The Dwarf 3 and Dwarf Mini take a lot of the guesswork out of Milky Way photography — the mount tracks, the app has a dedicated Milky Way mode, and the onboard computer stacks frames as it shoots. But "point and shoot" only gets you the first 60% of the image. The rest — clean calibration, a real capture plan, and a proper post-processing pass — is what separates a flat stack from a print-worthy shot. Here's the workflow I use.
Dark Sky Requirement
Wide-field Milky Way work is more forgiving of light pollution than deep-sky nebula imaging, but you still want Bortle 6 or darker — the wide lens's f/2.4 aperture pulls in a lot of ambient glow along with the stars. If you're shooting DSO targets with the Dwarf from your backyard, plan on driving further out for a Milky Way session.
Calibrating with Dark Frames
Don't skip this. The Dwarf 3's wide-angle lens needs its own set of dark frames, separate from anything you've shot with the telephoto lens — the Mini captures these automatically, but on the Dwarf 3 you do it manually: barrel closed, unit inside a bag or box, Darks mode in the app. Match the exposure time, gain, and ambient temperature to what you'll actually shoot that night — I keep a 15-second and a 30-second set on hand and re-shoot them if the temperature has drifted since the last session. These get subtracted from your light frames during stacking to kill hot pixels and sensor noise.
Dialing In Milky Way Mode
Switch lenses from the shooting-mode icon in the top right of the app and select Milky Way. To find the core, I use the telephoto lens to "Goto" a nearby bright object — the Antares/Scorpius region works well — then switch back to the wide lens once I'm roughly on target. The app's AR overlay (search "locate Milky Way" in-app) will show you the core's position and flag any obstructions before you've even started exposing.
Capture Settings
Exposure time depends on your alignment mode: 10–15 seconds on Alt-Az, 30–60 seconds if you're tracking in EQ mode — and drop back to 10–15 seconds regardless of mode if it's windy. Keep gain low, in the 0–60 range; the wide lens already gathers plenty of light at f/2.4, and pushing gain higher just blows out star color and adds noise. For frame count, I aim for 150–300 frames, which works out to roughly 25–50 minutes of total integration — enough to build a signal-to-noise ratio that holds up under aggressive post-processing.
Processing in Stellar Studio
The app's built-in Stellar Studio handles auto-stacking, star reduction, and even artificial diffraction spikes if you want them, plus basic brightness and noise adjustments in-app. For anything beyond a quick share, export as FITS or TIFF rather than JPG — you'll want the bit depth for what comes next. Pull the file off the device via USB, or through Google Photos set to "Original Quality." From there: Photoshop opens PNG/TIFF directly and benefits from a pass through the Camera Raw filter; Affinity Photo's Develop persona or a Curves/Levels adjustment layer does similar work; Siril can open the TIFF for color calibration or green-noise removal, but treat it as a finishing tool only — it can't restack a file that's already been stacked.
Recovering a Lost Meteor
The stacking algorithm treats any sudden streak as an outlier and throws it out of the final stack — which means a meteor that crossed your frame mid-session is sitting on the SD card, not in your export. To get it back: connect the Dwarf to a computer, point Siril at the raw frame folder, and switch the view from Linear to AutoStretch (raw frames are essentially black otherwise). Scroll through the individual frames looking for a single frame with a gradient streak, sometimes with a brighter flare at one end — rejected frames are usually named with "failed" in the filename, which narrows the search. Once you find it, run a Histogram Transformation stretch, optionally follow with background removal and an Asinh stretch, then export as 16-bit TIFF or PNG. Bring that into your stacked Milky Way image as a new layer sized to match, set the blend mode to Screen so only the bright streak shows through, and nudge it into alignment with a rotation transform if needed. A touch of motion blur along the meteor's angle smooths out pixelation, and a lasso-and-delete pass removes any stars that didn't align cleanly before you flatten.
Mosaics and Foreground Blending
For a multi-panel mosaic, always stitch the raw, unedited panels together before doing any color work — Photoshop's Auto-Align Layers (Perspective or Cylindrical projection), Affinity's New Panorama, or Lightroom's Photo Merge > Panorama all handle the sky's curvature better than manual alignment, which is really only viable on tighter telephoto frames. For the foreground, shoot it separately in Alt-Az mode with tracking off: a 30-second exposure at low gain (0–30) captures a sharp, naturally lit landscape without the hot-pixel noise a long exposure at higher gain would introduce. Drop that shot underneath your tracked sky stack and separate the two with a layer mask.
Building the Final Composite
The philosophy that ties all of this together is building the image from the ground up rather than editing a single flat file. Start from the untouched base stack, then layer in targeted passes on top of it: a star-removal pass (RC Astro's StarX or similar) to isolate nebulosity from the starfield, one or two duplicated "screen" layers with levels adjustments to pull out faint structure in the core and dust lanes, a faded pass of the original starfield blended back in for texture, localized color and curves work rather than one global adjustment, your recovered meteor layer, and a final pass that isolates only the brightest anchor stars so they don't get lost under everything else. It's more layers than a typical landscape edit, but each one is doing one specific job — which makes the whole thing easier to walk back and adjust than a single heavy-handed edit would be.