A portrait video can be sent to YouTube Live inside a landscape 16:9 output without stretching the people or objects in it. First decide what should occupy the empty space on each side; then build that composition with FFmpeg and check it at the resolution you intend to send.
For most phone footage, fitting the complete portrait frame is the safest starting point. It leaves side space, which you can keep black, design as panels, or fill with a blurred enlargement; cropping is a different choice because it removes part of the original image.
Why portrait footage needs a landscape layout choice
A 9:16 phone recording and a 16:9 landscape broadcast have different shapes. If you place the portrait image in the middle of the landscape output and preserve its proportions, it occupies only a narrow central area. The remaining width is part of the composition, not a technical fault that must be hidden.
There are three basic approaches. Fit keeps the full portrait image visible and leaves space at the sides. Crop enlarges the footage until the landscape frame is covered, cutting off some of the top and bottom. Stretch changes the image proportions and makes subjects look wider or shorter; avoid it for people, devotional idols, products, or any other footage whose shape matters.
This article describes one landscape output made from portrait footage. It is not the same workflow as sending both a horizontal and a separate vertical live feed. YouTube documents vertical layout choices such as centre crop, fit-to-phone-width, and stacked in its Live Control Room; third-party dual streaming involves a separate vertical stream key and the appropriate content format for each output. Check YouTube’s live-streaming setup guidance if that is the result you need.
Before opening a filter graph, decide what viewers should see in the landscape frame. A temple programme might keep the full singer and instruments visible, with a restrained side panel for the channel name. A lofi station might use the same complete portrait animation over a softened enlargement. A simple loop assembled from existing files may have no need for either treatment; the guide to making a YouTube Live stream from a video playlist covers that broader source workflow.
Create a 16:9 output canvas
Choose the output dimensions before scaling the portrait image. For a common full-HD landscape composition, use a canvas of 1920×1080: the width-to-height ratio is 16:9. The source can remain portrait; FFmpeg’s job is to create the landscape frame and place the source within it.
The canvas dimensions do not force the foreground to fill the width. If the portrait source is scaled to fit within 1920×1080 while maintaining its aspect ratio, its height can reach 1080, but its width remains much narrower than 1920. That gap is expected. It is where the bars, panels, or background will appear.
For a simpler output, scale the source to fit and pad the result to the canvas size. A filter segment for that operation is:
scale=1920:1080:force_original_aspect_ratio=decrease,pad=1920:1080:(ow-iw)/2:(oh-ih)/2
The scale filter reduces the image as needed to fit inside the requested dimensions without changing its proportions. pad adds canvas area around it and centres the fitted image. With no padding colour specified, FFmpeg uses black. This is an illustrative filter segment, not a complete command guaranteed to match every input or installed FFmpeg build; consult the FFmpeg filters documentation for filter options and syntax.
Keep the intended frame shape consistent throughout the command. If you set a landscape output size elsewhere but later scale or crop the image differently, you can end up with a result that is not the layout you previewed. Also check sample aspect ratio: setsar=1 in a filter graph sets square pixels for the output, a useful explicit choice for ordinary web video. It does not repair an incorrectly composed frame.
If the source dimensions or orientation vary across a playlist, test more than one clip. A portrait file may be stored with rotation metadata rather than dimensions that look portrait at first glance. Your preview or probe should confirm what FFmpeg actually reads, and you should inspect the encoded output rather than assuming the source’s phone-gallery appearance will carry through unchanged.
Fit the portrait video without distortion
The fit operation preserves the entire visible frame by scaling it uniformly until it fits inside the landscape canvas. force_original_aspect_ratio=decrease is useful because it limits the scaled image to the canvas dimensions without forcing it to fill both dimensions. The shorter dimension remains shorter; it is not an instruction to stretch.
For a centred portrait image on a black background, the scale-and-pad segment above is often enough. If you need a more deliberate layout, build the same steps as labelled branches in a filter_complex graph. The labels make it possible to process the source in separate ways, then combine the resulting images with overlay.
