Short-form video glossary

Short definitions for product and SEO terms used across HypeNest docs and guides.

Terms for creators and marketers. Not legal advice.

Terms

Video repurposing

Turning one long source video into multiple shorter assets — clips, titles, thumbnails, and posts — for other formats and platforms. Repurposing is the whole job after the recording exists: extract, package, review, then publish. It is not a synonym for “make a clip.”

AI clip generator

Software that proposes short vertical clips from a long video — often with captions — for human review before anything is published. The generator is a draft engine, not a publisher: it ranks moments, cuts them, and leaves keep-or-kill to an editor.

TikTok clip

A short vertical video built for TikTok’s mute-first feed. Openings, on-screen text, and pacing decide whether the viewer stays. A TikTok clip is not “the leftover of a YouTube edit” — it is a full packaging job.

Hook

A hook is the opening of a short video — usually the first one to three seconds — that makes a scrolling viewer decide to keep watching. It has a visual part (the first frame: a face mid-sentence, a result, something in motion) and a verbal or text part (the first line spoken or shown on screen). Common hook types: a bold claim, a specific question, a visible result shown before the explanation, a mistake or warning, or a story dropped in at the most tense moment. A good hook makes a promise the rest of the clip keeps. A clickbait hook that the clip does not pay off may win the first second and lose the rest, which shows up as a cliff in audience retention.

Burned-in captions

Subtitles rendered into the video frames so they appear even when the platform player is muted, downloaded, or re-uploaded. They are a packaging choice, not just accessibility garnish.

Direct publish

Sending a reviewed asset to a platform from the product instead of downloading and re-uploading. HypeNest direct publish is live for YouTube and TikTok only today. Instagram, Facebook, and LinkedIn are not live publish targets.

Review before publish

Human check of AI-generated clips and packaging before anything goes live — core to brand-safe automation. The metric is time-to-approved-post, not time-to-first-AI-cut.

Source video

The original long recording — podcast, webinar, tutorial, demo, interview — used as input for repurposing. Clip quality is capped by source quality: muddy audio and rambling structure yield weak openings no generator can invent.

Packaging

Titles, descriptions, thumbnails, tags, and on-screen text that frame a clip for search and clicks. The cut is not the whole job. A tool that only exports the cut leaves this work on the table.

Batch clipping

Creating multiple short clips from one source video in a single pass so you can schedule a week of posts instead of one-off edits. Batch clipping only helps if review keeps quality from collapsing.

Clip pacing

How quickly cuts, captions, and visual changes move inside a short video so viewers stay through the payoff. Pacing is not “faster is always better” — mismatched speed to the niche feels like spam.

Hook rate

A practical retention diagnostic: the share of viewers who continue past the opening seconds of a short clip. It is more useful than a vanity view count when you are choosing which opening to reuse.

Completion rate

The share of views that reach the end of a short video. Useful for comparing hooks and lengths inside a niche. A high completion on a 7-second clip is not the same achievement as on a 45-second lesson.

Upload cadence

The planned rhythm of short-form publishing — for example 1/day or 5/week — that clipping workflows are designed to support. Cadence without reviewed keepers is a content debt plan.

Short-form video

Mobile-first vertical video built for swipe feeds such as YouTube Shorts, TikTok, and Instagram Reels. YouTube Shorts is YouTube’s version: square or vertical videos up to three minutes, shown in the Shorts feed, and able to route viewers back to long-form. Short-form is a consumption context — mute, swipe, speed — not merely a file duration.

B-roll

B-roll is footage that plays over or between the main shot, the A-roll, while the main audio keeps running. In an interview, the A-roll is the guest talking; the B-roll is the product they mention, the street they describe, the chart behind the number. Common types: cutaways to the thing being discussed, screen recordings, close-up inserts of hands or objects, establishing shots of a place, archive photos, and licensed stock. In a vertical Short, B-roll does two jobs: it shows what the speaker describes, so a viewer with the sound off still follows, and it breaks up a single static frame that would otherwise fill the whole clip. It works when it matches the exact sentence it covers; generic stock of hands on a keyboard fills seconds and explains nothing.

