting, visual treatment, composition, lighting, and the intended output.
- Prompt engineering
- The practice of writing, testing, and refining instructions so an AI model produces a more useful, controllable result. It is a creative process, not a magic formula.
- Seed
- A value that makes a generation repeatable when the model supports it. Reusing a seed can help you compare prompt changes while keeping the starting randomness similar.
- Text-to-image
- A workflow that turns a written description into a new image. It is commonly used for concepts, illustrations, product scenes, and social media visuals.
- Text-to-video
- A workflow that turns a written scene description into a short video. Clear direction about subject, action, camera movement, and atmosphere usually improves results.
- Upscaling
- Increasing the resolution of an image or video. Good upscaling preserves edges and textures, but it cannot reliably restore detail that was not present in the original.
- Visual consistency
- Keeping important visual choices stable across multiple generations, such as character traits, color palette, lens style, setting, and lighting.
- Workflow
- A repeatable sequence of creative steps. For example: draft an image, select a favorite, turn it into a video, upscale the result, and export for a specific channel.
Create with Vynix
Put the idea into practice.
Explore 120+ AI models for image, video, audio, and creative workflows in one mobile app.