How do I convert a 16:9 video to 9:16 without losing the subject?

Upload it to Valmera and type “make it 9:16.” The agent finds the speaker’s face in the frames and aims the crop at it, re-aims at every camera cut, and fits shots whose content runs to the edges over a blurred backdrop instead of cutting them off. Review the preview, fix any shot in chat, then export.

How do I keep the speaker centered when cropping a video?

Ask for the crop and Valmera centers it on the speaker’s face rather than the middle of the frame, measuring again in every camera shot. To adjust it, say “move the crop left” or “aim at the guest from 2:10 to 3:40,” or drag the crop box in Studio’s framing panel.

Does auto reframe track a moving subject?

Not continuously. Crop positions are set per camera shot and per section, and they change as a cut, never as a pan. A presenter walking across one long shot is not followed: fit that shot, or set a new crop position from a given time.

What happens when two people share one shot?

Valmera doesn’t guess the speaker from face size. It fits the shot and flags it, then the agent sets a crop position for each speaker section using the speaker-labeled transcript and the frames. Check those sections in the preview.

Will a 9:16 crop of 1080p footage look soft?

It can. The crop window is 608×1080 pixels and the export scales it to 1080×1920, so it holds the detail of a 608-pixel-wide picture. Shoot 4K (the window is then 1215×2160) or frame wider when you plan to cut verticals.

Can I get 16:9 and 9:16 versions from one upload?

Yes. Export the 16:9 version, ask for 9:16, review it and export again. Earlier versions stay in the project history, and asking for the source ratio restores the original shape without undoing cuts or captions.

Is auto reframe free?

Creating an account and uploading are free. Editing requires a subscription. Exports are full-resolution H.264 MP4s rendered from your original file and end with a brief Valmera end card.

Auto Reframe Video to 9:16 Without Losing the Subject

To auto reframe a video to 9:16, upload it to Valmera and type “make it vertical.” The AI agent finds the speaker’s face in the frames, aims the crop at it and re-aims at every camera cut. Footage whose content runs to the edges, like gameplay, slides or a screen recording, is fitted over a blurred backdrop instead of being cut off.

It works the same way for 1:1 and 4:5. The agent checks its own preview before it replies, and you can move any crop with a sentence.

Make Your Video Vertical in One Request

Upload, type “make it 9:16,” review the preview. Creating an account and uploading are free. Editing requires a subscription.

Create an account →

See pricing →

How to Auto Reframe a Video to 9:16 in Valmera

1
Upload the horizontal video
MP4, MOV, M4V, MKV or WebM, up to 100 GB and 3 hours. Valmera indexes it once: faces and shot changes in the frames, plus a word-level transcript with speaker labels.
2
Ask for the shape
Type “make it 9:16” (or 1:1, 4:5, 16:9). Add what must stay visible: “keep both hosts” or “keep the whole game screen.”
3
Let the agent measure and choose
It aims the crop at the face, re-aims at each camera cut, and fits footage that runs to the edges over a blurred backdrop. A shot with two people gets a crop position per speaker section.
4
Review the preview
Watch every cut. Fix any shot in a sentence (“move the crop left at 0:42”) or drag the crop box in Studio’s framing panel and click Apply.
5
Export
Click Export in Studio for a full-resolution H.264 MP4. Need another size? Ask for it and export again.

PROMPTS TO PASTE

Make it 9:16 for TikTok and keep the speaker centered.

Aim the crop at the guest from 2:10 to 3:40.

Move the crop a little left in the opening shot.

It's gameplay: fit the whole screen, don't crop.

How Valmera Decides Between Crop and Fit

In auto mode, Valmera measures two things in the footage you kept. First, where the subject is: faces are found in the pixels and confirmed across up to about 180 indexed frames, so one false detection can’t steer the crop. Second, whether to crop at all: if the crop window would keep less than 55% of the picture’s detail, cropping would truncate the shot, so it fits the whole frame instead. An explicit request always wins: say “crop it” and you get a crop.