Jump cut

A jump cut is an edit between two shots of the same subject from the same (or nearly the same) camera position, so the subject appears to jump forward in time. In classic film grammar it was treated as a mistake. On YouTube and TikTok it became the default way to tighten a talking-head recording: cut the pauses, false starts, and filler, and the speaker “jumps” to the next useful sentence. Jump cuts raise pacing without new footage. Overused, they make speech sound breathless and remove the natural rhythm that makes a joke or a pause land. A small punch-in (a slight zoom on alternate cuts) is a common way to soften the jump.

Safe zone

A safe zone is the part of a vertical frame that platform interface elements do not cover. On YouTube Shorts, TikTok, and Instagram Reels, the app draws the username, caption text, music label, and action buttons (like, comment, share) on top of your video — mostly along the bottom and the right edge, with a smaller band at the top. Anything you place under those elements — burned-in captions, a face, a product, a call to action — can be hidden on a phone. The exact overlay size varies by app version, device, and whether the post is organic or an ad, so safe-zone guides are approximate. The practical rule: keep text and faces in the centre band and leave the bottom and right edge for the UI.

Aspect ratio (9:16)

Aspect ratio is the proportion between a video’s width and height. 16:9 is the landscape shape of most YouTube videos and TVs; 9:16 is the same shape turned upright — the full-screen vertical format of YouTube Shorts, TikTok, and Instagram Reels. A 9:16 video is commonly delivered at 1080×1920 pixels. Other ratios you will meet: 1:1 (square), 4:5 (a tall feed post), and 4:3. Aspect ratio is not resolution: a 720×1280 and a 1080×1920 file share the same 9:16 shape at different detail levels. To check a ratio, divide width by height: 1080 ÷ 1920 = 0.5625, the same as 9 ÷ 16. Mismatched ratios are why a file shows black bars or gets cropped by the player — the platform fits the video to its own frame. Instagram, for example, accepts anything between 1.91:1 and 9:16.

RPM (revenue per mille)

RPM stands for revenue per mille — how much a creator earns per 1,000 views, after the platform’s share. On YouTube, RPM is a creator-side metric in YouTube Analytics that includes all revenue sources counted there (ads, and on long-form also memberships, Super Chat, and similar), divided by total views, not only monetized ones. That is why RPM is lower than CPM. Shorts RPM is calculated separately because Shorts ad revenue is pooled and shared, not attached to ads on individual videos. RPM varies with audience country, niche, season, and format, which is why screenshots of someone else’s RPM tell you little about your own. Use your own analytics as the source of truth.

CPM (cost per mille)

CPM means cost per mille: the price an advertiser pays for 1,000 ad impressions. It is the buyer’s metric. In YouTube Analytics, creators see CPM as the estimated amount advertisers pay per 1,000 impressions on their monetized playbacks — before YouTube takes its share — and a related playback-based CPM. CPM moves with advertiser demand: audience country, topic, time of year (Q4 is usually higher), and device all change it. A high CPM does not guarantee high earnings, because not every view shows an ad and the creator receives a share of the revenue, not the full CPM. For planning income, RPM is the more honest number; CPM explains why some niches and months pay better.

SRT file

An SRT (SubRip Subtitle) file is a plain-text caption file. Each cue has a sequence number, a start and end timestamp in the form 00:00:01,000 --> 00:00:03,500, and one or two lines of text, separated by a blank line. Because it is plain text, you can open it in any editor, fix a misheard name, and upload it again. YouTube accepts SRT uploads as closed captions, and most editing tools can import or export it. An SRT stores words and timing only: no font, colour, or position. That is the key difference from burned-in captions, which are pixels in the video. SRT is also a common hand-off format for translators working on subtitles in other languages.

Open vs closed captions

Closed captions are a separate text track the viewer can turn on or off — the CC button. The platform renders them, so they can be edited after upload, indexed as text, and shown in the viewer’s chosen size. Open captions are part of the picture: rendered into the frames, always visible, styled exactly as the editor designed them, and impossible to switch off. Burned-in captions are open captions. Short-form feeds lean on open captions because many people scroll with the sound off and styled word-by-word text holds attention. Closed captions remain the accessibility standard for long-form, and they can be translated. Many channels use both: open captions on the vertical clip, a closed-caption track on the full video.

