Lucid
Copyright (c) 2026 Braedon Saunders

Lucid itself is licensed under the MIT License; see LICENSE.

This file records third-party work used by Lucid and its development tools.
The application includes the full notices in Resources/ThirdPartyNotices.txt.

================================================================================
NanoVSR — production reconstruction architecture
================================================================================

NanoVSR architecture adaptation
Source: https://github.com/filippawlicki/nanovsr
Lucid's production network adapts the RepVGG blocks and bilinear residual to
single-frame, pixel-unshuffled, 2x reconstruction without BatchNorm. All shipped
weights were trained by this project; no pretrained NanoVSR weights are used.

MIT License

Copyright (c) 2026 Filip Pawlicki, Marcel Kańduła, Marcin Pucek, Kamil Dobies

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.


================================================================================
SPAN — historical models and development tools
================================================================================

Source: https://github.com/hongyuanyu/SPAN
License: Apache License, Version 2.0.
The development code adapts the SPAN architecture for pixel-unshuffled codec
restoration. Earlier releases used locally trained SPAN weights. Lucid 1.0.0
ships the Nano-derived model described above. The SPAN license and attribution
are retained in Lucid/Resources/ThirdPartyNotices.txt and Tools/architectures.


================================================================================
Algorithms implemented from published descriptions
================================================================================

The filters in Lucid/Metal/DetailEnhancer.swift are written here from the
published descriptions of the following work. No code was copied from these
projects, and they are named because the ideas are theirs:

  * CDEF, the constrained directional enhancement filter, from AV1
    — Steinar Midtskogen and Jean-Marc Valin. The per-direction divisor table
      matches the one used by dav1d.
  * Temporal accumulation with neighbourhood clipping
    — Brian Karis (temporal anti-aliasing) and Playdead's shipped implementation.
  * Stochastic debanding
    — the approach used by libplacebo and mpv.
  * Contrast-adaptive sharpening
    — the lobe from AMD FidelityFX CAS.
  * Oklab
    — Bjorn Ottosson.
  * The deblocking filter
    — ITU-T Recommendation H.264, Tables 8-16 and 8-17, run without a bitstream.


================================================================================
Test material
================================================================================

Test clips are not distributed in this repository. Those referenced by
TestSite/ and by the training corpus generator are re-encodable from public
sources:

  * Big Buck Bunny and Sintel — (c) Blender Foundation, CC BY 3.0.
  * Netflix Chimera test sequences — CC BY, via media.xiph.org.


================================================================================
Trademarks
================================================================================

Lucid is not affiliated with, endorsed by, or sponsored by NVIDIA, Apple,
Microsoft or Google. RTX Video Super Resolution is a trademark of NVIDIA
Corporation and is named in the documentation only to describe, by comparison,
what this project does.