Prompt
What is Prompt?
A Prompt is the text you type into an AI generation tool to describe what you want it to create. Better prompts ( more specific, more detailed, better structured ) produce better results.
At a glance
- Also known as
- Text promptGeneration promptInput promptQuery (in some contexts)
- Used for
- Instructing an AI model to generate a specific image, video, or textCommunicating creative vision, style, composition, and quality requirements to a generation modelThe primary interface between human intention and AI generation capability
- Common tools
- All AI generation interfaces (midjourney, stable diffusion, ChatGPT, claude, kling, runway, morphic)Prompt builders and structured prompt toolsPrompt libraries and community resources
- Related terms
- Prompt engineeringNegative promptCFG scaleSeedModelIteration
Ready to create?
Direct scenes, design characters, and ship full films
All-in-one AI creative platform with simple, transparent pricing, no speed throttles, and an infinite Canvas for max creativity.
How it compares
Compared with related concepts
A prompt is to an AI generation model what a brief is to a human creative: it communicates what is needed, in what style, to what standard, and for what purpose. Just as a good creative brief produces better creative work than a vague one, a good prompt produces better generation than a vague one. The difference is that a human creative can ask for clarification; a generation model works only from the information provided. This is why prompt quality is so directly consequential: the model has no recourse for ambiguity except to fill in gaps from its training distribution, which may or may not align with the creator's intention.
Think of it like…
A prompt is like giving instructions to a highly capable but very literal assistant who will do exactly what you describe, nothing more and nothing less: so the quality of what they produce is entirely dependent on the quality of the instructions you give. If you ask for 'a painting of a house', you will get some house. If you ask for 'a warm-toned oil painting of a Victorian terraced house at dusk, windows lit from inside, garden overgrown, impressionistic brushwork', you will get something far closer to your actual vision.
Pro tip
Build prompts from the most specific and critical information first ( the subject, key visual characteristics, and style ) before adding secondary details. Models attend to earlier elements in a prompt more consistently than later ones in many implementations, so burying your most important descriptors at the end of a long prompt may reduce their influence on the output. Test this with your specific model by comparing prompts with the same information in different orders, and develop an understanding of how your chosen model weights prompt position.
Types and variations
- The simple prompt is a brief, natural-language description with minimal structure: most effective with newer, more instruction-following models.
- The structured prompt organises information into explicit categories: subject, environment, style, lighting, mood, technical quality.
- The weighted prompt uses syntax to indicate the relative importance of different elements: parentheses or brackets to increase emphasis, specific platforms offer their own weighting syntax.
- The conversational prompt (primarily in language model contexts) frames the request as natural dialogue rather than structured instruction.
- The negative prompt specifies elements to avoid or exclude (see Negative Prompt entry).
Ready to make your first scene in Morphic?
Try MorphicCommon use cases
- Prompts are the fundamental input for every AI generation task: generating images for commercial, editorial, or creative use; creating video sequences for film, content creation, or marketing; generating text for writing, editing, coding, or research assistance; creating audio, music, or sound design through generative audio tools; and any other AI generation application in which human intent must be communicated to a model to produce a specific output.
- The ability to write effective prompts is the most universally relevant skill for any creator working with AI generation tools.
Ready to create?
Direct scenes, design characters, and ship full films
All-in-one AI creative platform with simple, transparent pricing, no speed throttles, and an infinite Canvas for max creativity.
FAQs
A prompt is the text input provided to an AI generation model that instructs it to produce a specific output. It is the primary interface between human creative intention and the model's generation capability: the means by which a creator tells the model what to generate, in what style, with what characteristics. The quality and specificity of the prompt directly affects the quality and relevance of the output.
A prompt should include enough detail to give the model clear, specific guidance on the most important aspects of the desired output: subject, style, composition, lighting, mood, quality. Beyond that, the optimal level of detail depends on the model and the task. Some models respond well to very detailed, keyword-rich prompts; others perform better with natural-language descriptions. Test incrementally: start with core description, add detail progressively, and assess where additional specificity improves versus over-constrains the output.
A good image generation prompt typically includes: a clear subject description, compositional and framing specification (shot type, angle), visual style and aesthetic description, lighting description, mood or atmosphere, and quality indicators. It uses specific, evocative vocabulary rather than vague terms — 'warm late-afternoon golden-hour sidelight' communicates more than 'nice lighting'. It avoids contradiction (specifying incompatible elements) and is ordered with the most important information first.
A prompt (positive prompt) describes what the generation should include and what it should look like. A negative prompt describes what the generation should avoid or exclude: artefacts, unwanted styles, specific elements. The two work together: the positive prompt pulls the generation toward the desired outcome; the negative prompt pushes it away from undesired outcomes. Most generation platforms support both as separate input fields.
Yes, significantly. Different models have different training data, architectures, and fine-tuning, which affects how they interpret and respond to prompt vocabulary and structure. Some models respond strongly to cinematic terminology and stylistic vocabulary; others respond better to natural language descriptions. Some are sensitive to prompt order and weighting; others treat all prompt elements roughly equally. Developing familiarity with how a specific model interprets prompts is as important as understanding general prompting principles.
Yes. In image and video generation, prompts can include cinematographic specifications: shot type (close-up, wide shot, over-the-shoulder), lens characteristics (shallow depth of field, 35mm lens, anamorphic), camera position (low angle, bird's eye view), lighting setup (three-point lighting, Rembrandt lighting, golden hour), and film or photographic style (35mm film grain, cinematic colour grade). These specifications are often highly effective at directing the model toward specific visual qualities.
Prompt weighting is a technique in which specific words or phrases within a prompt are given increased or decreased emphasis using syntax supported by the generation platform: typically parentheses to increase weight or brackets to decrease it, sometimes with numerical values (word:1.5 for 150% weight). It allows creators to control which elements of the prompt exert stronger influence on the generation, boosting underperforming descriptors and reducing the influence of elements that are over-dominating the output.
Effective prompting is a skill developed through practice, observation, and study. Community platforms and Discord servers for specific generation tools (Midjourney, Stable Diffusion) contain extensive shared prompt resources and examples. Anthropic's prompting documentation at docs.claude.com covers prompting for language models in depth. Dedicated prompt engineering resources, model-specific guides, and the practice of systematically testing prompt variations on your chosen model are all effective learning approaches.
