Mini-Film: Open-Source Tool Beats Shopify AI for Batch RAW Processing

Everyone’s obsessed with the AI generate button. Watch any demo of a product photography tool, and the conversation inevitably pivots to the quality of the generated backgrounds. “Look at this lifestyle scene,” they say. “It preserved every stitch on that denim jacket.” But there’s a silent killer hiding in your workflow long before you touch any AI slider. It’s the import stage.

Your studio just shot a large batch of RAW frames for a new collection. You fire up your tool, drag the folder into the queue, and wait. Your GPU sits idle while your CPU chokes on decoding each proprietary camera file one by one. That “seconds per image” promise evaporates when you’re staring at a loading bar for a long time before editing even begins.

Mini-Film is an open-source batch processor built specifically to solve this pre-generation logjam. It ingests thousands of RAW files in parallel, surfaces culling decisions instantly, and hands SnabPic a clean, optimized set of images ready for bulk background replacement. No more waiting on your machine to catch up with your creative intent.

Speed Becomes a Liability

That creative momentum you build during a shoot evaporates the moment you open your editor. Lightroom users since version 1.0 know this pain intimately. You import hundreds of RAWs from a product line shot, and suddenly you’re watching a progress bar crawl across the screen. Catalog generation alone can eat a significant chunk of editing time. Capture One isn’t much kinder — its session-based approach loads faster but chokes on batch operations just as hard.

The bottleneck isn’t your camera. It’s your software’s single-threaded architecture. Most RAW processors were designed when 20-megapixel sensors were the ceiling. Today’s cameras push 45, 50, even 61 megapixels per frame. Your machine struggles because these tools never anticipated four-digit import counts from a single ecommerce session. You end up culling in passes: load fifty images, mark rejects, load another batch. That workflow doubles your review time at minimum.

Mini-Film sidesteps this entirely. It doesn’t build catalogs or sessions — it reads your file system directly and processes in parallel threads tuned to whatever CPU and GPU combination sits on your desk. The difference between waiting and working becomes measurable in seconds instead of minutes. That saved time goes directly into what matters: getting consistent product shots ready for SnabPic’s bulk background replacement pipeline, not staring at loading spinners while Lightroom sorts out its metadata index.

The RAW Workflow Nobody Optimizes

That saved time evaporates fast when you’re still waiting on JPEG proxies. Lightroom Classic renders 1:1 previews from Canon CR3 files at a noticeable delay per image on an M2 Mac. Multiply that by a hundred frames from a pack shot session. Suddenly a lot of idle thumb-twiddling appears out of nowhere.

Most photographers work around this by culling first, rendering proxies second. It’s a sequential trap that wastes half your editing time before you’ve even seen a single final image. Darktable offers an alternative here, but its caching layer has quirks. The database backend can bloat past a large amount of storage after processing many raw files, forcing manual purges through darktable-dbdelete to reclaim space.

The core problem is architectural: raw processors decode full sensor data for every thumbnail generation because they assume you might zoom to 100% at any moment. That’s computational overkill for batch review where the average decision happens at fit-to-screen magnification. Mini-Film sidesteps this entirely by generating lossy JPEG strips alongside the raw decode step during import.

The strip files run about 800 KB each compared to a 30 MB CR3 original—meaning the entire review library fits into memory rather than thrashing disk I/O on every scroll event. This approach changes the feedback loop fundamentally. You flip between images at sub-second latency instead of waiting for Adobe’s DNG converter to rebuild its cache.

The difference between smooth scrubbing and frame-by-frame stutter adds up across hundreds of files in measurable hours lost per week, not seconds saved per image.

The Core Architectural Difference

Mini-Film reads only the pixel data and discards everything else until export. The results are stark on real hardware. A 50MP folder with 300 RAWs takes a while to render full-resolution previews in Finder or Explorer. Same folder in Mini-Film: much faster. That’s not a optimization trick—it’s a fundamental design decision about what to prioritize. OpenSourceRAW developer Marcus Chen documented this bottleneck last year.

His tests showed embedded XMP sidecar parsing alone consumes a noticeable delay per file on typical SSD-backed systems. Scale that across hundreds of images and you’ve lost minutes before seeing a single frame. Mini-Film queues metadata extraction as a background process that runs only when the viewer is idle. You browse immediately; the database populates later. It reverses the traditional order from “process everything, then show me” to “show me now, process later.” The tradeoff is invisible for most workflows.

GPS coordinates and lens details don’t affect your crop or exposure decisions during selection rounds. They matter only when Lightroom needs keywords for archiving or stock agencies demand complete EXIF sheets before uploads. For tethered shooting sessions this compounds dramatically. A photographer firing many frames per hour using Capture One sees each image appear in Mini-Film roughly significantly faster than waiting for their primary catalog software to finish reading every embedded field first time around.

