No, direct overwriting is not appropriate; the README explicitly recommends extracting into a new folder and migrating only the specified user data and LUTs.

• The 2.1.1-to-2.1.2 upgrade instructions require closing Veyra and copying only runtime_local/user-data-2.0.0 and runtime_local/luts.
• The README warns against copying the old veyra.ini, the entire runtime_local folder, or mixed DLLs over the new package because stale components or startup failures may result.
• An in-place overlay using the same GPU flavor is possible, but it leaves stale components and is therefore not the preferred path.
• The new package must match the active GPU, and NVIDIA and AMD package contents should not be mixed during migration.

The appropriate decision is to retain the old version for rollback, run 2.1.2 from a new directory, and copy only the two explicitly listed data paths. The README does not guarantee compatibility if the entire old configuration is copied.

Yes, because the README explicitly puts live processing, screenshots, and offline export under the same settings system and provides a delivery-oriented queue.

• The Highlights section says that presets, List/Node sessions, PNGs, and video exports share the settings, reducing divergence between preview and final output.
• The new export workflow supports an ordered queue, trimming, MP4/MKV, multiple audio tracks, embedded subtitles, cancellation, and retry.
• The Architecture section states that the export queue stores a per-job settings snapshot, so queued jobs do not simply follow later global-setting changes.
• Export is an independent encoding and muxing path rather than a recording of the preview window, which is better for file production, although processing time and codec compatibility depend on the input and output specifications.

The project fits well when reusable enhancement settings, track/subtitle preservation, and multi-job export are priorities. The README does not promise lossless preservation for every subtitle format, codec, or container combination.

Yes, because Veyra’s List and Node modes are not separate algorithm stacks; they share the same EnhanceGraph engine and extend to preview, screenshots, and export.

• The Highlights section provides executable connections, independent NR and color parameters, separate List/Node configurations, sessions, and presets.
• Architecture explicitly states that List/Nodes share the engine and disconnected drafts do not execute, reducing accidental processing during graph editing.
• The graph includes multiple NR stages, single SR, color, and Video HDR, with connections to live preview or offline export.
• NR controls provide per-layer strength, Picture control, and parameters for hue, chroma, shadows, highlights, local compression, and temporal stability.

This fits advanced users who need different reusable processing plans for different sources. However, complex-chain behavior still depends on GPU, VRAM, input color space, and node order; the README gives no performance ceiling or automatic validation rules for arbitrary node combinations.

Yes, because the README’s main local validation used an RTX 5070, and the application explicitly targets Windows 11 x64 with DirectX 12.

• The NVIDIA package includes the RTX 50 NVIDIA-original path and exposes DLSS, XeSS, FSR, and RTX Video SR options.
• Version 2.1.2 adds true dual-pass RTX Video SR, separate second-pass controls, and VRAM-budget checks for high-resolution processing.
• Frame Generation includes full-frame deduplication, smoothed anchors, and optional VFG regulation, but the README also warns that FG can increase VRAM usage, latency, or motion artifacts.
• High resolution, layered NR, and FG all raise GPU and VRAM requirements, so being able to launch the application does not guarantee stable 120 FPS for every combination.

If the goal is minimum input latency in competitive games, the README does not provide RTX 5070 frame-time or latency measurements for specific games, resolutions, or FG settings.

It depends: the paths exist, but the README explicitly places RX 7000 RedDuke NR and parts of the FSR 4 capability in experimental or unverified territory.

• The AMD portable package is an official installation route. Version 2.1.2 adds experimental RX 7000 RedDuke NR, Small/Original modes, and per-layer scaling while retaining a separate RX 9000 path.
• The README also lists “FSR 4 ML hardware” among unresolved or unverified cases, so the presence of a menu option does not mean every AMD model is covered.
• Users must download the package matching the active GPU and must not mix runtime files from different vendor packages.
• Strong NR, super resolution, and Frame Generation increase GPU and VRAM pressure; the README also records some NR TDR and VRAM-growth reports.

It is reasonable to try if you accept model- and driver-specific validation. It is not appropriate to promise long-term stability, vendor certification, or a complete FSR 4 compatibility matrix from the README alone.

It depends: Veyra covers this workflow, but capture format, VRR, and audio-carrier behavior remain decisive external constraints.

• FrameSource accepts files, capture cards, PS5, and screen capture sources, while live preview, screenshots, recording, and export share the enhancement pipeline.
• Version 2.1.2 adds WGC high-refresh cadence, ready-immediate capture without FG, first-copy 60-to-30 handling, and Dolby/DTS carriers on generic PCM pins.
• The Architecture section specifies a native D3D12 video window, color/timestamp handling, and an export path that does not simply record the preview window.
• However, Limitations says console/VRR capture still needs field retesting. There is also no complete compatibility matrix for HDR state, capture-card pixel format, and refresh-rate combinations.

It is suitable for consolidating preview, enhancement, and export, but the README cannot guarantee stable operation for your exact PS5, capture card, and HDR/VRR chain.