Convert, compress, crop, watermark — and generate, upscale, cut out, or erase parts of images with on-device AI. No uploads. All inference runs locally on Apple silicon via MLX.
Current: v0.4 (pre-release)
OtterPix is in active development. Check back here for launch updates.

Eight built-in tools, all offline, all free.
PNG · JPEG · TIFF · BMP · HEIC · WebP · AVIF
JPEG quality, PNG re-encoding, WebP / AVIF presets
Free angle or quick 90° flips
720p / 1080p / 4K presets or custom dimensions
Interactive canvas with draggable handles
Text watermarks with font, color, opacity, angle
Configurable error correction and size
Render PDF pages to JPEG / PNG at any DPI
Chain operations into reusable presets, batch-process whole folders
Every AI feature runs locally on Apple silicon via Apple's MLX framework. Weights are downloaded once and reused — no per-image API fees, no images leaving your Mac.
Text-to-image with Flux 2 Klein 4B. Reference images supported for restyling and multi-image composition.
4× super-resolution with Real-ESRGAN. Four variants: general, x4plus, anime, anime video.
Salient-object background removal with BiRefNet. Swin-T for speed, Swin-L for fine details on hair and complex edges.
Paint over unwanted objects, LaMa fills in the background. Iterative — paint, erase, paint more, erase more.
Final pricing is being finalized — basic tools will always be free.
All eight basic tools · Pipeline · Batch processing · Format conversions including HEIC / WebP / AVIF
Everything in Free, plus all on-device AI features: Generation, Upscale, Cutout, Eraser
Read and write: JPEG, PNG, TIFF, BMP, HEIC, WebP, AVIF. PDFs can be rendered to images at any DPI. Most operations preserve color profiles and EXIF orientation.
No. Every operation — including all AI features — runs locally on your Mac. OtterPix's sandbox entitlements grant network access only for downloading model weights from Hugging Face on first use of each AI feature, and for App Store subscription verification. There is no analytics SDK, no telemetry, no image upload.
Only for the one-time model download. After that, every AI feature works fully offline. Model sizes range from ~2.5 MB (smallest upscale variant) to ~6 GB (Flux 2 image generation). You choose which models to download from the AI settings panel.
Four: AI Generation (Flux 2 Klein 4B, Apache 2.0), AI Upscale (Real-ESRGAN ×4), AI Cutout (BiRefNet, with Swin-T or Swin-L backbones), and AI Eraser (LaMa big-lama). All inference runs on MLX, accelerated by the Apple Neural Engine and GPU on Apple silicon.
macOS 14 (Sonoma) or later, on an Apple silicon Mac (M1 or newer). Intel Macs are not supported because MLX targets the unified-memory architecture. For Flux 2 image generation we recommend 16 GB RAM or more.
Yes — the Pipeline feature lets you chain any combination of operations (convert, resize, crop, watermark, upscale, cut out…) into a single preset and run it across a whole folder of images.
Check back at otterpix.app or write to hello@otterpix.app — we'll reply with TestFlight and App Store availability as soon as they're live.