Background removal, passport photos, Magic Eraser, 4× upscaling and high-accuracy Indian-language OCR — with models that run on this computer, not in a cloud. Unlimited use: no credits, no per-image fees, no subscription.
All nine tools are in the app today: HD background removal (with batch mode), passport photo maker, Magic Eraser, 4× upscale, Hindi / English OCR, scan cleanup, scanned table to Excel, click-to-select and auto-redaction. New models arrive as app updates.
Launch price (regular ₹1,499) · incl. 18% GST · one-time per PC · Windows 10 / 11
Online payment opens soon — you can buy by email today. How to buy
Every tool installs separately, so you download only what you use. Sizes are what the app downloads once; after that the tool works offline.
Cuts out people, products and objects in one click. The HD model gives sharp edges and hair detail; small PCs use a light model automatically. Batch mode cuts out a whole folder of product photos in one go.
Finds the face, crops to the official size, puts it on a white or blue background and saves it at the exact size and KB a form asks for.
Paint over an object, a stamp or a stain and it disappears, with the gap filled from its surroundings — offline on a normal 4–8 GB laptop.
Enlarges small photos, logos and scans 2× or 4× without the usual blur.
A modern on-device OCR engine for Hindi, Marathi and other Devanagari text, and for English invoices and statements, with optional recognisers for Tamil, Telugu and Urdu. (Basic OCR in 14 languages is free without the AI Pack.)
Flattens a curled page photographed on a phone and turns every page the right way up.
Reads the rows, columns and cell text of a scanned statement or invoice into a spreadsheet.
Click or drag on an object in a photo and it is selected — ready to cut out, recolour or remove.
Finds Aadhaar numbers (checksum-verified), PAN numbers and faces on scanned pages and marks them for true redaction.
After activation the app shows the size of each model before it downloads it. Install only the tools you want, and remove them any time to free space.
| Tool | With a graphics card | Without |
|---|---|---|
| Background removal (light) | 4.4 MB | 4.4 MB |
| HD background removal | 221 MB (124 MB if you run HD on the card) | 221 MB (200 MB with 6–8 GB RAM) |
| Passport photo maker | 25 MB | 25 MB (6.5 MB under 6 GB RAM) |
| Magic Eraser | 27 MB | 27 MB |
| 4× upscale | 9.3 MB | 9.3 MB |
| Hindi / English OCR | 20 MB | 20 MB |
| Scan cleanup | 37 MB | 37 MB |
| Scanned table to Excel | 26 MB | 26 MB |
| Click-to-select | 13 MB | 13 MB |
| All tools | about 370 MB (about 275 MB if you run HD on the card) | about 370 MB (about 350 MB with 6–8 GB RAM) |
Optional extras — the best-quality models big-LaMa (198 MB) and RealESRGAN_x4plus (64 MB), and the Tamil, Telugu and Urdu recognisers for the high-accuracy OCR (under 8 MB each) — bring the total to about 655 MB at most.
Cloud AI tools upload your photo or document to a server. The AI Pack does not.
Photos, scans and PDFs are processed in memory on your computer. No upload, no account and no telemetry — so ID cards, signatures and client documents stay where they are.
There are no credits and no per-image fees. Remove the background from five photos or five thousand: the price is the same one-time ₹999 (launch price).
No special hardware needed — a graphics card just makes it faster.
One-time per PC, prices include 18% GST, and MokshX Private Limited issues a GST invoice.
Online payment opens soon — you can buy by email today. Email [email protected] with the plan you want and your request code (in the free app press Ctrl+K, run Activate licence… and press Copy). We reply with UPI / bank-transfer details and a GST invoice, then email your licence key after payment — usually within one working day.
₹999 one-time
Regular price ₹1,499Launch
₹1,799 one-time
Regular price ₹2,499Launch · ₹1,998 if bought separately
₹99 7 days
Pro + AI Pack · exam season (February to July)
Several PCs? Business packs of Pro + AI Pack: 3 PCs ₹4,999, 10 PCs ₹14,999 — see all plans. Terms · Refunds.