A representative blurred-background graph has this structure:
[0:v]split=2[bgsrc][fgsrc];
[bgsrc]scale=1920:1080:force_original_aspect_ratio=increase,crop=1920:1080,boxblur=20:10[bg];
[fgsrc]scale=608:1080:force_original_aspect_ratio=decrease[fg];
[bg][fg]overlay=(W-w)/2:(H-h)/2,setsar=1[outv]
Here, one copy is enlarged until it covers the landscape canvas, cropped to the canvas dimensions, then blurred. The second copy is fitted to a narrower portrait region and placed in the centre. The 608 foreground width is an example layout value for the stated canvas, not a universal width; adjust it to suit the source and desired margins. The foreground and background are different uses of the same source, so the blurry sides do not imply that the full portrait image itself has been cropped.
This filter graph is illustrative, not a tested full command. Filter availability and accepted options depend on the FFmpeg build, and sources can differ in frame rate, pixel format, rotation metadata, or audio streams. If you use filter_complex, map the resulting video label, such as [outv], and separately map audio from the input if it has audio. Check the final output’s dimensions and listen to its sound before sending it to YouTube.
Choose black bars, panels, or blurred sides
The side treatment affects how legible and calm the finished stream feels. Black padding is simplest: it is visually neutral, adds no competing detail, and needs no second image branch. Its limitation is equally plain: the canvas has empty-looking black areas, which some channels may consider too stark.
Designed panels can use the side areas for a logo, schedule, language label, or a short note that helps viewers understand the programme. Keep text away from the portrait subject and test it at the size people will watch. A panel that looks balanced on a desktop preview may be difficult to read on a phone, especially if it is crowded with small text. Do not treat a side panel as extra room for an unchanging block of information that competes with the video.
A blurred enlargement gives the entire widescreen canvas colour and movement from the same footage. It can work well for devotional visuals, lofi animation, or a calm nature clip, but motion behind the sharp foreground can be distracting. A strong blur reduces detail but does not make a busy source inherently restful. Compare it with black padding using a representative moving section, not only a still frame.
| Treatment | What it does | Main trade-off | Useful when |
|---|---|---|---|
| Black padding | Fits the whole portrait frame and leaves black at the sides | Least complex, but the side areas remain plain | You want the source to remain the focus |
| Designed panels | Fits the source and uses side space for graphics or text | More layout work; small or busy content can distract | Viewers need a clear label or schedule |
| Blurred enlargement | Uses a scaled, cropped, blurred copy behind the fitted source | More filtergraph complexity and potentially more visual movement | Plain black sides do not suit the programme |
| Crop to landscape | Enlarges the source to cover the canvas and cuts off top or bottom | Fills the frame but loses image area | The subject remains safely inside the crop |
The first three options preserve the full foreground portrait image. The final row is deliberately different: it trades visible image area for a landscape fill. A playlist channel with varied source material may need different treatment per programme; the piece on scheduling recurring streams from pre-recorded videos can help you think through repeatable programming, but each video still needs its own composition check.
Crop only when losing image area is acceptable
Cropping can be right when the portrait source has enough unimportant space at the top and bottom, and the important content stays within the landscape frame. It can also suit footage where the subject is moving within a known safe area. But it does not preserve the entire portrait image: fitting a narrow portrait image across a wide landscape canvas requires a large enlargement, and the top and bottom must be removed to reach the landscape shape.
That loss matters in practical ways. A person’s head, subtitles at the bottom of a phone clip, a temple spire, a product in someone’s hands, or the lyric line in a music video can disappear. If the source is already tightly framed, cropping can make the result unusable. Do not decide from dimensions alone; preview the actual cut and check the whole duration for movement into the removed area.
In a crop-first layout, the background or foreground branch is scaled to cover 1920×1080 and then cropped to those dimensions. That is the same kind of operation used for the background in the blurred example, but there is no smaller fitted foreground over it. The distinction is the outcome: a cropped single image fills the output by losing its edges, whereas a fitted image shows the complete frame with side space.