Footage	What auto reframe does	What to check
Talking head (one person)	Crops to the new shape, aimed at the face.	Hands, props or a product near the edge of the frame.
Multi-camera podcast or interview	Measures the subject in each camera shot and re-aims the crop at every cut. A shot with no clear face (a screen, a product) is fitted instead of inheriting the last crop.	Every cut lands on the right person.
Two people in one wide shot	Fits the whole frame and flags the shot. The agent then sets a crop position for each speaker section from the speaker-labeled transcript and the frames.	Each section shows whoever is talking.
Gameplay, screen recordings, slides, wide scenes	Fits the whole picture over a blurred copy of itself, so nothing is cut off. Or ask for the picture in a band across the middle, with room for a headline above and captions below.	Small text is still readable on a phone.
One person walking across a long shot	Holds one crop position for the whole shot. This is the limit: the crop never pans to follow movement.	Fit that shot, or ask for a new crop position from a given time.

Crop, Fit or Blur: The Four Modes

Mode (Studio control)	What it does	Best for
auto (default)	Measures each camera shot and picks a crop or a blurred fit.	Most videos. Start here.
crop (Fill)	Fills the new frame edge to edge, aimed at the subject. The sides are cut off.	Talking heads and close-ups you want big on a phone.
pad (Fit)	Fits the whole picture inside the new frame with black bars.	Archive or client footage whose composition must stay intact.
pad_blur (Blur)	Fits the whole picture over a blurred, enlarged copy of the same footage.	Gameplay, slides, screen recordings and wide scenes.

The blurred backdrop is your own footage, enlarged and blurred behind the fitted picture.

Fill, Fit and Blur are the same modes in Studio’s framing panel. Padding is always black bars or your own footage blurred; nothing is generated.

Keep the Speaker Centered When Cropping

A plain center crop only works when the speaker sits in the middle of the frame. A host framed on the left of a 16:9 shot ends up half outside a centered 9:16 window. Valmera centers the crop on the detected face instead and measures again in every camera shot. The crop changes position as a cut, never as a pan: a reframe on a cut reads as an edit, while a crop that drifts mid-shot reads as a slide.

A centered 9:16 crop misses an off-center speaker. Aiming the same 608-pixel window at the face keeps them in frame.

To adjust by hand, say it in chat. Crop positions are stored against the source footage, so they stay on the right moment after later trims or speed changes.

“Move the crop left” or “a little right in the opening shot.”
“Aim at the guest from 2:10 to 3:40” sets a position for that section only.
“Fit the whole screen, don’t crop” switches a shot, or the whole video, to a fit.
Or drag the crop box in the framing panel and click Apply.

Resolution: What a Vertical Crop Keeps

A crop doesn’t add pixels. From a 1920×1080 source, the 9:16 window is 608×1080, and the export scales it to 1080×1920, so it carries the detail of a 608-pixel-wide picture. A 4K source gives a 1215×2160 window, more detail than a native 1080×1920 vertical, and exports at 2160×3840. If you know you’ll cut verticals, shoot 4K or frame wider with the subject near the middle.

Target (from 1920×1080)	Crop window	Width kept	Export size
9:16	608×1080	31.6%	1080×1920
4:5	864×1080	45%	1080×1350
1:1	1080×1080	56.25%	1080×1080

The export keeps the source’s short side and is rendered from your original file, not the preview. Below HD, finals are composed on an HD canvas so captions and titles stay sharp; that adds no picture detail.

Limits

Auto reframe sets crop positions per camera shot and per section; it is not continuous motion tracking. The crop never pans, so a presenter who walks across one continuous shot is not followed. Shots with two or more people are fitted and reviewed, because face size doesn’t tell you who is speaking, and there’s no automatic two-speaker split-screen layout for 9:16. For footage that needs a crop following a moving object, such as sports or a speaker walking on stage, OpusClip’s ReframeAnything model tracks moving speakers and objects (opus.pro, checked October 6, 2026).

Related Workflows

Platform sizes and which ratio to pick are on resize video for TikTok, Instagram and YouTube. To go vertical for one section and back, use an aspect shift. To turn a long recording into vertical shorts, each reframed on its own, use the AI clip maker, as TikTok creators do. The glossary entry explains auto reframing in general, and the trim-and-reframe example puts a real source next to its final export.

Using Claude, ChatGPT or Cursor? The same auto_reframe and set_frame tools are available through Valmera's MCP server.

Creating an account and uploading are free. Editing requires a subscription — see plans.