Dejo aquí en spoiler la sección dedicada a los requisitos de vídeo y donde se encuentra la información que puede ser más interesante para no tener que mirar el tocho entero:
- Spoiler: mostrar
- 7) [ Video Codec ]
7.1) Video must be:
7.1.1) H.265/MPEG-H HEVC encoded with x265 10-bit for 720p/1080p HDR and 2160p SDR/HDR content.
7.1.2) H.264/MPEG-4 AVC encoded with x264 8-bit for 720p and 1080p SDR content.
7.1.3) Custom builds of x264/x265 are allowed and must be based off the current x264/x265 codebase. e.g. tMod, kMod.
7.2) x264/x265 headers must remain intact and must not be modified or removed.
7.3) x264/x265 must be kept up to date, with a maximum allowance, or grace period, of 60 days before groups are required
to update to the latest revision. [...]
7.4) Segmented encoding is not allowed.
7.5) Constant Rate Factor (--crf) must be used.
7.5.1) Decimal values may be used.
7.5.2) Starting with an initial value of 16 or 14 (for especially compressible sources) for 2160p and 17 for 720p and
1080p is recommended.
7.6) While the encoded video stream bitrate exceeds x percent (where x is defined below) of the source video stream
bitrate, the CRF value must be incremented by 1 or 0.1, until it does not.
7.6.1) 30% for 720p.
7.6.2) 60% for 1080p.
7.6.3) 70% for 2160p.
7.7) Use of 2-pass is accepted for all resolutions. However, this method should be used only in extreme cases and not as a
primary replacement for CRF methods. e.g. If the source contains an excessive amount of grain or black and white
scenes.
7.7.1) The NFO must provide sufficient detailed evidence of 2-pass producing a visual improvement, bitrate or file
size advantage over CRF in regards to the source used.
e.g. Source has heavy grain throughout, which required the use of 2-pass to remain transparent at a reasonable
bitrate. Included in the proof directory is test encodes of some troublesome areas at CRF and 2-pass
showing a substantial improvement in quality at a more reasonable bitrate. CRF needs 6.5GB to stay
transparent 2-pass only needed 4GB is detailed evidence.
e.g. The source was grainy so 2-pass was used is NOT detailed evidence.
7.7.2) Exceeding CRF 24 on any resolution is a good indication to consider the use of 2-pass.
7.7.3) 2-pass encodes must follow the percentage values in 7.6, it is recommended to target the maximum value allotted
and work down.
7.8 ) Exceeding the percentage values in 7.6 is allowed, but very detailed justification must be included in the NFO.
e.g. 720p encoded at CRF 13 resulting in an encode 50% of the source, this value was picked as the sourced content
had a lot of strobing effects (e.g. 1m10s-15m30s) and confetti scenes (e.g. 20m55s-50m60s) which compressed
poorly and required a higher bitrate to maintain transparency. Proof directory contains samples of these
troublesome scenes at 30% and 50% showing a substantial picture improvement.
7.9) Using unreasonably high CRF values or low target bitrates with 2-pass for the source content is considered a
technical flaw.
e.g. Producing a 1080p encode at 50% of the source bitrate, offering no justification in the NFO and having poor
transparency to the source IS a technical flaw.
e.g. Producing a 1080p encode at 30% of the source bitrate (This may occur on animation content to avoid a bloated
encode), offering justification in the NFO and having good transparency to the source is NOT a technical flaw.
7.10) Encoded video bitrate must not exceed the source video bitrate.
7.10.1) Video bitrate refers only to the bitrate for the video stream, and not the overall muxed bitrate of all
streams within the container.
7.10.2) The following algorithm can also be used to estimate a CRF value if the value originally used results in a
bitrate which exceeds the source's.
ValidCRF = [-6 * (DestinationBitrate - ExcessiveBitrate) / ExcessiveBitrate] + CRFUsed
e.g. Source bitrate: 4,500Kbps
Target bitrate: ~3,400Kbps
Encoded bitrate @ CRF17: 5,900Kbps
ValidCRF = (-6 * (3400 - 5900) / 5900) + 17
7.11) Settings cannot go below what is specified for preset (--preset) 'slower' for x264 or 'slow' for x265 by the
revision used.
7.12) Level (x264: --level, x265: --level-idc) must be:
7.12.1) '4.1' for 720p
7.12.2) '4.1' for 1080p, '4.2' for 1080p above 30fps
7.12.3) '5.1' for 2160p, '5.2' for 2160p above 30fps.
7.13) Custom matrices are not allowed.
7.14) Zones (--zones) may be used, but should be used sparingly with a high degree of caution, the NFO must provide
sufficient detailed evidence for each zone used.
e.g. Used a zone at CRF 13 between frames 1634-5343 due to confetti spam and heavy motion resulting in very poor
compressibility. Proof directory contains two samples of this scene with and without the use of the zone
showing a drastic improvement in quality while still keeping a reasonable file size.
7.15) GPU offloading or any other forms of acceleration (e.g. --opencl, nvenc) is not allowed.
