OCR documents, transcribe speech in 40+ languages, classify images, detect documents and barcodes, smart-crop photos, upscale with Core ML, colorize B&W. Every model runs on your device's Neural Engine — your content stays private.
Extract printed and handwritten text from photos using Vision.
Transcribe audio in 40+ languages using Apple's Speech framework.
Detect the language of any block of text.
Identify objects, scenes and concepts in any photo.
Find faces with bounding boxes and landmark points.
Auto-crop to the salient subject of the image.
Find document edges in a photo for scanning.
Read every common barcode and QR code type.
Generate QR codes from URLs, contacts, Wi-Fi credentials and more.
2×–16× upscale using SPAN, RealPLKSR, DRCT and Real-ESRGAN Core ML models.
Restore old or damaged photos with InstructIR, plus GFPGAN for face restoration.
Bring black-and-white photos to life with DDColor.
12 one-tap presets: Face Restore, Old Photo Fix, AI Denoise, JPEG Fix, Sharpen & Clarify, Low Light, HDR Effect, Auto Color, Tint B&W, AI Sharpen, Portrait and Color Pop.
Models like HVI-CIDNet (8 MB), DDColor Tiny (59 MB), InstructIR (64 MB), FBCNN (70 MB).
DRCT, DDColor, GFPGAN, BiSeNet for advanced edits.
Depth Anything, Real-ESRGAN x4, and full-size DDColor for maximum quality.
Yes. OCR uses Apple's Vision framework, which runs entirely on-device. No internet connection is needed and your photos are never uploaded.
Filemorph supports speech-to-text in 40+ languages including English, Spanish, French, German, Japanese, Korean, Chinese, Russian and Arabic, using Apple's on-device Speech framework.
Yes, the Core ML models are downloaded once on first use, then run locally on the Neural Engine for every subsequent conversion. Models are tier-based: small for iPhone, larger for iPad Pro and Mac.
No. After the initial model download, every AI operation runs on your device. Nothing about your content leaves your phone.
20+ AI tools running on your Neural Engine.
Download on theApp Store