# AI video glossary: face swap, deepfake, reference photo and more > Plain definitions of the words you meet when you put people into videos with AI, from face swap and deepfake to reference photo, take, artifact and 480p. Source: https://vswp.app/guides/ai-video-glossary/ · Updated 2026-10-08 A face swap replaces a person's face in a video with another face, a whole-person swap also changes clothes and build, and a deepfake is any realistic AI fake of a real person. A reference photo is the picture the AI uses to learn what the new person looks like. Below are these and about 20 other terms you'll meet when you put yourself or friends into a video with AI, in plain words. ## Face swap, person swap, deepfake: what's the difference? These three get mixed up most often. - **Face swap.** The AI changes the face (and usually the hair) of someone in a video, while the body, clothes and moves stay as they were. - **Whole-person swap** (also "body swap" or "character swap"). The new person's clothes, hair and silhouette come from the photo too, so you see them head to toe. - **Deepfake.** The everyday name for a video, photo or voice made or changed with AI so that a real person seems to do or say something they never did. A realistic face swap is technically a kind of deepfake; what separates a joke from a fake is consent and honesty. | Term | What changes | Typical use | |---|---|---| | Face swap | Face and hair | Trends, memes, costumes | | Whole-person swap | Face, hair, clothes, silhouette | Seeing yourself in your own clothes | | Deepfake | Anything, made to look real | Ranges from labeled jokes to scams | For the full picture, see [how AI video face swap works](/guides/ai-video-face-swap-explained/). ## What do you put in? - **Source video.** The clip you start from. On [vswp](/) it's either a video from the Trends library or one you upload: 1 to 120 seconds, up to 150 MB, any common format. - **Trend.** A popular video, dance or meme that many people repeat. In the [Trends library](/trend/) the people in each video are already found, so you only add photos. - **Processed part.** The stretch of the video the AI works on, up to 30 seconds. For a longer clip you can choose which part. - **Reference photo.** The photo that shows the AI what the new person looks like. It's a model to follow, not a sticker pasted on top. One sharp photo of one person in good light works best; see [which photos work best](/guides/best-photos-for-ai-video-swap/). - **Portrait vs full-body photo.** A portrait shows the face and hair. A full-body photo also carries clothes and silhouette. A portrait can't pass on clothes it doesn't show. ## What does the AI do? - **Generative AI.** AI that creates new content instead of editing existing pixels. That's why every result is a new variation, not a frame-exact copy. - **Person detection and tracking.** Finding every person in the video and following each one from frame to frame. On vswp this runs on its own after you upload, usually in a couple of minutes, and it's free. - **People cards.** The people found, shown one by one with the main people first. People seen briefly or far away are grouped as the **background**. - **Keep the original look.** A vswp switch. On: clothes stay as in the video and only faces and hair change. Off: the look, clothes and hair, comes from the photo. - **Render.** The run in which the AI actually makes your new video. You see the estimated time and the price before it starts. - **Take.** One version of the result. Pressing **Create video** again makes a new take, which is a new render. - **Photo-to-video (image-to-video).** A different kind of AI video: it animates a single still photo into a short clip. A person swap starts from a real video, so the moves, camera and sound come from that video. ## What do the words about the result mean? - **Resolution: 480p and 720p.** The height of the picture in pixel rows. 720p is sharper; 480p is lighter and fine for a quick look on a phone. vswp gives you an MP4 at either. - **Frame rate (fps).** How many pictures per second the video shows. On vswp it follows the source, up to about 30 fps. - **Original audio.** The sound from the source clip. The swap changes how people look, not how they sound. - **Likeness.** How much the new person resembles the photo. It can vary from take to take. - **Artifact.** A visible glitch in AI video: merged fingers, a face that smears during a fast turn, or details mixed up in a crowd. If you see one, [here's how to get a better take](/guides/ai-face-swap-does-not-look-like-you/). - **Watermark.** A small mark on the video. On vswp, the free preview carries a small "vswp.app" mark; paid videos have none. - **Free preview.** A short sample before you pay. Every new vswp user gets one: a few seconds of their video at 480p, from where the people with photos are on screen. The full video is paid in Telegram Stars (⭐); the price depends on length, quality and the number of people, and the exact price is shown before you start. ## Which words are about using it fairly? - **Consent.** The person in the photo agreed to be put into the video. Ask before you use anyone's photo, and only use videos you have the right to use. Our guide to [face swap consent and safety](/guides/face-swap-consent-and-safety/) explains more. - **AI label (disclosure).** A note such as "made with AI" on a post, or a platform's "AI-generated" tag. In the EU, the AI Act's rule on disclosing deepfakes applies from August 2026; rules differ by country, and this is not legal advice. - **Deepfake detection.** Checking whether a video is a fake. No tool gives certainty; see [how to spot a deepfake](/guides/how-to-spot-a-deepfake/). ## FAQ ### What is a reference photo in AI video? It's the photo the AI uses to learn what a person looks like before it redraws them in the video. A sharp photo of one person, with the face clearly visible, gives the best likeness. ### Is every face swap a deepfake? A realistic face swap is a kind of deepfake in the technical sense. A labeled joke with friends who agreed is very different from a fake made to deceive someone. ### What does "take" mean in AI video? A take is one version of the result. Because the AI generates new frames each time, a second take of the same video and photo looks a little different. ### Does a face swap change the voice? No. On vswp the result keeps the original audio of the source clip, so the voices and music stay the same.