The full feature set
Import and creation, vector and raster processing, the heat map preview, placement, material testing and repeat work, organised around the job rather than around the toolbar.
Getting started
The first hour produces a part, a machine profile and a parameter set every later job is emitted from
Install the macOS or Windows build, start the 14 day trial from inside the app, and connect the machine. At connect the controller-aware core reads $30 and $31 off the controller and scales every emitted power value into that machine's real range, then queries $110 and $111 to derive and validate the effective maximum feed rate. From there the sequence is short: bring artwork in, pick the processing mode against the physical result you want, judge the generated toolpath in the heat map preview colour coded by power, and stream it. The trial carries full feature access, the same as the top software tier. Every figure on this page belongs to the desktop application and the controller-aware core underneath it.
- macOS and Windows builds, with every update fetched from a signed version manifest and verified against a published SHA-256 checksum before it installs
- The machine profile is assembled from the controller's own registers at connect, so the dialect, the power scale and the effective feed ceiling are read rather than assumed
- The heat map preview draws the toolpath the core will actually emit, colour coded by power, so a setting is judged on screen and not on the workpiece
Six moves from a downloaded build to a finished part
The order is the argument. Each move settles a value the next one would otherwise have to guess at.
- Step 1
Install the build for your operating system
Take the macOS or Windows build from the downloads page and run it. From then on every update arrives through a signed version manifest, and each download is checked against a published SHA-256 checksum before it installs.
Why it matters: The update path you get on day one is the path still feeding the machine in two years. It is signed and checksum verified from the first install rather than from a later hardening pass.
- Step 2
Start the 14 day trial and bind the device
The trial starts from inside the app and carries full feature access, the same as the top software tier. The app mints a Device Setup Code carrying the device identifier, an Ed25519 public key, device name and platform, and verifies the returned RS256 activation and lease signatures, issuer, device identifier and public key hash entirely locally.
Why it matters: Laser machines sit on workshop benches, isolated segments and networks with no route out. Every licence state transition completes offline, so the network a machine happens to be on has no say in when work starts.
- Step 3
Connect the machine and let the profile assemble itself
Pick the profile that matches your laser and controller, or build a custom one. On connect the core reads $30 and $31 and scales every emitted power value into that machine's real range, then queries $110 and $111 to derive and validate the effective maximum feed rate. The dialect, the travel strategy and the work envelope live in that profile from that point on.
Why it matters: Travel strategy alone changes what leaves the app. On GRBL a travel move is emitted as G1 F rapid with S0 rather than G0, because GRBL rejects G0 carrying an F parameter. The core applies that itself, so error 20 never reaches the operator on job one.
- Step 4
Bring artwork in, or draw the job here
Import PNG, JPEG, BMP and SVG, or build it in the app: text, shapes, QR codes in square, rounded and dot module styles, and linear barcodes in Code 128, Code 39, EAN-13, UPC-A and ITF-14. An imported SVG is parsed into a hierarchical sublayer tree where every element is individually shown, hidden, nudged, scaled or made semi-transparent, with the original markup left unmutated and the edits merged at render time. Place it with the millimetre-accurate tools before a single setting is touched.
Why it matters: Barcode input is validated at entry rather than at the scanner, and the merged SVG result is exactly what the generator consumes, so the canvas and the machine agree on what the job is.
- Step 5
Choose the processing mode against the physical result
Line work and logos take the vector path, in outline or fill, with closed paths ordered by descending signed area so outer contours run before inner detail. Photographic and tonal work takes the raster path, in grayscale, threshold or dither, with three error models to pick from: Floyd-Steinberg, Atkinson, and Bayer on a 4 by 4 ordered threshold matrix.
Why it matters: On a laser that choice is a materials decision. Error diffusion leaves isolated single-pixel dots that a beam with a finite spot size will not resolve on a coarse substrate, while ordered dithering builds clustered structures that survive the same material.
- Step 6
Preview it, frame it, then stream it
The heat map preview shows the generated toolpath colour coded by power rather than the source artwork. Framing traces the extents in constant power mode clamped to a maximum of 10 percent, and a dry run walks the path with the beam off. Then stream it and watch live progress: bytes in flight are tracked against the controller's 128 byte receive buffer and the buffer is kept full rather than waiting for an acknowledgement per line.
Why it matters: Material and machine time are spent once. Every move before this one exists so that spend lands against the real toolpath instead of against a picture of the artwork.
What the first run leaves behind, and every later job is emitted from
Run it on the stock you actually sell into. The first hour is worth an offcut of real material, because what it produces is a machine profile, a material setting and a template that the tenth run is emitted from.
Why the order holds
Machine context is settled first, because the power scale and the feed ceiling come off the controller rather than out of an assumption. Artwork comes second, edited without mutating the original markup. The processing mode comes third, chosen against the substrate. The preview comes fourth, run on the generated toolpath rather than on the source image. Streaming comes last, with the stop and the recovery sequence already in hand before there is anything at stake.
- The machine profile carries the dialect, the travel strategy, the power scale and the work envelope, saved against the file, so a setting is recovered rather than rediscovered at the machine
- Material settings and templates are stored with the job, and repeat work lays out in step and repeat up to a 20 by 20 grid, so the tenth run is emitted from the same numbers as the first
- Recovery is a first-hour drill rather than a first-incident one: stop a throwaway part on purpose, run the recovery, and watch the machine come back to the origin you set, so the sequence is already familiar the day it interrupts a real job. What each stop level issues is on the software page
- Secrets sit in the operating system credential store from the moment the trial starts: macOS Keychain, Windows Credential Manager or freedesktop Secret Service, never a flat file
- Estimate calibration sharpens timing against how your machine actually moves, so it earns its place once a few real jobs have given it measured runs to learn from
- The assistant features are the single part that needs a live connection, and they sit outside the path that generates and emits motion
Where to go next
Controller paths and status
What each dialect is emitted as, what changes between them, and the exact status label carried on every path including the Ruida bridge.
