How Siril Handles Deep-Sky Lucky Imaging and Large Frame Sets

Updated

Siril is built to process deep-sky lucky imaging natively: it reads SER video files directly, runs its algorithms across all CPU cores, and can handle more than 60,000 frames. Registration can be switched between all-stars, single-star, and nebula-body modes so alignment still succeeds when stars are faint. If you work with SER or FITS-Cube data, each preprocessing step adds only one extra file to your working directory.

Why SER support matters for lucky imaging

Lucky imaging means capturing thousands of very short exposures and keeping only the sharpest frames, so the format you record in determines how much conversion work sits between capture and stacking. Siril processes SER videos natively, so you can feed the capture file straight into preprocessing instead of splitting it into individual frames first.

The same applies to FITS-Cube data. According to the Siril site, taking advantage of the SER format or FITS-Cube means only one more file is created for each preprocessing step — the intermediate results stay compact rather than exploding into one file per frame.

Handling very large frame sets

Siril's algorithms use all cores of the computer, which the site describes as making it possible to process more than 60,000 frames "with no effort." Two practical consequences:

  • Parallel stacking. Each stacking method runs in parallel, so the combine stage scales with your CPU rather than running single-threaded.
  • Compact intermediates. Because SER and FITS-Cube inputs produce one output file per step, a 60,000-frame run does not leave you managing 60,000 intermediate files.

The limiting factor is usually disk speed and RAM rather than the frame count itself, so keep the SER file and the working directory on fast storage.

Choosing a registration method

Siril lets you adapt the registration method to the case, which is the main reason alignment rarely fails on difficult data:

Method Use when
All stars The field has plenty of usable stars and you want the most robust global alignment
One star You want to lock onto a single reference point, e.g. a bright central star
Nebula body Stars are faint or sparse, but the nebula itself has enough structure to align on

Switching to nebula-body registration is the usual fix when a faint-star field refuses to align under star-based methods.

Where this fits in the workflow

Lucky imaging is one stage inside a larger pipeline that Siril covers end to end:

  1. Calibration — remove unwanted signal and pattern from RAW frames using bias, dark, and flat master files.
  2. Registration and stacking — align frames and combine them to improve signal-to-noise ratio, with outlier rejection cutting noise further.
  3. Enhancement — colour calibration, deconvolution, histogram stretching, and similar steps on the stacked result.
  4. Science — analyse star brightness variation and use the astrometry tools to identify objects in the image.

You can drive these steps automatically with dedicated scripts or run them manually if you want control over each stage.

Practical starting conditions

  • Download the current release from siril.org; the site lists version 1.4.4 (released June 17, 2026) with builds for Linux (AppImage, plus a third-party Flathub Flatpak), Windows 64-bit and portable, and macOS for both Intel and Apple Silicon. Note that the MINGW64 Windows builds are marked deprecated.
  • Record or convert your lucky-imaging capture to SER, or use FITS-Cube, to get the one-file-per-step behaviour.
  • If alignment fails on a faint field, switch the registration method before assuming the data is unusable.

If you publish work that used Siril, the site asks that you cite C. Richard et al., Journal of Open Source Software, 2024, 9(102), 7242, DOI: 10.21105/joss.07242.