7.16) Optional tuning (--tune) parameters allowed are: 'film', 'grain', or 'animation'.
7.17) Optional settings recommended for tuning per source, see forum.doom9.org for further information:
7.17.1) For complex video, --preset veryslow is encouraged.
7.17.2) --aq-mode 3 --aq-strength x.x
0.5-0.7 for grainy films.
0.6-0.9 for digital films.
0.9-1.1 for animation.
7.17.3) --psy-rd x.x:0.0
0.8-1.2 for films.
0.5-0.8 for animation.
7.17.4) --deblock x:x
-3:-3 for films.
1:1 for animation.
7.18) Sample Aspect Ratio (--sar) must be '1:1' (square).
7.19) Disabling deblocking (--no-deblock) is not allowed.
7.20) Frame rate must be passed to x264/x265 such that the keyframe interval (--keyint) and minimum GOP length
(--min-keyint) can be correctly set by the encoder. Changing these values is not allowed.
7.21) Color space must be 4:2:0.
7.22) Color matrix (--colormatrix) may be optionally set to 'bt709' for SDR content, but is not required.
7.23) x265 specifics:
7.23.1) Range (--range), color primaries (--colorprim), transfer (--transfer), color matrix (--colormatrix) and chroma
sample location (--chromaloc) must be set to the same value as the source, or omitted if the source value is
undefined.
7.23.2) Ultra-HD Bluray support (--uhd-bd) is not allowed.
7.23.3) High tier (--high-tier), repeat headers (--repeat-headers), access unit delimiter (--aud) and HRD signalling
(--hrd) must be enabled.
7.23.4) HDR releases must be encoded with:
7.23.4.1) HDR10 signalling (--hdr10) and HDR10 optimization (--hdr10-opt) enabled.
7.23.4.2) Master Display (--master-display) and Max CL (--max-cll) set to the same value as the source, or omitted
if the source value is undefined.
7.23.4.2.1) Master Display and Max CL values must be taken from the whole concatenated source, not a partial
source. e.g. the first m2ts file.
7.24) Any form of tone-mapping (e.g. HDR to SDR, DV to SDR, HDR10Plus to SDR) is not allowed, use INTERNAL.
7.25) HDR, DV and HDR10Plus must only be released at the source resolution, except in cases when 4.5.3 applies.
e.g. Source is 2160p and may only be released at 2160p, not 1080p or 720p.
e.g. Source is 1080p and may only be released at 1080p, not 720p.
7.26) Suggested command line:
7.26.1) x264 --preset slower --level ## --crf ##
7.26.1.1) Optional tweaks: --no-mbtree --no-fast-pskip --no-dct-decimate
7.26.2) x265 --high-tier --repeat-headers --aud --hrd --preset slow --level-idc ## --crf ## --range ## --colorprim ##
--transfer ## --colormatrix ## --chromaloc ##
7.26.2.1) If HDR append: --hdr10 --hdr10-opt --master-display ## --max-cll ##
7.26.2.2) Optional tweaks: --no-cutree --no-open-gop --no-sao --pmode --aq-mode 4
De lo que comenta
Dardo sobre primar "un tamaño determinado" antes lo hacían respecto a los tamaños de DVD5 para 720p y DVD9 para 1080p (esto es 4,37 y 7,95 GB) ahora parecen centrarse en un porcentaje respecto al bitrate de la fuente:
7.6) While the encoded video stream bitrate exceeds x percent (where x is defined below) of the source video stream
bitrate, the CRF value must be incremented by 1 or 0.1, until it does not.
7.6.1) 30% for 720p.
7.6.2) 60% for 1080p.
7.6.3) 70% for 2160p.
Claro que como dice Dardo cada fuente es un mundo y algunas muy compresibles necesitarán un porcentaje menor y otras más complejas no se ajustarán a esa exigencia sin un tratamiento más complejo. Aun así me parece un salto importante respecto a lo que venían haciendo, además han aumentado el requisitio mínimo sobre el preset a usar:
7.11) Settings cannot go below what is specified for preset (--preset) 'slower' for x264 or 'slow' for x265 by the
revision used.
Respecto a lo que comenta
professor keller sobre el x265 lo que me da la impresión es que mucha gente sobrestima su capacidad reductora respecto a x264. De ahí que se vean muchos bdrip en x265 que en realidad sólo son la versión actual de los bdrip 1080p de 3 GB que tanto le gustan a Dardo

.
A mí algo que me trae de cabeza por temporadas son las resoluciones. Durante mucho tiempo consumía exclusivamente 720p y poco a poco he aumentado mi acopio de 1080p, en gran medida gracias a alguien que todos conocemos. Al final el asunto se reduce a sacrificar un poco de calidad por más espacio de almacenamiento o al revés, sacrificar espacio de almacenamiento por calidad/resolución. Habrá quien diga que hoy los discos duros están a precios asequibles y que este debate no tiene sentido, aunque a mí me parece que sigue teniendo importancia. Además las tv actuales en 4k parecen favorecer lo de cuanto más grande mejor. Me gustaría que aportaseis vuestras opiniones al respecto.