All patternsPrompt actions
Transform
Description
Design considerations
Related patterns
Examples

Transform

Converting content from one modality to another while preserving intent, constraints, and fidelity is a keystone feature of generative AI. Examples include audio to text, text to audio, text to image, image to text (OCR or captioning), text to video, document to slides, code to diagram, and screenshot to HTML.

Different modalities are useful for different purposes. Text is the best tool for working through outlines and narrative. Seeing tokens translated into image can reveal tone and mood, while transforming text or data into diagrams allows users to review it easily. In this way, the transform pattern serves as a creative pipeline between the user and the AI.

Where to find the action

  • Right-click menus or inline actions (Midjourney)
  • Chained together on an open canvas (FloraFauna)
  • Direct generation via open input (Krea)
  • Background helpers, such as transcription (Descript)
  • Built into the core interface (Chronicle)

Design considerations

  • Maintain a connection to the original. Show outlines, prompts, and blended sources and make them accessible for additional iteration, even if the form of the generation has changed.
  • Add stop points between modalities. As transformational actions tend to be more expensive to fulfill, put action plans, sample responses, and verifications in front of the user before committing to a broader generative run.
  • Start small and work up. Show drafts of the transformed generation before committing to a full run, like generating an outline or snippet first.

Related Patterns

Branches

Encourage users to explore how the transform action will be interpreted by different models by supporting branching, along with clear descriptors of what each model excels in from the model selector.

Draft mode

Use the transform action to create a draft version of more expensive processes before they run, like an outline for a slide deck, action plan for a site, or outilne for a movie.

Examples

‍

Chronicle lets users fine tune their outline before transforming it into slides
Figma Make transforms visual designs into code
FloraFauna’s canvas cards support transformation between text, image, and video modalities
FloraFaunra focuses on seamless exploration by allowing transformations to flow seamlessly through forms as the user explores how to visualize a concept
Granola is one of many audio-to-text transcription tools, supported by AI features of synthesis and summarization
Julius can transform data tables into charts so users can visualize and interact with the data within the tool instead of moving it elsewhere
Text to media is a fundamental transformational mode of generators. Here, Krea transforms from text to video
Midjourney supports image to text transformation, using an image to generate a prompt that can be used for further generation
Midjourney supports image to video transformation as an easy-to-reach inline action
Midjourney’s suggest a prompt feature (previously “describe” in discord) allows users to transform image to text and back to image
Perplexity allows users to generate images of different styles from a search result

© Emily Campbell 2025 | Sharable under CC-BY-NC-SA | Contact

Sharable under the Creative Commons CC-BY-NC-SA license