Watch time

Watch time is the total time viewers spent watching a video or channel, usually reported in hours or minutes. It is different from view count: 1,000 views of a 60-second clip where everyone leaves after five seconds produce far less watch time than 300 views watched to the end. YouTube uses public watch hours from long-form videos as one route into the YouTube Partner Program, and Shorts views are counted separately for that threshold. On short-form, average view duration and percentage viewed tell you more than raw watch time, because clip lengths differ so much. Watch time rewards clips that deliver what the first second promised; it punishes padding and slow openings.

Audience retention

Audience retention shows what share of viewers are still watching at each moment of a video. YouTube Analytics draws it as a curve; TikTok shows a similar retention graph and average watch time. Read the shape, not only the average: a steep drop in the first two seconds means the opening failed; a gradual slope is normal; a sudden cliff mid-clip marks a confusing cut or a dead stretch; a bump means people rewound or replayed. On Shorts, retention can exceed 100% at the start of the curve because of replays and loops. Compare curves between clips of similar length and topic — a 10-second joke and a 90-second tutorial are not comparable.

Click-through rate (CTR)

Click-through rate is the share of impressions that turned into a view: clicks divided by the number of times the thumbnail or post was shown. On YouTube, impressions CTR measures how often people clicked your thumbnail after seeing it on YouTube surfaces such as Home, search, and suggested videos. It is the packaging metric: title and thumbnail drive it; the video itself does not. On Shorts and TikTok the main feed plays the next video automatically, so there is no click to measure there — CTR matters for Shorts mostly on search, the channel page, and other browse surfaces. A rising CTR with falling retention usually means the packaging promised something the video did not deliver.

Thumbnail

A thumbnail is the still image that represents a video before it plays — in search results, suggested videos, the channel page, and playlists. Together with the title it makes the click decision. Effective thumbnails are readable at the size of a fingernail: one subject, strong contrast, a face or object that shows the topic, and little or no text that repeats the title. For YouTube Shorts, you choose one of YouTube’s auto-generated options or upload a 9:16 image in YouTube Studio on a computer; the cover appears on the Shorts shelf, the channel’s Shorts tab, and search. On TikTok, the cover frame appears on your profile grid, which is where returning viewers decide what to watch next.

Talking head

A talking-head video is one where the main picture is a person speaking to camera or to an interviewer — podcasts, interviews, webinars, tutorials, commentary. It is the most common source for short-form clips because the value is in what is said, and a single sentence can stand alone. The weaknesses are visual: one static frame, little motion, and nothing on screen that shows what is being described. That is why talking-head shorts usually add burned-in captions, jump cuts that remove pauses, a slight punch-in on emphasis, and B-roll that illustrates the point. The strongest talking-head clips start on the claim, not on the greeting or the setup question.

Lower third

A lower third is a text graphic in the lower part of the frame — traditionally the bottom third — that identifies a speaker, a place, or a topic: a name and title, a location, a quote attribution, or a social handle. It comes from broadcast news and is standard in interviews, webinars, and corporate video. In vertical short-form, the classic lower-third position collides with the platform interface: TikTok, Shorts, and Reels draw the username, caption, and buttons over the bottom of the frame. So on vertical video, a “lower third” usually has to move up — just above the bottom UI band or into the upper-middle — and be shorter than on a landscape video.

Loop (looping Shorts)

A loop is a short video that plays again from the start when it ends. YouTube Shorts and TikTok replay automatically, so a viewer who does not swipe watches the opening a second time. A seamless loop is edited so the last frame flows into the first — the final sentence leads straight into the opening line, or the motion at the end matches the start — so the replay feels like a continuation rather than a restart. Loops increase replays and total watch time. YouTube counts a view the moment a video starts to play (its Help Center dates this rule, across all formats, to August 24, 2026), while Partner Program earnings are still based on “engaged views.” A loop helps only if the content is worth a second watch; tricking viewers into replays does not build an audience.

Vertical video

Vertical video is video that is taller than it is wide — usually 9:16, sometimes 4:5 — designed to fill a phone screen held upright. It is the native format of YouTube Shorts, TikTok, Instagram Reels, and Stories. Filming vertically from the start keeps the subject framed for mobile; converting landscape footage means cropping and reframing, which works well for one speaker and poorly for wide shots, slides, or several people across the frame. Vertical video changes composition rules: less horizontal space, more headroom decisions, and interface overlays along the bottom and right edge that you must design around.

