Quick answer: A gigapixel panorama is just a very disciplined multi-row panorama: shoot a grid of overlapping frames with a longer lens, let CPFind’s multirow alignment and the optimizer solve the grid in Hugin, raise Enblend’s memory cache for the render, make sure you have several times the final file size free as temporary disk space, and queue the job in PTBatcherGUI so it can grind away unattended. The stitching pipeline is identical to a small panorama — only the discipline and the logistics change.
Plan the grid before you shoot
Gigapixel images are made with focal lengths that would be absurd for a single-row panorama — moderate telephotos that see a small slice of the scene per frame. Work backwards from the print or the zoom experience you want: decide the final pixel dimensions first, then the focal length and grid follow. A 100-frame grid at 25–50% overlap is unremarkable; large projects run to many hundreds of frames in regular rows and columns. Regularity is the point: a pano head with click stops, or careful manual stepping, so every frame has neighbours on all four sides. Missing one frame leaves a hole that no amount of processing fills, so overlap generously and shoot a few spare frames at the edges. The camera discipline — manual exposure, focus and white balance, rotation around the entrance pupil — is the same as for any Hugin set, and is laid out in our shooting guide.
Shoot the grid
Work row by row, in a consistent serpentine or a strict left-to-right pattern you can keep track of, and keep the camera level as the rows stack. Telephoto frames have shallow depth of field and show focus errors cruelly at gigapixel zoom levels, so focus carefully once — on a mid-distance subject — and lock it. Watch the light: a grid takes minutes to shoot, and moving clouds change the scene between the first and last frame, printing a patchwork of brightness into the sky. On changeable days, shoot faster, or accept the sky as it is and rely on the photometric correction to smooth the steps. Bracket only if you truly need the dynamic range; brackets multiply the frame count and the render time by the same factor.
Focus and depth of field at telephoto lengths
Telephoto frames magnify focus errors as surely as they magnify the scene. A gigapixel viewer invites the audience to zoom to 100%, where a frame focused a metre too close announces itself as a soft patch in the mosaic. Focus once, deliberately, at a distance that keeps both the subject plane and the important detail inside depth of field at your chosen aperture — then lock focus and do not let the camera refocus between frames. If the scene has genuinely near and far layers, a mid-distance focus at a moderate aperture is the compromise; focus stacking a grid multiplies an already large job beyond reason. Check the first and last frames of the shoot at full magnification on the camera before you pack up: focus drift across a long grid is a reshoot, not a repair.
Align a grid in Hugin
Load the whole set into one project. This is the workload CPFind’s current default was designed for: since 2024 it defaults to the multirow strategy, matching each image against its grid neighbours instead of every image against every other, which keeps alignment time sane on hundreds of frames (the older all-pairs behaviour remains available as an option). The optimizer then solves the grid as one rigid puzzle — and rigidity is your friend: if the shooting was regular, misfits show up as isolated bad pairs you can inspect in the control-point editor, rather than global warping. Calibrating the lens beforehand pays off disproportionately here: a small distortion error, repeated across a hundred frames, becomes visible drift at the far side of the grid. Our lens calibration guide shows how to measure your lens’s distortion parameters once and reuse them.
Memory, cache and temp disk: the real bottlenecks
Rendering a gigapixel panorama is an exercise in moving data. Enblend blends the remapped frames in strips and keeps a memory cache — by default 1 GB — which you can raise in the preferences for very large jobs so it swaps less. Temporary disk space is the usual hard limit: the intermediate remapped images (nona writes one per source frame, at output resolution) plus the final file can require several times the size of the finished panorama, on top of your source photos. Put the project and its temp files on your fastest, emptiest drive, and check free space before you start, not after hour three. RAM pressure shows up as a render that crawls rather than fails — if the machine is swapping, a bigger Enblend cache makes it worse, not better; balance the cache against the memory nona and the operating system need. GPU remapping in Hugin remains experimental, so plan on CPU time and patience.
Queue it and walk away
Gigapixel renders are the textbook case for the batch queue. Save the aligned project, add it to PTBatcherGUI, and let it process unattended — overnight is a legitimate schedule for a big grid. The full queue workflow, including running several projects back to back, is in our batch stitching guide. When the render completes, inspect the output systematically at 100%: sweep the seams in a grid pattern of your own, because a single misaligned frame hides in a gigapixel file like a typo in a novel. If one frame is wrong, fix that frame’s control points in the saved project and re-render — the project file is the asset; the render is a printout.
Output size planning
Decide the deliverable before rendering, because re-rendering is the expensive step. A web zoom viewer wants tiles derived from a big master; print wants the full-resolution TIFF at final size. Render the master once, at the full solved resolution, as a TIFF, and derive everything else from it. Resist rendering at a guessed “probably enough” size: with the project saved you can re-render, but a fresh full render is hours, while a downscale from the master is minutes. The general render settings workflow is the same as in our step-by-step tutorial — only the numbers get bigger.
After the stitch: upscaling for print
One expectation to set before you fall in love with the render button: Hugin outputs your panorama at the resolution your source photos support — it assembles pixels brilliantly, but it does not invent new ones. When a gigapixel stitch is destined for a very large print, a gallery wrap, or a client who keeps asking for ‘a bit bigger’, the standard finishing step is a dedicated AI upscaler, and the one most panorama shooters use is Topaz Gigapixel: it enlarges the finished TIFF with detail reconstruction that plain interpolation cannot match, and it is far better applied once, to the flat master, than to individual frames before stitching. Topaz lists its Personal plan at $149/year ($12/month billed annually) on its site (checked 21 October 2026) — see Topaz Gigapixel (affiliate link — we’re reader-supported; see our disclaimer). Upscaling is a finishing luxury, not a stitching requirement: get the master right first, and only then decide whether the print size truly needs more pixels than you captured.
FAQ
How many photos does a gigapixel panorama need?
It depends on the lens and the target size. With a moderate telephoto and 25–50% overlap, a gigapixel-class result typically means one to several hundred frames in a regular grid. The frame count follows from your final pixel dimensions, not the other way round.
Why does my gigapixel render run out of space or crawl?
Intermediate files. nona writes a remapped image per source frame at full output resolution, and Enblend needs scratch room too — several times the final file size is a realistic temp-space budget. A render that crawls is usually swapping; check RAM and the Enblend cache setting.
What Enblend cache should I set for a very large panorama?
The default cache is 1 GB. Raising it in the preferences helps Enblend swap less on big jobs, but only up to the free RAM you actually have — an oversized cache forces the operating system to swap instead, which is slower.
Can Hugin use my GPU to render faster?
Hugin’s GPU remapping (via nona) is experimental and not a reliable production path. Plan gigapixel renders around CPU time, temp disk and the batch queue.