You pick selects faster when they aren’t hidden behind unnecessary overhead disguised as thoroughness.

Batch Culling and Marking at Keyboard Speed

The system is dead simple. Key 2 assigns two stars for keepers. Key 3 marks picks worth editing further. Color red mapped to key 1 flags immediate rejects — corrupted files, misfocused shots, lighting disasters that will never survive post-processing. Yellow on key 4 tags tethered check frames requiring review against the original composition reference. A wedding photographer with many RAW files from a ten-hour shoot ran this workflow in our test environment.

She sat down at her MacBook Pro M3 Pro in the evening with coffee and a USB-C card reader attached through a Thunderbolt cable pulling images off a V90 SD card. Quickly she had all reject batches deleted, keeper files triple-checked against color overlays, and export queue already populating with a handful of final selects ready for Lightroom Classic ingest.

Muscle memory locks in after roughly thirty images worth of repetition. Your index finger learns key positions faster than any menu navigation ever could — the star and color system sits right where your left hand naturally rests during reviewing sessions spent pressing arrow keys to advance frames.

While scanning focus points at full resolution zoom levels around two hundred percent magnification for critical edge sharpness verification checks across frame corners. And subject eye regions simultaneously displayed inside split-view panel modes showing before-and-after rating status changes reflected instantly throughout entire sorted collection of remaining unrated images awaiting decisions.

The real trick lives in Mini-Film’s auto-delete threshold. Open config.yaml in any text editor before inserting your first card. Locate the hotkeys block near line 42. Set "delete_flagged" to Ctrl+Shift+X — a two-finger stretch that becomes muscle memory after 50 frames. Two blocks down: "auto_delete_threshold": "5". This tells Mini-Film to purge any frame flagged for deletion once it sees six consecutive keepers afterward. No confirmation dialog appears. No trash bin holds the files.

Why this matters: Adobe Bridge requires three clicks per rejected image — mark, confirm, then empty rejected files from the folder afterwards. At many frames per hour that is many extra mouse movements, each one breaking your visual rhythm between takes. Mini-Film cuts that to zero clicks on discardable frames during active shooting periods. You only open external editors on the images that survive past threshold five. Compare workflows side by side.

Traditional culling demands first pass through Bridge or Photo Mechanic at a second or two per frame just reading thumbnails and metadata tags, then a second pass deleting rejects while watching load spinners for each trashed file. Mini-Film shows full RAW previews in a fraction of a second per image on a USB 3.2 card reader connected to any recent MacBook Pro M-series machine.

Thunderbolt reads are even faster, depending on card write speed class V60 versus V90 rated SD cards showing noticeable gap during burst sequences over ten frames deep.

Ready the Final Output

Ratings and selections complete, you’re now staring at a curated subset of your original batch. The stars tell the story — three-star keepers, four-star selects, five-star hero shots. Most guides stop here. They assume you’ll export everything individually or drag files into another application manually. Mini-Film’s bulk flattening feature collapses your selected RAWs into a single TIFF sequence in seconds.

Handle to File > Export > Selected as Flattened Sequence and choose a destination folder on an SSD or fast external drive. The output lands as sequential 16-bit TIFFs stripped of individual RAW metadata but preserving camera body serial, lens focal length, shutter speed, aperture value, and ISO rating in each file’s EXIF data block.

Web-hosted services like Shopify Magic Studio automatically discard this metadata during upload — losing provenance that might matter for portfolio attribution or insurance records down the line. A large keep set flattens in roughly ninety seconds on any modern NVMe-equipped machine running Linux or macOS with enough RAM headroom above sixteen gigabytes.

Windows users should ensure their output directory sits on an NTFS volume formatted with default allocation size for predictable write speeds under sustained load. Your timeline now contains every usable shot from the session arranged chronologically without gaps where rejects once lived.

Each frame carries its original capture timestamp alongside your editorial decisions encoded permanently into the file’s metadata structure — accessible years later through any EXIF reader application installed on future hardware platforms nobody has designed yet. This sequence is ready for direct import into ComfyUI workflow nodes expecting sequential TIFF frames for video diffusion pipelines, or drop-in deployment inside Krita diffusion plugins requiring contiguous numbered image stacks for image-to-image batch processing across consistent camera origin files.

Sharing identical white balance values determined during initial RAW decode processing pipeline execution stages earlier in this workflow example scenario described step-by-step from SD card insertion onward toward final ready-for-AI-batch output format completion state.

This is the real win. You go from a long import to a quick cache generation. Your GPU sits idle less, your edits happen faster, and that creative momentum stays intact through the whole batch. Mini-Film solves the bottleneck no one talks about because it isn’t glamorous. No AI slider, no generated background, just raw speed where you need it most.