Every model in the AI Pack is published by its authors under a permissive licence (Apache-2.0, MIT or BSD-3-Clause). MokshX re-hosts the files (a few converted to ONNX by MokshX from the official weights) and ships their licence notices in the app. What you pay for is the one-click install, the tuned tools around the models, and privacy — not exclusive model weights.
| Model | Used for | Licence | Source |
|---|---|---|---|
| Background removal — BiRefNet_lite (ONNX: fp32 1024 px and 768 px for the CPU, fp16 1024 px for the graphics card)Peng Zheng, Dehong Gao, Deng-Ping Fan, Li Liu, Jorma Laaksonen, Wanli Ouyang, Nicu Sebe (BiRefNet); ONNX exports by MokshX Pvt Ltd from the official MIT weights | HD background removal | MITNotice | UpstreamCode |
| Background removal — U²-Net-p (u2netp, 4.4 MB)Xuebin Qin, Zichen Zhang, Chenyang Huang, Masood Dehghan, Osmar R. Zaiane, Martin Jagersand (U²-Net); ONNX conversion distributed by Daniel Gatis (rembg) | Background removal (light model, low-RAM PCs) | Apache-2.0Notice | UpstreamCode |
| Inpainting — big-LaMa (ONNX fp32, 512×512)Roman Suvorov, Elizaveta Logacheva, Anton Mashikhin, Anastasia Remizova, Arsenii Ashukha, Aleksei Silvestrov, Naejin Kong, Harshith Goka, Kiwoong Park, Victor Lempitsky — Samsung AI Center (LaMa / big-lama); ONNX export by Carve (Carve-Photos/lama) | Magic Eraser, large areas (optional) | Apache-2.0Notice | UpstreamCode |
| Inpainting / object removal — MI-GAN 512 Places2 pipeline (ONNX)Andranik Sargsyan, Shant Navasardyan, Xingqian Xu, Humphrey Shi — Picsart AI Research (ICCV 2023) | Magic Eraser | MITNotice | UpstreamCode |
| OCR text detection — PP-OCRv5 mobile det (ONNX)PaddleOCR team, Baidu (PaddlePaddle) | Hindi / English OCR, table to Excel (finding text) | Apache-2.0Notice | UpstreamCode |
| OCR text recognition — Devanagari (Hindi, Marathi, Nepali…) PP-OCRv5 mobile rec (ONNX)PaddleOCR team, Baidu (PaddlePaddle) | Hindi / Marathi OCR (reading Devanagari) | Apache-2.0Notice | UpstreamCode |
| OCR text recognition — English/Latin PP-OCRv5 mobile rec (ONNX)PaddleOCR team, Baidu (PaddlePaddle) | English OCR, table to Excel | Apache-2.0Notice | UpstreamCode |
| OCR text recognition — Tamil PP-OCRv5 mobile rec (ONNX)PaddleOCR team, Baidu (PaddlePaddle) | Tamil OCR (optional) | Apache-2.0Notice | UpstreamCode |
| OCR text recognition — Telugu PP-OCRv5 mobile rec (ONNX)PaddleOCR team, Baidu (PaddlePaddle) | Telugu OCR (optional) | Apache-2.0Notice | UpstreamCode |
| OCR text recognition — Arabic script (Urdu, Arabic, Persian…) PP-OCRv5 mobile rec (ONNX)PaddleOCR team, Baidu (PaddlePaddle) | Urdu OCR (optional) | Apache-2.0Notice | UpstreamCode |
| Face detection — YuNet (2023mar, 227 KB)Shiqi Yu, Wei Wu, Hanyang Peng (YuNet / libfacedetection); packaged by OpenCV Zoo | Passport photo maker, face redaction | MITNotice | UpstreamCode |
| Portrait matting — MODNet (ONNX fp32 + int8)Zhanghan Ke, Jiayu Sun, Kaican Li, Qiong Yan, Rynson W.H. Lau (MODNet, City University of Hong Kong / SenseTime); ONNX by Xenova (Hugging Face) | Passport photo maker (portrait cut-out) | Apache-2.0Notice | UpstreamCode |
