AI Vertical Video Generator

A vertical AI video is ready to deliver when the required subject and action remain visible through the whole used range, captions survive the destination preview, and the exported file has the intended dimensions. Selecting 9:16 in a generator is only the first step. Generate and download the shot in VidVid; use an external editor for reframing, subtitles, assembly and final export.

Updated: October 14, 2026. This guide covers delivery of a finished clip. For preparing the starting image, use the image-to-video guide.

Choose the crop by what must remain visible

Situation in the whole used range	Delivery choice	Reject the choice when
One subject stays inside the same portrait window	Fixed crop	A hand, product edge or required action leaves the window
One subject travels across the source	Keyframed reframe in an editor	Following it cuts off a second required subject or makes the camera move distracting
Two required objects cannot fit together	Fit the full source into a designed portrait background	The subject becomes too small to inspect on the intended screen
The shot cannot preserve the required action at a useful size	Change the shot or generate a portrait composition	A different prompt still fails the same visibility check

A static first frame cannot approve a moving crop. Scrub the entrance, widest action, change of direction and exit. If a person demonstrates a product, both the hand and the product may be required; centering only the face can remove the meaning of the shot.

How much does a landscape crop remove?

For a 1920×1080 source filling a 9:16 frame at its original height, the retained width is 1080 × 9 ÷ 16 = 607.5 pixels, about 608 pixels. That is 31.64% of the original width. A centered crop removes roughly 656 pixels from each side. This is geometry, not a measured quality result; a different source size or fit strategy changes the calculation.

Draw this window before deciding to regenerate. If the entire required action fits, editing may preserve the existing shot. If two subjects separate beyond the window, increasing export resolution cannot restore the removed content. A moving crop can follow one subject, but cannot make two distant subjects fit simultaneously.

A five-step delivery check

1. Lock the used range. Record the source file and in/out timecodes. Judge only the range you plan to publish, including handles needed for a transition.
2. Preview the crop through motion. Check required edges at the start, largest movement and end. Watch once at the intended display size, with sound off if the message must work without narration.
3. Place actual captions. Test the longest line and a two-line subtitle. Use the destination’s current upload preview to inspect controls and overlays. There is no single safe-area rectangle that guarantees every app, device and placement.
4. Inspect the exported file. Check dimensions, duration, frame rate, audio and caption spelling after export. Reopen that file, rather than approving only the editor timeline. A portrait file is not automatically an exact 9:16 file.
5. Approve a named version. Record the filename, crop method and caption version. A changed used range, crop, subtitle placement or source clip triggers another preview; an approval of the previous file does not transfer automatically.

A low-resolution draft can help assess framing. It cannot establish whether small text, material detail or the final upscale will survive delivery. Choose the final resolution using the actual placement requirements and inspect the export; this guide does not prescribe 480p as a universal publishing standard.

If the shot needs another generation

Select an aspect ratio supported by the chosen model in the studio, and inspect the displayed credit quote before submitting. Put the composition in the reference image where possible, then describe the movement that must stay inside it. Runway’s official image-to-video prompting guide distinguishes the input image’s composition from the prompt’s motion instructions. That is useful drafting guidance, not evidence that every VidVid model follows identical rules.

Example motion brief: The camera remains fixed. The centered product turns slowly, with both outer edges visible throughout. The hand enters from above, touches the lid, then leaves without covering the label. This is an untested template. Check the output instead of treating wording such as “visible throughout” as a guarantee. Add final labels and exact text in the editor when readability is a delivery requirement.

Before another paid generation, identify the failed condition: missing subject, clipped action, unstable shape or unsuitable camera path. Changing the model without naming the failure makes the next result harder to evaluate. Compare available models at the model directory; use the credit planning guide for budgeting rather than dated example rates.

Existing H3 files: inspect dimensions, not just the setting

The following files and run records were already published on this page. They are historical examples, not new October 14 tests or a controlled model comparison. A is an image-to-video landscape shot; B requested a portrait 9:16 setting; C is a square text-to-video shot. C does not establish which of three hooks performs best. The recorded times are individual runs, not a latency promise.

Open this clip in a new window →

Prompt we sent (as run): A slow macro push-in on a matte ceramic object, realistic reflections, black studio background, premium commercial lighting, the object keeps its exact shape, no deformation, 5 seconds.

Open this clip in a new window →

Prompt we sent (as run): Vertical 9:16 product hook, energetic but controlled camera move, subject fully visible in the first second, bright accent colour, shallow depth of field, object shape stays fixed.

Open this clip in a new window →

Prompt we sent (as run): Extreme macro of a brushed metal surface, slow lateral camera move, reflected light crawling across the material, no distortion, commercial product photography look.

Run	Mode	Settings	Output file	Submit → ready
A	image → video	5s · 768P	1344×768, 24 fps, 5.2 s, 726 KB	196 s
B	text → video	5s · 768P · 9:16	768×1344, 24 fps, 5.2 s, 1.0 MB	117 s
C	text → video	4s · 768P · 1:1	768×768, 24 fps, 4.5 s, 341 KB	83 s

The recorded B file is 768×1344: its width/height ratio is 4:7, not exactly 9:16. At a height of 1344, an exact 9:16 frame would be 756 pixels wide. Inspect the downloaded dimensions and explicitly choose crop or fit at final export. This arithmetic does not assess the clip’s visual quality. These records do not verify today’s credit price, model availability or engagement performance.

Copyable delivery record

Source file / version:
Used range (in–out):
Required subjects and action:
Crop: fixed / keyframed / fit / new shot
Target aspect ratio and export dimensions:
Caption language / version / longest line checked:
Export filename:
Destination preview checked / date:
Approved by / date:
Changes requiring recheck:

FAQ

Can I reuse one export across Shorts, Reels and TikTok?

Use the same source when appropriate, but inspect each destination preview separately. Caption placement and the used range may need another export.

Does requesting 9:16 guarantee exact export dimensions?

No. Inspect the downloaded file and final export. The historical B record on this page requested 9:16 and reports a 768×1344 output, which is 4:7.

Does VidVid perform this final editing workflow?

Use VidVid for generation and downloading the shot. Reframing, subtitle placement, assembly and final delivery checks in this guide take place in an external editor and the destination preview.

For multi-shot continuity, continue with the storyboard workflow. For a product demonstration, see the product video guide.