Resolution and scaling¶

The Resolution setting on the Output tab decides the pixel size of the combined video. Its real job is to put every clip on one canvas so they can be joined at all — not to improve them.

ForgeAssembler resizes. It does not enhance.

Choosing 4K on 1080p sources gives you a genuine 3840×2160 file with exactly the detail the source had, interpolated to fill four times as many pixels. There is no super-resolution, no sharpening, and no detail synthesis anywhere in ForgeAssembler. If you want real enhancement, do it before you import — with a tool built for it, such as Topaz Video AI — and assemble the enhanced files.

Why a resolution setting exists at all¶

Concatenation needs every clip to share one frame size. A compilation of a 1920×1080 scene and a 1906×1080 scene cannot be stitched until both are the same shape, so each clip is scaled to fit the chosen canvas and padded to fill it exactly:

scale=<W>:<H>:force_original_aspect_ratio=decrease:flags=lanczos,
pad=<W>:<H>:(ow-iw)/2:(oh-ih)/2,
setsar=1

force_original_aspect_ratio=decrease means aspect ratio is never distorted — a clip is scaled until it fits inside the canvas, then centred, and whatever is left over becomes black bars.

Odd source widths produce thin bars

A 1906×1080 source on a 1920×1080 canvas scales to 1906×1080 and gains a 7-pixel black bar on each side. On a 4K canvas the same clip becomes 3812×2160 with 14-pixel bars. The bars are baked into the output. If your sources are an unusual width and you want no bars, set the resolution to source (below), or crop them before importing.

The scaler¶

Every resize uses Lanczos (flags=lanczos), not ffmpeg's default bicubic. Lanczos is sharper in both directions — downscaling a 4K clip to 1080p and upscaling a 720p clip to 1080p both come out crisper — and it costs nothing measurable next to the encode.

This applies to the video, to scaled overlay images, and to clip thumbnails.

It is still a resampler. A sharper interpolation is not new detail.

Choosing a value¶

Setting	Canvas	Use it when
`source`	Matches the first video clip in the project	All your clips are the same size and you want no resampling at all
`1080p`	1920×1080	The common case
`1440p`	2560×1440	Sources are genuinely 1440p or larger
`4k`	3840×2160	Sources are genuinely 4K
`uw_1080p` / `uw_1440p`	2560×1080 / 3440×1440	Ultrawide sources
`4_3_hd`	1440×1080	4:3 material
`3_4_hd` / `9_16_hd`	1080×1440 / 1080×1920	Vertical / phone-shaped output

Pick the resolution your sources actually are. Upscaling costs real time and disk for no visible gain:

A 13-minute 1080p compilation forged at 1080p was 812 MB at 8.2 Mbps.
The same compilation forged at 4K was 2.45 GB at 24.5 Mbps — three times the size, three times the encode, and not one extra pixel of detail.

source skips the guesswork: ForgeAssembler probes the first video clip and uses its dimensions as the canvas. Clips that already match are scaled by a factor of exactly 1 — the cheapest, cleanest path. Note that source needs at least one video clip to probe; a project of nothing but title cards has nothing to measure.

What about downscaling?¶

Downscaling is the case where the resolution setting genuinely helps. Taking 4K sources to a 1080p output is a real quality operation — Lanczos downsampling averages four pixels into one and the result is sharper and cleaner than the source viewed at that size. Mixed-resolution projects are normalised the same way, so a 4K clip and a 1080p clip end up visually consistent.

Frame rate¶

Frame rate is separate and works the same way: source takes the first video clip's rate, or you can force 24 / 30 / 60. Forcing a rate resamples the timing — it does not interpolate new frames, so raising the number does not make motion smoother.