Six first jobs, and what each one settles
Pick one for the first run. Each exercises a different part of the path and ends with a saved setting rather than a one-off result.
- A name or a short line of text in a vector fill mode: settles the origin, the power scale and the profile inside a minute of machine time
- An SVG logo: exercises the sublayer tree and the vector pipeline, including undirected edge dedup at 0.01 mm, Douglas-Peucker simplification at 0.05 mm and closed path ordering by descending signed area
- A QR code with a centre logo at 5 to 29 percent of symbol size: error correction is raised automatically to level H when a logo is enabled, because embedding a logo means deliberately destroying part of the symbol, and the test that counts is whether it reads off your material on your own scanner
- A photograph in a dither mode: shows which error model your substrate actually resolves, with percentile clipping at the 0.5th and 99.5th setting the tonal range before conversion
- A small tag laid out in step and repeat: proves the placement tools and the repeat path before a production run leans on either
- A material test grid: hands you the power and speed pair to write into the material setting, so the second run of that material starts from a number
The first good result should be a file, not a memory
Run one real piece on real material. A name, a tag, a logo, a QR code on the stock you sell into. What you want out of it is a physical result in your hand with a written setting behind it, rather than a tour of the interface.
The result is reproducible because nothing in the chain was guessed. The power scale came off the controller's own $30 register. The processing mode was chosen against the substrate. The toolpath was judged in the heat map preview before material was committed. Save that as a material setting and a template, and the run six months from now is emitted from the same arithmetic rather than re-derived by whoever is on shift.
A setting nobody wrote down is a setting somebody rebuilds at the machine later, usually on the day the job is already late. The core is built so the first good result survives as a file.
Status labels, read them before a workflow rests on a path
Confirmed
- GRBL, Marlin, Smoothieware and Generic G-code are supported paths, each with its own laser-on command, travel strategy, power scale, precision and streaming discipline
- The 14 day trial carries full feature access, the same as the top software tier, so material testing, templates, placement and the supported controller paths are all evaluated inside it
- Offline activation is a supported path rather than a fallback, so the trial is evaluated on the machine that actually runs the work even when that machine has no route out
Not confirmed
- Ruida runs as an external bridge and carries the alpha label, in the product and here. Validate it against the specific controller in front of you before a production workflow rests on it
- Arc fitting carries the Experimental label. Evaluate it against your own geometry and run the first jobs on the line segment path
First run questions, answered with the constants
How much do I need to know before the first job?
Enough to pick a machine profile and a processing mode. The core reads the machine's own $30, $31, $110 and $111 values at connect, so the machine context is established for you, and the heat map preview puts the generated toolpath on screen before anything is streamed. Templates, material settings, step and repeat and calibration are all sitting there for the moment the work asks for them.
What does the 14 day trial include?
Full feature access, the same as the top software tier. Vector and raster processing, the heat map preview, placement, material testing, templates, code generation and the supported controller paths are all live during it, so the evaluation runs on the real product.
Does it activate on a machine with no internet connection?
Yes. The app mints a Device Setup Code carrying the device identifier, an Ed25519 public key, device name and platform. The portal returns an activation package, and the app verifies the RS256 signature, issuer, device identifier and public key hash entirely locally. Nothing in that exchange asks the machine to reach the internet, so a laser on an isolated bench segment is an ordinary first install. Renewing the licence later, or standing it up on a different machine, runs the same way, and the key handling behind it is on integrations. One thing worth knowing before the first run: licence state expires if the system clock is set more than an hour before lease issuance.
Do I need to draw the artwork somewhere else first?
Only if you prefer to. Import PNG, JPEG, BMP and SVG, or build the job in the app: text, shapes, QR codes in square, rounded and dot module styles, and linear barcodes in Code 128, Code 39, EAN-13, UPC-A and ITF-14, each validated at entry rather than at the scanner.
When is the right time to run estimate calibration?
After a handful of real jobs. Calibration sharpens timing estimates against how your machine actually moves, so it works from measured runs. The register read at connect is separate and happens every session regardless, which is why power scaling and the effective feed ceiling are correct from the first connection.
What do I do if I stop a job part way through?
Use the stop system rather than the power switch, then run the recovery before anything else is touched. On job one the difference is the whole point: cutting power leaves the machine at an unknown position with the origin gone, so the job is set up again from scratch, while the recovery returns the controller to a known state and puts the user origin back where you set it, so the same job is simply re-run. Stopping a laser is easy. Getting it back to a coordinate frame you trust is the part worth doing once, deliberately, on the first job rather than on the late one. The stop levels and what each one issues are on the software page.
Where does the workflow go after the first job runs cleanly?
Into repeatability. Save the machine profile, the material setting and the template, run a material test grid on each new substrate, and lay repeat work out in step and repeat up to a 20 by 20 grid. The features and controller pages cover both at parameter level.
Start the trial, connect the machine, run a real part
Install the build for your operating system, start the 14 day trial from inside the app, and connect the machine so the core can read its registers. The first job answers what a specification only describes: how your own material reacts at the power scale the controller itself reports.
Last updated August 21, 2026