NOTICE — Super-resolution ×4 — Real-ESRGAN general v3 (SRVGGNetCompact, ONNX fp32, 4.9 MB) + weak-denoise twin
Model id: upscale-realesrgan-general (v1)   Redistributed by MokshX Pvt Ltd as a downloadable component of Moksh PDF Editor.

Authors / copyright: 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
Licence: BSD-3-Clause
Licence text: https://raw.githubusercontent.com/xinntao/Real-ESRGAN/master/LICENSE
Upstream: https://github.com/xinntao/Real-ESRGAN/releases/tag/v0.2.5.0
Source repository: https://github.com/xinntao/Real-ESRGAN

Licence evidence recorded at publish time:
  - Real-ESRGAN LICENSE (https://raw.githubusercontent.com/xinntao/Real-ESRGAN/master/LICENSE, fetched 2026-10-09): "BSD 3-Clause License" / "Copyright (c) 2021, Xintao Wang" / "All rights reserved." — full text reproduced in NOTICE.txt
  - GitHub repository xinntao/Real-ESRGAN "About" / files navigation label: "BSD-3-Clause license"
  - The .pth weights are release assets of that same repository (no separate weights licence is stated anywhere in the repo, README, docs/model_zoo.md or the release notes), so they are distributed under the repository BSD-3-Clause licence
  - Architecture code realesrgan/archs/srvgg_arch.py lives in the same BSD-3-Clause repository (copy kept in staging models/upscale-realesrgan-general/_meta/_srvgg_arch.py next to _LICENSE.Real-ESRGAN.txt)

Files (SHA-256):
  3f55a7665b409d43f7d4b796f31130ddcdea60bb4d4c8b62ea10dfb1f8479dec  realesr-general-x4v3.onnx  (4866428 bytes)
  7e151549a60f101dfc1dadfc68294063d341e0a010f80ca5ada77f8ae1b9618c  realesr-general-wdn-x4v3.onnx  (4866428 bytes)

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the conditions of the BSD 3-Clause License (see licence URL above) are met, including retention of the above copyright notice.
These ONNX files are a format conversion of the unmodified upstream weights (see sourceCheckpoints in the manifest for the .pth SHA-256s); the network definition was transcribed from the upstream architecture source.

Full licence text of Real-ESRGAN (verbatim from the licence URL above, fetched 2026-10-09):

BSD 3-Clause License

Copyright (c) 2021, Xintao Wang
All rights reserved.

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:

1. Redistributions of source code must retain the above copyright notice, this
   list of conditions and the following disclaimer.

2. Redistributions in binary form must reproduce the above copyright notice,
   this list of conditions and the following disclaimer in the documentation
   and/or other materials provided with the distribution.

3. Neither the name of the copyright holder nor the names of its
   contributors may be used to endorse or promote products derived from
   this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

Notes: Exported 2026-10-09 by MokshX from the official v0.2.5.0 release assets (no ONNX is published upstream); checkpoint state-dict key "params". Upstream docs/model_zoo.md: "A tiny small model (consume much fewer GPU memory and time); not too strong deblur and denoise capacity", "X4 (can also be used for X1, X2, X3)" — i.e. upscale ×4 then downsample for smaller factors. Smallest/fastest upscaler; recommended default for the app (≈15× less compute than RealESRGAN_x4plus). Output dims are named height_x4/width_x4 on purpose so shape tools do not assume output == input size.

Generated 2026-10-09 by mokshx-models-staging/tools/make_manifest.py (same layout as scripts/publish-models.mjs)