From f3cefbc3ad104825edbb863df88d57eb7021a324 Mon Sep 17 00:00:00 2001
From: Jonas Smedegaard <dr@jones.dk>
Date: Fri, 5 Feb 2021 21:31:05 +0100
Subject: docs: add comment on how to establish/verify rendering settings

---
 localvideowebencode | 28 ++++++++++++++++++++++++++++
 1 file changed, 28 insertions(+)

diff --git a/localvideowebencode b/localvideowebencode
index 87b1a64..bb1ce63 100755
--- a/localvideowebencode
+++ b/localvideowebencode
@@ -158,6 +158,34 @@ samplestart=0
 samplelength=150
 compression=normal
 
+# establish/verify rendering settings for reference codec
+#
+# 1) pick reference frame:encoding format: 360p30:h.264
+# 2) pick reference codec (incl. normal settings)
+# 3) render reference frame format with --compression=quality
+#    * check that bpp is near "refbpp" of 0.12
+#      (otherwise adapt settings or do research and update refbpp)
+# 4) render misc. frame formats (esp. modulus 16) with --compression=quality
+#    * check that bpp derives from refbpp by "rule of .75"
+#      (otherwise adapt settings and/or do research and update rule)
+# 5) render misc. frame formats with --compression=normal
+#    * check that bpp derives from refbpp by "rule of .75"
+# 6) optionally pick dirty/hq settings
+#    * check that --compression=quality has filesize similar as normal
+
+# establish/verify rendering settings for non-reference codec
+#
+# 1) pick normal settings
+# 2) render reference frame format with --compression=quality
+#    * check that this "codec-bpp" deviates reasonably from refbpp
+# 3) render misc. frame formats (esp. modulus 16) with --compression=quality
+#    * check that bpp derives from codec-bpp by "rule of .75"
+#      (otherwise adapt settings and/or extend rule)
+# 4) render misc. frame formats with --compression=normal
+#    * check that bpp derives from codec-bpp by "rule of .75"
+# 5) optionally pick dirty/hq settings
+#    * check that --compression=quality has similar filesize as normal
+
 # VP8 is rumored to compress ~20% worse than H.264
 factor_vp8=120/100
 # VP9 compresses ~30% better than H.264 for practical use-cases
-- 
cgit v1.2.3