SVT-AV1 Parameters for Sharper Video Output
This article outlines the specific internal parameters within the
SVT-AV1 encoder (libsvtav1) that directly influence and
control the sharpness of the final video output. We will examine the
core settings—including the loop filter sharpness, quantization
matrices, and film grain synthesis—explaining how to configure them to
prevent blurriness and preserve high-frequency details in your
encodes.
The Sharpness Parameter
(sharpness)
The most direct way to control image sharpness in SVT-AV1 is through
the internal sharpness parameter
(configured via -svtav1-params sharpness=X in FFmpeg). *
Range: -2 to 7 (with 0 as the default). *
Function: This parameter directly adjusts the loop
filter’s sharpness metric. Higher positive values restrict the
deblocking filter from smoothing out edge transitions, resulting in a
noticeably sharper image. Negative values allow the loop filter to
smooth out more edges, reducing blockiness at the expense of
sharpness.
Deblocking Loop Filter
(enable-dlf)
AV1 uses a Deblocking Loop Filter (DLF) to smooth out block artifacts
caused by compression. * Function: While the DLF
improves objective metrics, it can introduce unwanted blur. If
preserving maximum sharpness is your primary goal, you can disable this
filter entirely using enable-dlf=0 (or
disable-dlf=1 depending on your SVT-AV1 version). *
Caveat: Disabling the loop filter altogether may
introduce visible block artifacts, especially at lower bitrates.
Quantization Matrices
(enable-qm)
Quantization matrices control how distortion is distributed across
different frequency bands. * Function: By enabling
quantization matrices using enable-qm=1, the encoder
allocates bits more aggressively to high-frequency details (which
represent sharp edges and fine textures) rather than flat surfaces. *
Adjustment: You can fine-tune this with
qm-min and qm-max (ranging from 0 to 15) to
set the minimum and maximum flatness allowed for the quantization
matrices, preventing the encoder from over-smoothing complex scenes.
Film Grain Synthesis
(film-grain)
At lower to medium bitrates, AV1 encoders naturally discard
high-frequency noise and fine textures to save space, resulting in a
“washed out” or overly smooth look. * Function:
Enabling film grain synthesis (e.g., film-grain=8 or
higher) instructs the encoder to analyze the original video’s grain,
discard it during compression, and then synthetically regenerate it
during playback. * Result: While this does not increase
actual resolution sharpness, the simulated grain restores perceived
texture and sharpness, preventing the video from looking artificially
soft.
Visual Quality Tuning
(tune)
The mathematical algorithm the encoder uses to evaluate video quality
heavily impacts sharpness. * Function: Setting
tune=0 (Visual Quality / VQ) optimizes the encoder for
subjective human vision. This tuning path prioritizes the preservation
of sharp edges and fine details. * Alternative: Avoid
tune=1 (PSNR) or tune=2 (VMAF) if sharpness is
your priority, as these modes often smooth out textures to achieve
better mathematical compression scores.