bitrate=<value>

Sets the average bitrate to be used in kbits/second (default:off).

Since local bitrate may vary, this average may be inaccu

-rate  for very short videos (see ratetol).

Constant bitrate can be achieved by combining this with vbv_maxrate,

 at  significant reduction in quality.

-B, --bitrate <integer>

Set bitrate (kbit/s)

qp=<0-51>

This selects the quantizer to use for P-frames.  I- and B-frames

are offset from this value by ip_factor and  pb_factor,  respec-

tively.    20-40  is  a useful range (default: 26).  Lower values

result in better fidelity, but higher bitrates.  0 is  lossless.

Note   that   quantization   in  H.264  works  differently  from

MPEG-1/2/4: H.264's quantization parameter (QP) is  on  a     loga-

rithmic  scale.    The  mapping  is  approximately H264QP = 12 +6*log2(MPEGQP).  For example, MPEG  at  QP=2  is    equivalent  to

H.264 at QP=18.

-q, --qp <integer>

Set QP (0=lossless) [26]

crf=<1-50>

Enables  constant     quality  mode,     and selects the quality.  The　scale is similar to QP.  Like the bitrate-based modes, this  al-lows each frame to use a different QP based on the frame's com-

plexity.

--crf <float>

Quality-based VBR (nominal QP)

pass=<1-3>

Enable 2 or 3-pass mode.    It is recommended to always encode  in

2     or  3-pass  mode as it leads to a better bit distribution and

improves overall quality.

　1    first pass

　2    second pass (of two pass encoding)

　3    Nth pass (second and third passes of three  pass encod-ing)

Here is how it works, and how to use it:

The  first  pass    (pass=1)  collects statistics on the video and

writes them to a file.  You might want to deactivate  some  CPU-

hungry options, apart from the ones that are on by default.

In  two pass mode, the second pass (pass=2) reads the statistics

file and bases ratecontrol decisions on it.

In three pass mode, the second pass (pass=3, that is not a typo)

does  both: It first reads the statistics, then overwrites them.

You can use all encoding options,     except     very  CPU-hungry  op-

tions.

The  third  pass (pass=3) is the same as the second pass, except

that it has the second pass' statistics to work from.   You  can

use all encoding options, including CPU-hungry ones.

The  first pass may use either average bitrate or constant quan-

tizer.  ABR is recommended, since it does not require guessing a

quantizer.  Subsequent passes are ABR, and must specify bitrate.

-p, --pass <1|2|3>

Enable multipass ratecontrol

- 1: First pass, creates stats file

- 2: Last pass, does not overwrite stats file

- 3: Nth pass, overwrites stats file

turbo=<0-2>

Fast first pass mode.  During the first pass of a     two  or  more

pass  encode  it is possible to gain speed by disabling some op-

tions with negligible or even no impact on the final pass output

quality.

   0    disabled (default)

   1    Reduce  subq, frameref and disable some inter-macroblock

    partition analysis modes.

   2    Reduce subq and frameref to 1, use a diamond  ME    search

    and disable all partition analysis modes.

Level 1 can increase first pass speed up to 2x with no change in

the global PSNR of the final pass compared  to  a     full  quality

first pass.

Level  2    can  increase first pass speed up to 4x with about +/-

0.05dB change in the global PSNR of the final pass compared to a

full quality first pass.

該 当無し

keyint=<value>

Sets maximum interval between IDR-frames (default: 250).    Larger

values save bits, thus improve quality, at the cost  of  seeking

precision.   Unlike  MPEG-1/2/4,    H.264 does not suffer from DCT

drift with large values of keyint.

-I, --keyint <integer>

Maximum GOP size [250]

keyint_min=<1-keyint/2>

Sets minimum interval  between  IDR-frames  (default:  25).   If

scenecuts appear within this interval, they are still encoded as

I-frames, but do not start a new GOP.  In H.264, I-frames do not

necessarily  bound a closed GOP because it is allowable for a P-

frame to be predicted from more frames than just the  one     frame

before it (also see frameref).  Therefore, I-frames are not nec-

essarily seekable.  IDR-frames restrict subsequent P-frames from

referring to any frame prior to the IDR-frame.