Avoid stretching as a shortcut between fit and crop. It changes the proportions of the source to occupy the canvas, so faces, instruments, text, and familiar objects can look unnaturally broad. If full-width presentation matters more than preserving the complete portrait image, crop deliberately and inspect what is removed rather than distorting the image.
Check scaling, overlay, and encoding order
A filter graph is easier to troubleshoot when you think of it as a sequence: split the source if you need two versions, scale each branch for its purpose, crop the background branch if it must cover the canvas, blur that branch if desired, then overlay the fitted foreground. Finally, set the output sample aspect ratio and map the completed video label. The order matters because a crop before scaling and a crop after scaling can produce different composition and detail.
Use the same output dimensions for the canvas and the encoded video. If the encoded stream is 1920×1080, inspect that exact output size. A lower-resolution preview can hide edge issues, make overlays look different, or conceal whether the portrait region is centred. For designed panels, review the text at realistic playback size as well as at full resolution.
Encoding settings are a separate decision from composition, but they affect whether the feed reaches YouTube reliably. YouTube’s current encoder guidance lists RTMP or RTMPS ingest, constant bitrate, and a recommended keyframe interval of two seconds, not exceeding four seconds. For H.264 at 1080p30, its table gives 5 Mbps minimum and 14 Mbps recommended. Those are platform settings, not a guarantee that your upload connection can sustain a stream; check the current YouTube encoder settings and choose a quality your reliable upload capacity can support.
YouTube also says it transcodes live streams to create output formats for viewers on different devices and networks. That does not remove the need to send a correctly composed and stable input. A portrait image with the wrong aspect ratio, missing audio, or an unstable feed still needs fixing at the source. For a long-running channel, distinguish the visual layout question from the decision about how to keep a stream running; the guide to switching a YouTube loop from a cloud service to FFmpeg on a VPS addresses a different operating arrangement.
Preview the composition before going live
Render or preview a short representative segment before broadcasting publicly. Include movement near the top and bottom of the portrait image, any on-screen text, and a section with the kind of audio viewers will hear. A still frame can show whether the layout is centred, but it cannot reveal that a moving hand, face, caption, or instrument disappears during a crop.
Check the encoded file’s width, height, and playback, then watch it at the intended player size. Verify that the fitted portrait content is not stretched, that the side areas are the treatment you chose, and that overlays do not cover important details. Listen through headphones or speakers you know; verify that the audio is present and in sync rather than assuming a successful video filter graph means the complete programme is sound.
When ready, test privately or unlisted as appropriate, and monitor stream health in Live Control Room before relying on the setup. YouTube recommends checking the stream and testing representative motion and audio. Its recommended upload encoding settings are also worth reviewing when your source is a pre-rendered file. Repeat the check after changing dimensions, filter options, frame rate, or encoder settings.
If the computer used to keep an FFmpeg broadcast running is itself the weak point in an overnight schedule, StreamNeo removes the need to leave that computer on by turning an uploaded video into a YouTube live stream. That addresses the ongoing-operation problem, not the composition decision: prepare and check the landscape layout before you upload it.
Before committing, compare the operating options on the pricing page. When the file and channel are ready, start free — 24-hour trial, no card.
FAQ
Can I fit a 9:16 video into a 16:9 YouTube Live output without stretching it?
Yes. Scale it proportionally until it fits inside the landscape canvas, then use the remaining side space for black padding, panels, or a background treatment. Fitting preserves the complete portrait image; it does not fill the whole landscape width.
Does cropping keep the whole portrait frame visible?
No. Cropping enlarges the image to cover the landscape frame and removes some of the portrait image, commonly from the top and bottom. Check that no face, caption, or other important detail falls outside the crop before choosing it.
Is a blurred background required by YouTube?
No. It is one possible composition created by processing a second copy of the source behind the fitted portrait image. Black side areas or designed panels are also valid layout choices; choose based on what suits the programme and remains readable.
Is one landscape FFmpeg output also a vertical YouTube stream?
No. A single landscape composition is one output. YouTube documents a separate workflow for vertical live output, including vertical layout choices and, for third-party dual streaming, a separate vertical stream key; check the current Live Control Room guidance if you need both formats.