A focus-stacking GUI for macro entomology photography.
Record a short video of a specimen while rotating the lens's focus ring through the depth of the subject, drop it into StackAnt, and get back a single publication-ready composite in which the whole specimen is in focus — all without manually picking and stacking frames.
Pictured: a Ugandan ant specimen, stacked from a single manual-focus-pull video.
- Load a video (MP4/MOV/AVI/MKV) or an existing folder of images.
- Extract frames through
ffmpegwith optional decimation. - Automatic blur detection (Laplacian variance) flags unusable transition frames; a slider tunes the threshold live.
- Even-spaced decimation caps the set at a target frame count (~75 by default) before stacking.
- Stack with either an in-process Laplacian-pyramid backend (pure
Python/NumPy/OpenCV, the default) or
focus-stack— pick one with the method radio, let Auto choose by stack size and resolution, or hit Compare to run both and toggle between the results. Defaults work out of the box, with an Advanced panel for the rest. - Preview the result at a scaled size, draw a rectangle to inspect any region at 1:1, and toggle between the stacked output and a selected input frame for comparison.
- Export to TIFF (lossless) and/or JPEG (quality 60–100).
- Batch mode (
File ▸ Batch…): point it at a folder and it runs the whole extract → score → filter → stack → export pipeline on every video in sequence, writing each composite next to its source and skipping ones already done. - Keyboard shortcuts, persistent settings, cancellable subprocesses.
Two things StackAnt does that Helicon Focus and Zerene Stacker don't:
- Video-first. A manual-focus-pull video is often the fastest way to capture a specimen — in the field, or at a microscope whose stock software doesn't ship with focus-stacking (or where it's a paid add-on). StackAnt takes the video directly and extracts frames for you. Commercial stackers expect you to arrive with N already-separated image files.
- Automatic frame-quality filtering. Transition frames where the lens is still sliding are motion-blurred and useless. StackAnt scores each frame's Laplacian variance on ingest and rejects the bottom tail automatically, with a live slider for override. Other stackers require manual pre-culling.
It's also free and open-source (MIT), and the log panel records every subprocess command so any run is reproducible.
A few Ugandan ant specimens, each stacked from a single manual-focus-pull video:
 Ugandan ant 01 |
 Ugandan ant 02 |
 Ugandan ant 03 |
 Ugandan ant 04 |
- Python 3.10 or newer (tested on 3.12).
- ffmpeg on your PATH.
- focus-stack on your PATH (Petteri Aimonen's CLI — not any other "focus-stack" tool).
- Linux (primary) or Windows.
StackAnt shows a blocking dialog at launch if either tool is missing, with exact install instructions for your platform.
Debian / Ubuntu / Linux Mint
sudo apt install ffmpeg libopencv-dev libxcb-cursor0 build-essential git
git clone https://github.com/PetteriAimonen/focus-stack.git
cd focus-stack && make
# Install to either a system path (needs sudo):
sudo cp build/focus-stack /usr/local/bin/
# …or to your user-local bin (no sudo, assumes ~/.local/bin is on PATH):
mkdir -p ~/.local/bin && cp build/focus-stack ~/.local/bin/libxcb-cursor0 is required by PyQt6 ≥ 6.5 on X11. Without it the app
aborts at launch with "no Qt platform plugin could be initialized".
Note on OpenCL: on some machines focus-stack fails mid-run with a
CL_OUT_OF_RESOURCES error from the GPU driver. If that happens, open
the Advanced panel and check Disable OpenCL (GPU) — stacking
falls back to CPU-only processing (slower but reliable).
Windows
ffmpeg: download a static build from https://www.gyan.dev/ffmpeg/builds/ and add itsbin/folder toPATH.focus-stack: grab a release binary from https://github.com/PetteriAimonen/focus-stack/releases and put it in a folder onPATH.
git clone <this-repo-url> stackant
cd stackant
python3 -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -r requirements.txt
python main.py- Open a video (or an image folder) from the File menu or the Controls panel.
- Pick a decimation factor if the video is long (e.g.
3keeps every third frame), then click Extract Frames. - StackAnt scores each frame, auto-picks a blur threshold, and flags unusable frames in red. Nudge the slider or double-click (or press Space) any frame to override.
- Pick a stacking method (Pyramid, focus-stack, or Auto) and click Stack Frames. The chosen backend runs in the background and its output streams to the collapsible log panel; Compare runs both and shows the two results with a view toggle.
- The composite appears in the preview pane. Drag a rectangle to see a 1:1 detail. Toggle between the stacked result and any input frame with the button above the preview.
- Export to TIFF and/or JPEG with a single click. Output folder opens automatically on success.
To process many videos unattended, use File ▸ Batch…: it takes the current settings once and runs the full pipeline on every video in a folder, writing each composite alongside its source.
pip install -r requirements-dev.txt
pytest # unit + headless-smoke tests
ruff check stackant testsThe codebase is split cleanly between pure-logic modules (frame_filter, stacker, exporter, frame_extractor command building) and Qt UI widgets, so most logic has non-Qt tests.
Shipped so far: v0.2 — Laplacian-pyramid fusion (a pure-Python
stacking method alongside focus-stack, same algorithm family as Helicon
"Method C" and Zerene "PMax") and v0.3 — Batch processing (queue a
folder of videos and run the full pipeline on each). Next up:
- v0.4 — Polish: optional input downscaling for memory-constrained integrated GPUs, reproducibility manifest on export, Windows end-to-end validation.
- v0.5 — Bundled installers so end users don't need a Python toolchain.
- v1.0 — Stable release once everything above is shipped and verified without caveats.
See CHANGELOG.md for details.
Bug reports and pull requests welcome. Please include the contents of the log panel (use the Copy log button) when reporting a problem.
MIT — see LICENSE.