Upscale PowerDirector video — and tell a resize from real detail

In a consumer edit, "the video looks bad" usually means one of two different things, and they have different fixes. The first is a fit problem: a 4:3 camcorder clip in a 16:9 project, a vertical phone clip in a landscape property walkthrough, or a 720p shot inside a 4K sequence. The clip does not fill the frame, so the editor scales or crops it, and the result looks soft. The second is a source problem: the clip fills the frame correctly and is still soft, because the camera had a small sensor, the subject moved, the light was poor, or the phone compressed the file hard before you ever imported it. A fit problem is an editing decision and the editor's own scale is the right tool. A source problem is the case an upscaling model exists for, and even then only partly.

A full-size comparison of one frame settles which problem you are looking at, and it takes about a minute. Put the problem clip on the timeline and view a single frame at 100 percent, not fitted to the preview window, then do the same with a frame you know is good — a face, brickwork, a price tag on a product. If the problem frame is visibly made of fewer, larger pixels, you have a resolution gap. If it carries the same pixel count as the good frame but the detail is smeared, doubled by motion, or broken into blocks, those pixels were never clean, and enlarging them cannot make them clean.

The preview window is exactly why this check is worth doing. Every editor, PowerDirector included, shows a fitted preview smaller than the frame, so both interpolation and softness vanish at that size. Judging a clip in the preview is judging a thumbnail. Judge it at 100 percent on the monitor you will deliver for, and write down which of the two problems you actually have before you spend money on either fix.

Plenty of real jobs need a 4K file: a client's delivery spec, a listing platform's requirement, a screen in a venue, an upload target that only accepts certain formats. There is nothing wrong with producing that file. The trap is expecting the 4K label to add detail. A 1280 by 720 frame delivered at 3840 by 2160 has been enlarged to nine times its pixel count, and a 1920 by 1080 frame delivered at 4K is four times. Those extra pixels have to come from somewhere, and in an ordinary export they come from interpolation between the pixels that already existed.

So separate the two questions. Will this file be accepted, and will it play at the size the job needs? That is a compatibility question, and your editor's export settings answer it today at no extra cost. Will the picture hold more real detail on a large screen? That is a reconstruction question, and no export setting answers it. If the deliverable only has to be technically 4K — a lobby screen looping a slideshow, a platform that re-encodes whatever it receives — producing 4K from the editor is the correct and cheapest answer, and a paid pass would buy a difference nobody sees.

The upscale earns its place in the middle case: the file must be 4K and the frame will be watched large, on a television or a big monitor where the softness of an interpolated enlargement is obvious. Even there, be honest about what the pass buys. It makes the enlargement read as a plausible photograph instead of a stretched one. It does not turn a 720p capture into a 4K capture, and a viewer standing close to a large screen can still tell the difference.

The workflow that wastes the least is short. Finish the cut, then identify the shots that need help — usually a handful out of dozens — and export each one separately at its native frame size from the original media. Not the finished sequence, and not at the delivery size. A five-to-ten second slice of the worst shot is the right first export, because it shows what the model does with your footage before you commit a whole job to it.

Mild compression and a generous bitrate are the export settings that matter. Preset names move between versions, but the intent does not: spend the bitrate on detail, do not scale up on the way out, and export once. A file that has already been exported and re-exported carries a second generation of compression into the pass before the model has done anything at all. Audio is not carried through the request, so the soundtrack stays in the project and is re-attached when the enhanced clip returns.

The handover is either a file upload or a public HTTPS link. A supported video file up to 500 MB covers most single shots; a larger intermediate is better passed as a URL the service fetches itself. One source can run to 10 minutes and billing rounds up from a five-second minimum, so a long cut is processed shot by shot rather than submitted whole — which is also the cheaper way to do it.

The pass only sees the video file you upload. Titles, transitions, music, voice-over, colour adjustments and speed changes live in the project, not in that file, so nothing in the project is at risk while the pass runs. The risk appears when you put the result back, and it is almost always a placement mistake rather than a lost edit: the enhanced clip is dropped at the wrong start point, trimmed to a slightly different length, or left fighting the original clip underneath it.

The safe sequence is to import the enhanced file, place it on a video track above the original, align it to the same start frame, and mute or hide the original. Compare the two at full size across the whole clip, not only the frame you tested — a model that handles a face well can turn a background fence or a row of leaves into a shimmer. Only when that comparison holds do you swap it in: hide the original permanently, or point the clip at the new file so trims and transitions attached to it survive.

Two details decide whether the swap looks seamless. The frame rate has to match the project's, or the clip is conformed and motion can stutter or drift against the music it was cut to. And anything burned into the exported frames — burned-in captions, a logo bug, a look baked in at export — was part of what the model saw, so it can be sharpened or reshaped along with the picture. Keep those elements as project objects where the version allows it, and check any burned-in text by eye after the pass.

The commonest case where an upscale is the wrong purchase is a clip that is only soft because of how it was exported. If you previously exported at a low bitrate, sent the file through a messaging app, or re-imported an export and exported it again, the softness may be entirely a delivery artefact. Going back to the original media and exporting once at a proper bitrate costs nothing in credits and often recovers more than any model would. Try that first: it is free, and it is faster than a queue.

The second wrong purchase is a softness the model cannot fix. Missed focus and motion blur are missing information, not hidden information — a model will make a blurred frame look like a sharper blurred frame. Underexposure and sensor noise respond better to a denoise or enhancement pass than to an upscale, and running the two in the wrong order can bake noise into the result. And if the clip will only ever be watched small — a thumbnail, a phone feed at arm's length, a background panel in a listing — the improvement may be invisible at the size it is actually seen.

There is also a scheduling case. If a job ships today and the clip holds up on the screen it is made for, adding a pass adds a queue and a review step for a difference nobody will notice. An upscale is worth the credits when the frame will be watched large and the source is clean but small. Those two conditions together are the whole justification.

Billing follows the duration you upload, and it is quoted before you submit. On the Standard model, 1080p footage bills 540 credits per minute of source, which is 32,400 credits an hour, and the per-second rate is simply that divided down — a forty-second clip is billed as forty seconds, rounded up from a five-second minimum. The preflight quote is the actual cost for that clip, so there is no reason to guess: trim, upload, read the quote, and then decide.

Converted at the Creator yearly rate of $0.0093 per credit, a minute of 1080p Standard work is about $5.01 and an hour about $300. Creator's 106,800 credits cover roughly 197 minutes of 1080p Standard footage, or about 53 minutes at the 4K tier. The 4K tier of the same model bills more per second, and that premium buys a bigger frame for a delivery spec rather than more recovered detail.

The saving that matters is editorial. A twenty-five minute wedding cut may hold ninety seconds that genuinely need help — the ceremony recorded in dim light, the walkthrough room where the camera had to lean on digital zoom. A ninety-second job at those rates is a small fraction of submitting the same cut whole, and the difference is entirely in what you choose to export. Trimming in PowerDirector costs nothing and is the largest single lever on the bill.