720p source to 9:16: a 404x720 strip, upscaled 2.67x to 1080x1920

If you crop a 1280x720 clip to 9:16, you keep a strip only 404 px wide and 720 px tall. Scaling that to 1080x1920 enlarges it 2.667x on each axis, so every source pixel covers about seven output pixels. The picture will look soft. The crop itself is correct and cheap; the problem is the starting resolution, and no filter adds real detail.

The numbers

The width fraction is the same as for any 16:9 source, 0.3164, because it is a ratio. The pixel result is what differs: 0.3164 x 1280 = 405.0, which the compiler rounds down to the even value 404. The strip is 404x720, and 1080 / 404 = 2.673, 1920 / 720 = 2.667.

Three ways to get a sharper vertical file

First, do not crop at all: leave the 720p frame whole at the center of a 1080x1920 canvas with Timeline fit: "blur". The 1280x720 picture then scales to 1080 px wide (a 0.84x reduction, so it stays sharp) and occupies 608 of the 1920 rows, with blur above and below.

Second, crop and accept the softness for a short clip with a big, simple subject, such as a talking face. Soft video reads better than soft text, so keep any small text out of the source and add it later as captions.

Third, re-shoot or re-generate at a higher resolution, since sources that produce vertical frames natively avoid the crop entirely. Sume's Timeline renders 1080x1920 by default, but it composes the pixels you give it.

# 1) free contract check, no job, no charge
curl -X POST https://api.sume.com/v1/video-filter/check \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/webcam720.mp4",
    "ops": [{"op":"crop","x":0.3418,"y":0.0,"width":0.3164,"height":1.0}],
    "filtergraph": "scale=1080:1920,setsar=1"
  }'

# 2) same body to /v1/video-filter, with an Idempotency-Key header, to encode

Decision rule

At the documented public rate the encode is $0.02 per job and the check is free; confirm the live rate in GET /v1/catalog. The clip must already be a media.sume.com file of your workspace, so import it first with POST /v1/media-imports, and send an Idempotency-Key on the encode.

• Source under 1080 px tall: expect an upscale if you crop. Prefer blur-fit when the whole frame matters.
• Source 1080 px tall or more: crop and scale are safe; the strip is at least 608 px wide.
• Whatever you pick, check one frame at full size with video frames before captions go on.

Crop versus blur-fit, in numbers

The two routes treat a 1280x720 source very differently, and the numbers show why. With the crop, the picture is a 404x720 strip stretched to 1080x1920, which is an enlargement of about 2.67 times. With blur-fit, the whole 1280x720 frame is scaled to 1080 pixels wide, a reduction to 0.84 times, which gives a picture about 608 rows tall; the remaining 1,312 rows, about 656 above and 656 below, are blurred backdrop.

Whichever route you pick, finish with a proof still from video frames at the final size. Softness is easy to miss in a scrubbing preview and obvious on a phone screen at 1080 pixels wide, so look at one frame at 100% before you caption or upload.