Frame size and resolution
The reels feed runs vertical, and a 9:16 ratio fills the whole screen. The platform accepts a horizontal clip, but shows it as a strip in the middle of a black field, and a thumb swipes past before the viewer reads the first line.
The working resolution for filming is 1080 by 1920 pixels. Shooting four times larger gains little: on upload the platform recompresses video to its own dimensions, and the extra pixels vanish along with the time spent uploading.
The square 1:1 frame stayed with ordinary feed posts. If the same material goes to both places, film vertically and crop for the feed: the reverse leaves empty bands top and bottom.
- Aspect ratio 9:16, vertical
- Resolution 1080 × 1920 pixels
- Frame rate 30 per second, 60 for smooth motion
- H.264 codec, MP4 container
- Audio in AAC
Why a reels video looks worse than the original
The platform recompresses every uploaded clip, and fine detail suffers most: text on a light background, thin lines, subtle grain. Phone footage loses less than camera footage: a phone already hands over a compressed file, and a second pass adds little.
Material that passed through several programs before upload comes out noticeably worse. Every export adds its own loss, and by the fourth resave the picture breaks into blocks whenever anything moves.
The practical rule stays simple: edit in one program, export once, upload the original. Sending the file through a messenger before uploading destroys quality more reliably than any platform compression.
How long a clip should run
The upper limit for reels is three minutes. That is a technical capability rather than a recommendation: almost nobody watches such clips through, and the platform treats watch-through as the main sign of whether to distribute material further.
Short clips under fifteen seconds collect a full watch more often, which is exactly why they travel wider. A viewer reaches the end before deciding to leave, and the platform counts that as approval.
The middle zone of thirty to forty seconds works where something needs explaining. Reach for it when the material genuinely does not fit into fifteen seconds, not because longer feels more serious.
The first second and a half decides more than the rest
A viewer decides whether to keep watching almost instantly. An intro card with a channel name, a slow lead-in and silence at the start eat exactly that time, and the clip loses half its viewers before the content begins.
The opposite move works: start in the middle. Show the result straight away and give the explanation afterwards, once the viewer has already decided to stay.
Cover and first frame
The cover appears in the profile grid and in the reels tab. Nobody sees it in the feed itself: the video plays immediately there. One consequence follows: a cover works on people who already opened your profile, while the first frame works on everyone else.
A cover can be picked from the clip’s own frames or uploaded as a separate image. The second option frees up text and composition, but the image must match the video in mood: a mismatch reads as bait and raises the share of people who leave in the first second.
The profile grid crops covers at the edges. Everything that matters, text included, belongs in the middle of the frame, or the grid will show half-letters.
Sound, captions and on-screen text
A sizeable share of viewers watch with sound off, and a clip has to read in silence. Captions solve that completely: the platform can generate them, though the transcript needs checking by hand, because it mangles names and technical terms.
On-screen text competes with the interface. The post caption and buttons cover the bottom, the likes column covers the right. The safe zone is the middle of the frame with roughly fifteen percent clearance on every side.
Music adds a chance of reaching people who browse by sound, but it does not rescue weak video. A trending track on dull material produces a spike of impressions in the first hours and nothing afterwards.
What happens after publishing
The platform shows a new clip to a small slice of the audience and watches what follows. If the people who opened it watched through, liked, saved or shared, distribution widens. If they swiped past, it stops there.
That makes the first hours count for more than the following day. A clip published while the audience sleeps starts on a weak signal and rarely catches up.
Another common cause of a weak start is posting back to back. Two clips an hour apart split the same first wave of impressions between them, and both get less than one would have.
Five reasons a video does not travel
- A horizontal frame in a vertical feed
- An intro card or logo in the first seconds
- Text hidden behind the interface at the bottom and right
- A clip over a minute long for no reason
- Two clips in a row on the same day
How to tell what worked
Comparing raw view counts across clips of different ages makes no sense: the older one had more time. Compare shares instead, how many likes and saves landed per thousand views.
Saves show usefulness better than likes do. People like out of politeness and save what they intend to come back to, and the platform treats that as a strong signal.
Shares show that a clip made somebody want to discuss it. Material with a high share rate usually carries a long tail of impressions: it keeps travelling for weeks after publishing.
Working those shares out by hand across every post gets hard past the thirtieth clip. One table with every counter answers the question directly: which format works for you, and which one you keep filming out of habit.
Filming: light, steady hands, vertical frame
A phone films well enough for reels, and buying a camera will not rescue the picture. Light will. A window to the left or right of a face gives soft shadow and living depth, while a ceiling lamp overhead gives bags under the eyes.
In the evening any lamp placed in front of the face at eye level does the job, rather than above it. A ring light helps by filling evenly, though it gives itself away with a circular catchlight in the pupil, which shows on a close shot.
Viewers forgive bad light more readily than a shaking camera. A phone propped against a stack of books beats an outstretched arm. In-phone stabilisation helps while walking, but it crops the frame edges, and text placed right against them can drift outside.
Sound records worse than picture
A built-in microphone collects the whole room: the fridge, the street, the echo off bare walls. The difference is audible immediately, and editing can repair almost none of it.
Two moves close most of the problem. Stand closer to the phone than feels necessary, and film in a room with soft things: curtains, a rug and a sofa kill echo better than any processing.
What to do with a finished clip
Not everybody reads the caption, but it feeds search inside the platform. The first line shows before anyone taps “more”, and it should carry the point rather than a greeting.
Topic tags help a small account at the start: they tell the platform who to show it to. Three to five relevant ones beat thirty in a row, and a pile of generic popular tags achieves nothing, because competition there is beyond reach.
Pin three clips that explain what you do. Somebody arriving from the feed decides within seconds whether to follow, and they decide on the top row of the grid.
Deleting a weak clip for the sake of statistics accomplishes nothing. The platform rates every post separately, and one poor piece does not drag the others down. Deleting also takes away the views it might still have collected in its tail.
Re-uploading the same clip is not worth it either. The copy starts as a new post, but the platform keeps a record of the previous one, and the second attempt usually collects less than the first.