Auto-reframe

Auto-reframe is automatic cropping that moves a vertical (or square) window across a landscape video so the subject stays in frame. Instead of a fixed centre crop, the software tracks faces, the active speaker, or motion and pans the crop as the subject moves. Editors such as Adobe Premiere Pro ship an Auto Reframe feature, and AI clipping tools apply the same idea when they turn 16:9 recordings into 9:16 shorts. Auto-reframe works best with one or two people and a stable camera. It struggles with slides, wide group shots, fast action, and scenes where the important thing is not a face — which is why every automatic crop still needs a human check on a phone-sized preview.

Speaker detection

Speaker detection is the part of video and audio analysis that works out who is speaking and when. Audio-side, it is often called speaker diarization: splitting a transcript into “speaker 1” and “speaker 2” segments. Video-side, active speaker detection matches the voice to a face on screen, usually from lip movement and face tracking. Clipping and reframing tools use it to decide where to point the vertical crop in a podcast or interview, and to label captions by speaker. It fails in predictable places: people talking over each other, off-camera voices, profile shots, masks, and wide shots where faces are small. That is why speaker-based crops need review.

Montage clip

A montage clip is a short-form video built from several moments of one source video, cut together so they tell a single story. A continuous clip takes one uninterrupted segment, which suits a self-contained answer or anecdote. A montage clip takes, for example, the three best plays of a match, the key steps of a recipe, or the setup and payoff of a story told twenty minutes apart. It trades the natural flow of one segment for density: more payoff per second, but every join is a cut the viewer has to follow. Good montage clips keep one clear thread (one topic, one question, one result) and avoid joining moments that depend on context the viewer never saw.

Caption style preset

A caption style preset is a saved set of caption design choices (font, size, colours, outline or background box, word highlighting, animation and position) that you apply to a clip in one step instead of styling every video by hand. Presets keep a channel’s clips recognisable and make batch work consistent. Typical families are word-by-word highlighting in the karaoke style, text on a solid box for busy footage, large single words for fast edits, and quieter lower captions for interviews. A preset is a starting point, not a guarantee: long words, a crowded frame or the platform’s own interface can still call for a size or position change on a particular clip.

Export resolution

Export resolution is the pixel size of the video file you render out of an editor, for example 720×1280 (720p) or 1080×1920 (1080p) for a vertical 9:16 clip. Together with bitrate, it decides how much detail the file carries. Exporting above the source resolution adds no detail; it only scales the same picture up. Platforms such as YouTube and TikTok re-encode every upload, so a sharper export gives their processing more to work with, and small text and caption edges are the first details to suffer when the export is too low. Higher resolution also means larger files, which matters for batches, mobile uploads and drafts sent for review.

Watermark

A watermark on a video is a logo, name, or semi-transparent mark laid over the picture, usually in a corner, for all or part of the runtime. Creators and studios add their own to show who made a clip and to make reposts traceable. Free editing apps often stamp their logo on exports as the price of the free tier, and some social apps add their logo and the creator’s handle to videos saved from them. That is why a watermarked clip reposted elsewhere looks second-hand: viewers read the mark as “this came from another app”. Unlike burned-in captions, a watermark carries no information about the content; it only labels the source.

AI dubbing

AI dubbing replaces the speech in a video with a translation spoken by a synthetic voice and timed to the original. A full pass separates the voice from music and effects, works out who is speaking, translates each line, voices it, and mixes it back over the original background. Good dubbing keeps each speaker’s own voice, cloned from the source, so a two-person interview still sounds like two people. It also keeps the music and laughter, and fits each translated line into the time the original took, which often means a tighter translation rather than a faster voice. Subtitles translate the words; dubbing lets viewers watch without reading.

Voice-over

A voice-over is narration recorded apart from the picture and laid over it: you hear the speaker without seeing them say the words. Documentaries use it to explain what is on screen, tutorials to talk over a screen recording, and short-form clips to add a spoken hook to footage that opens without one. It differs from dubbing. Dubbing replaces what the people on screen say with another language; a voice-over adds a new voice on top, usually with the original sound turned down underneath (ducking). A voice-over can be recorded by the creator, read by hired talent, or generated from a written script with an AI voice.