Learn AI /Generative Media /Synthetic media
Synthetic media
- Starter
- Generative Media
- Communications teams
- Policy officers
Also called: AI-generated media.
Back to the milestone you came from
In five seconds
What it is
Synthetic media is text, images, audio or video generated or substantially altered by software, often AI, so it can resemble authentic material.
In plain words
Synthetic media is text, images, audio or video generated or substantially changed by software to create content that may resemble authentic material.
An everyday example
A communications team needs a poster about safe water storage for a public campaign. They do not have a photo of the exact scene they want, so they describe it, choose colours, and combine elements that resemble many real photos they have seen before. The software produces an image of a family, containers and a background that looks convincing. Viewers may assume the moment really happened, even though it was assembled digitally for illustration.
A design team usually works slowly and deliberately, while many synthetic media tools can produce many variations in seconds. Also, in the analogy, people manually choose parts they can inspect. Real generative systems do not literally paste together visible pieces from a small set of photos; they calculate outputs from learned statistical patterns, so the process is less transparent.
Remember thisSynthetic media can be useful, but realistic output is not evidence of a real person or event. Check provenance, disclose material generation or alteration, and verify the underlying content.
Understand it
Understand it
Why it matters
Synthetic media affects trust. It can save time in communication, translation, accessibility and training, but it can also blur the line between illustration and evidence. Non-specialists need to ask basic questions: What was generated, what was edited, who made it, how is it labelled, and should it be treated as proof? Those questions help with safer communication, procurement, risk assessment and public trust.
How synthetic media is made and why context matters
After viewing this, a reader should be able to explain that synthetic media starts from prompts or source files, is generated by a model, and then must be judged for purpose and authenticity.
01
Inputs
A person provides a prompt, source image, voice sample, video clip, or other instructions.
Draw a rounded rectangle labelled 'Inputs' with small icons for text, image, microphone and video inside.
02
Generation
A trained model uses learned patterns to produce a new media file or alter an existing one.
Draw a second rounded rectangle labelled 'Generation model'. Show gears or layered nodes, but keep them simple and abstract.
03
Output
The result may be text, image, speech, music or video that appears polished or realistic.
Draw a third rounded rectangle labelled 'Output' with thumbnail icons for document, picture, waveform and play button.
04
Use context
The same output can be acceptable as labelled illustration or misleading if presented as a record of reality.
Draw a fourth rounded rectangle labelled 'Use context' splitting into two tags below: green 'Labelled synthetic example' and red 'Claimed as real evidence'.
A left-to-right flow diagram with four boxes: Inputs, Generation, Output, and Use Context. Arrows connect each box. Below the final two boxes, a green tag reads “clearly labelled illustration” and a red tag reads “falsely presented as real evidence”. Read from left to right, then compare the two outcome tags.
Hover or tap each box to reveal one short example: prompt, model action, output type, and context label.
Describe this without the diagram
The diagram is a four-step flow from left to right. First, a user supplies inputs such as text instructions, an image, a voice sample or a video clip. Second, a generation model creates or alters media using patterns learned during training. Third, the system outputs text, images, audio or video that may look or sound realistic. Fourth, the meaning depends on use context: if the output is clearly labelled as synthetic or illustrative, it may be appropriate; if it is presented as proof of a real event or real speaker, it can mislead.
With reduced motion: Show the same four boxes and arrows without animation. Keep both end labels visible at all times: green for clearly labelled synthetic content, red for content falsely presented as real.
Synthetic media means media generated or substantially altered by computational systems to create or simulate content. The input can be a text prompt, a source image, a voice sample, a video clip, or a mix of these. The output can be new text, pictures, speech, music or video.
A common pattern is simple: a person gives instructions or example material, and the trained system uses patterns learned from training data to generate a new result. “Training data” means the examples used to tune the system before you use it. The result may look realistic, but realistic does not mean authentic.
This term is broader than “deepfake”. A deepfake is a fabricated or heavily altered image, audio clip or video made to imitate a real person, scene or event convincingly. Synthetic media also includes useful things such as translated voice-overs, draft illustrations, accessibility narration or simulated training videos.
Synthetic media and synthetic data are related but different. Synthetic media is made mainly for people to watch, hear or read. Synthetic data is usually made for analysis, testing or model training. Text generation names the process of producing written words. Synthetic media is the broader category of generated or substantially transformed media outputs, which can include generated text.
The key question is not only “Was AI used?” but also “What is this supposed to represent?” A generated image may be fine as an illustration if labelled clearly. The same image becomes a problem if it is presented as evidence of a real event.
How “A poster team building a scene from parts” maps onto the real thing
| In the story | In the AI system |
|---|---|
| The written description for the poster | A text prompt or other input given to a generative system |
| The software combining familiar visual patterns | A model generating media from patterns learned during training |
| The convincing campaign image | Synthetic media output |
| Viewers assuming the scene was photographed | Authenticity risk when generated content looks like evidence |
A worked example
A communications officer at a regional hub must clear a short video for a side event the next morning. The cleared concept note provides the message, but there is no footage or recorded presenter.