| Table structure recognition — SLANet_plus (ONNX)PaddleOCR team, Baidu (PaddlePaddle) | Scanned table to Excel (table structure) | Apache-2.0Notice | UpstreamCode |
| Table classification (wired / wireless) — PP-LCNet x1.0 (ONNX)PaddleX / PaddleClas team, Baidu (PaddlePaddle) | Scanned table to Excel (ruled / unruled tables) | Apache-2.0Notice | UpstreamCode |
| Document orientation (0/90/180/270) — PP-LCNet x1.0 (ONNX)PaddleX / PaddleClas team, Baidu (PaddlePaddle) | Scan cleanup (page orientation) | Apache-2.0Notice | UpstreamCode |
| Document dewarping — UVDoc (ONNX, PaddleX export)Floor Verhoeven, Tanguy Magne, Olga Sorkine-Hornung — "UVDoc: Neural Grid-based Document Unwarping", SIGGRAPH Asia 2023; trained & exported by PaddlePaddle (PaddleX) | Scan cleanup (flattening) | Apache-2.0Notice | UpstreamCode |
| Interactive segmentation — SlimSAM-77 uniform (ONNX, quantized)Zigeng Chen, Gongfan Fang, Xinyin Ma, Xinchao Wang (SlimSAM, NUS); Segment Anything by Meta AI; ONNX by Xenova (Hugging Face) | Click-to-select | Apache-2.0Notice | UpstreamCode |
| Super-resolution ×4 — Real-ESRGAN general v3 (SRVGGNetCompact, ONNX fp32, 4.6 MB) + weak-denoise twinXintao Wang, Liangbin Xie, Chao Dong, Ying Shan (Real-ESRGAN, Tencent ARC Lab / SIAT CAS; "Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data", ICCVW 2021); PyTorch→ONNX export by MokshX Pvt Ltd from the official .pth checkpoints | 4× upscale | BSD-3-ClauseNotice | UpstreamCode |
| Super-resolution ×4 — RealESRGAN_x4plus (RRDBNet 23 blocks, ONNX fp32, 64 MB)Xintao Wang, Liangbin Xie, Chao Dong, Ying Shan (Real-ESRGAN, Tencent ARC Lab / SIAT CAS; "Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data", ICCVW 2021); PyTorch→ONNX export by MokshX Pvt Ltd from the official .pth checkpoints | 4× upscale, best quality (optional) | BSD-3-ClauseNotice | UpstreamCode |
Listing a model here does not mean its authors endorse Moksh PDF Editor or MokshX. Model names are used only to identify the work they refer to. The full licence text and notice for each model are linked above and included in the app.
Yes. The models run on your PC, so there is nothing to meter: no credits, no per-image or per-page fees.
Only to buy the key and to download each model once. After that every AI tool works offline.
All of them: HD background removal (with batch mode), passport photo maker, Magic Eraser, 4× upscale, high-accuracy Hindi/English OCR, scan cleanup, scanned table to Excel, click-to-select and auto-redaction are in the app today. New tools and model improvements arrive as app updates.
No — they are separate. Pro adds the office tools (bank statements to Tally, GST e-invoices, batch unlock and recipes, compare, mail merge, stamps). Buy Pro + AI Pack to get both with one key.
Everything you installed keeps working for good. If you want the new and improved models that come after your first year, AI Updates cost ₹499/year — optional.
If it runs Windows 10 or 11 (64-bit) with 4 GB of RAM or more, yes. Smaller PCs get lighter models automatically, and a graphics card makes everything faster. Try the free app first — it is the same program.