Add optional line blending for low-latency downscaling (alternative to decimation) #35
Description
Hi marqs,
First of all, thank you for the continued development of OSSC Pro – the recent updates (up to v0.81) have made it an incredible tool for both upscaling and downscaling to CRTs.
I'd like to request an optional line blending (averaging) mode for downscaling in low-latency paths (Line Multiplier or Adaptive modes), as an alternative to the current line decimation (dropping).
The idea is similar to what was discussed in older community threads (e.g., users like e8root suggested blending input lines instead of dropping them for non-integer downscales, like 720p → 240p or similar). This would:
- Reduce shimmering/artifacts on moving content compared to pure decimation
- Maintain near-zero latency (much less than a full frame buffer), since it could be done line-by-line or with minimal buffering, leveraging the Cyclone V FPGA resources and available memory (internal SRAM + LPDDR2)
- Keep the "lagless" philosophy of OSSC Pro, unlike full scaler modes that add more delay
It wouldn't need to replace decimation – just an optional mode (e.g., "Downscale method: Decimation / Blending") for supported resolutions, perhaps starting with common ones like 3x/5x integer drops.
The hardware seems more than capable (plenty of logic and memory headroom), and since the project is open-source, this could greatly enhance downscaling quality for modern sources on CRTs without sacrificing input lag.
What do you think – is this feasible in a future firmware update?
Thanks again for your amazing work!