Starting from: A cleared script, approved terminology, licensed design elements and instructions for three language versions.
- The officer uses an approved tool to generate narration and simple animated scenes from the script.
- Language focal points check each version against the cleared text rather than assuming that fluent-sounding narration is accurate.
- The closing frame identifies the narration and visuals as generated, and the file metadata records the tool, source materials and editing steps.
- Before clearance, a reviewer checks that the video does not imitate a real person or imply that the scenes document an actual event.
Result: The side-event video is cleared with an accurate disclosure and a usable provenance record. Viewers can understand that the scenes illustrate the message rather than record a real presenter or event.
In a United Nations workplace
A legal officer in a specialised agency’s technical division is answering an internal auditor’s finding by close of business. The finding concerns a training video in which a contractor generated the presenter, voice and background without recording how those elements were produced.
The response needs to distinguish synthetic media from ordinary editing. Generating a voice or presenter, or materially transforming recorded media, creates new provenance, consent, licensing and disclosure questions even when the underlying script was cleared.
Before you try thisDo not upload the audit file, contractor records or procurement-sensitive material to a tool not approved for that information. A recognisable person’s likeness or voice must not be generated without the necessary consent and authority. The response should preserve provenance records and state where synthetic elements were used.
Where you may hear it
- A vendor demo during an ICT governance board meeting
- Clearance of a public information video before upload to the portal
- A legal review of consent and licensing for a generated voice
- An audit discussion about provenance records for generated images
Avoid confusion
Avoid confusion
Often confused with
Deepfake
A deepfake is an AI-generated or heavily altered image, audio clip or video that convincingly imitates an identifiable real person, usually in a recording or event that did not happen as shown.
Synthetic data
Artificially generated records designed to resemble data from a real population or process for uses such as testing, training or sharing.
Text generation
A system produces new written words from a prompt, patterning its output on examples seen during training.
Common misconceptions
Common misconception: Synthetic media always means deception.
Correction: Not always. Many uses are legitimate, such as clearly labelled illustrations, dubbing, accessibility narration or training materials. The risk comes from misleading context, weak disclosure or impersonation, not from every generated file.
Common misconception: If it looks real, it must show a real event.
Correction: Realistic appearance is not proof of authenticity. A system can generate convincing faces, voices or scenes that never existed. Verification still needs source checks, metadata, corroboration or trusted provenance methods.
Common misconception: Synthetic media only means fake videos of politicians.
Correction: That is a narrow case. The term also covers generated images, speech, music, text and other media, including benign uses. Deepfakes are one subset, not the whole category.
Quick check
Which statement best describes synthetic media?
Synthetic media is a broad category. It includes generated or heavily altered text, images, audio and video. Some uses are benign and clearly labelled; others are deceptive. It is broader than deepfakes and narrower than all digital editing.
Go deeper
Go deeper
The simple version
Synthetic media is text, images, audio or video generated or substantially changed by software to create content that may resemble authentic material.
The technical version
Synthetic media refers to text, images, audio or video generated or materially transformed by computational systems to create or simulate content, rather than merely store, compress or format it. Inputs may include prompts, conditioning signals, reference media or editing instructions. Depending on the system, generation may involve diffusion, adversarial training or other learned probabilistic mappings from input conditions to output content. Important issues include provenance, disclosure and evidential status: whether the output is labelled, traceable and appropriate for its claimed use.
Reviewed external resources
Primary sourceRecommendation on the Ethics of Artificial Intelligence (opens in a new tab)UNESCO
This UNESCO recommendation gives a policy and ethics framework for AI use and suits readers in policy, governance and public institutions.
Intermediatelink checked 2026-07-31
Primary sourceContent Credentials (opens in a new tab)Content Authenticity Initiative
This Content Authenticity Initiative resource explains content credentials and provenance approaches and suits readers focused on authenticity and labelling.
Starterlink checked 2026-07-31
EncyclopaediaSynthetic media (opens in a new tab)Wikipedia
This Wikipedia page gives a broad introductory overview of the term and suits readers who want a quick general starting point.
Intermediatelink checked 2026-07-31
VideoVideo explanations: Synthetic media (opens in a new tab)YouTube search
This YouTube search results page helps readers find different video explainers on the topic and suits people who prefer audiovisual learning; it is a search page, not one curated lesson.
search results, not one pagelink checked 2026-07-31
DiscussionCommunity discussion: Synthetic media (opens in a new tab)Reddit search
This Reddit search results page shows community discussions and examples people are debating and suits readers who want to see common questions and reactions; it is a search page, not one authoritative source.
search results, not one pagelink checked 2026-07-31
Where this came from
Drafted with a language model through an offline pipeline, checked against this site's schema, and every external link above fetched to see whether it answers. Treat it as a starting point and check anything that matters against the primary sources. How this site is built and what that means.
The one thing to remember
Synthetic media can be useful, but realistic output is not evidence of a real person or event. Check provenance, disclose material generation or alteration, and verify the underlying content.