-i, --min-keyint <integer>

Minimum GOP size [25]

scenecut=<-1-100>

Controls    how  aggressively  to  insert extra I-frames (default:

40).  With small values of scenecut,  the     codec    often  has  to

force  an     I-frame  when it would exceed keyint.    Good values of

scenecut may find a better location for the I-frame.  Large val-

ues  use    more  I-frames    than necessary, thus wasting bits.  -1

disables scene-cut detection, so I-frames are inserted only once

every  other  keyint frames, even if a scene-cut occurs earlier.

This is not recommended and wastes bitrate as scenecuts  encoded

as  P-frames  are     just as big as I-frames, but do not reset the

"keyint counter".

--scenecut <integer>

How aggressively to insert extra I-frames [40]

frameref=<1-16>

Number of previous frames used as predictors in B- and  P-frames

(default:     1).   This  is effective in anime, but in live-action

material the improvements usually drop off very rapidly above  6

or  so  reference frames.     This has no effect on decoding speed,

but does increase the memory needed for decoding.     Some decoders

can only handle a maximum of 15 reference frames.

-r, --ref <integer>

Number of reference frames [1]

bframes=<0-16>

maximum  number  of consecutive B-frames between I- and P-frames

(default: 0)

-b, --bframes <integer>

Number of B-frames between I and P [0]

(no)b_adapt

Automatically decides when to use B-frames and how many,    up  to

the  maximum  specified  above (default: on).  If this option is

disabled, then the maximum number of B-frames is used.

--no-b-adapt

Disable adaptive B-frame decision

b_bias=<-100-100>

Controls the decision performed by  b_adapt.   A    higher    b_bias

produces more B-frames (default: 0).

--b-bias <integer>

Influences how often B-frames are used [0]

(no)b_pyramid

Allows  B-frames    to  be used as references for predicting other

frames.  For example, consider 3 consecutive B-frames: I0 B1  B2

B3 P4.  Without this option, B-frames follow the same pattern as

MPEG-[124].  So they are coded in the order I0 P4 B1 B2 B3,  and

all  the    B-frames  are predicted from I0 and P4.     With this op-

tion, they are coded as I0 P4 B2 B1  B3.     B2  is     the  same  as

above,  but  B1 is predicted from I0 and B2, and B3 is predicted

from B2 and P4.  This usually results in slightly improved  com-

pression,     at almost no speed cost.  However, this is an experi-

mental option: it is not fully tuned and may  not     always     help.

Requires    bframes     >= 2.    Disadvantage: increases decoding delay

to 2 frames.

--b-pyramid

Keep some B-frames as references

(no)deblock

Use deblocking filter (default: on).  As it  takes  very    little

time compared to its quality gain, it is not recommended to dis-

able it.

--no-deblock

Disable loop filter

deblock=<-6-6>,<-6-6>

The first parameter  is  AlphaC0    (default:  0).     This  adjusts

thresholds for the H.264 in-loop deblocking filter.  First, this

parameter adjusts the maximum amount of change that  the    filter

is  allowed to cause on any one pixel.  Secondly, this parameter

affects the threshold for difference across the edge being  fil-

tered.   A  positive  value reduces blocking artifacts more, but

will also smear details.

The second parameter is Beta (default: 0).  This affects the de-

tail  threshold.     Very  detailed blocks are not filtered, since

the smoothing caused by the filter would be more noticeable than

the original blocking.

The  default behavior of the filter almost always achieves opti-

mal quality, so it is best to either leave it alone, or make on-

ly  small adjustments.  However, if your source material already

has some blocking or noise which you would like  to  remove,  it

may be a good idea to turn it up a little bit.

-f, --deblock <alpha:beta>

Loop filter AlphaC0 and Beta parameters [0:0]

(no)cabac

Use  CABAC (Context-Adaptive Binary Arithmetic Coding) (default:

on).  Slightly slows down encoding and decoding, but should save

10-15%  bitrate.    Unless you are looking for decoding speed, you

should not disable it.

--no-cabac

Disable CABAC

qp_min=<1-51> (ABR or two pass)

Minimum quantizer, 10-30 seems to be a  useful  range  (default:

10).

--qpmin <integer>

Set min QP [10]

qp_max=<1-51> (ABR or two pass)

maximum quantizer (default: 51)

--qpmax <integer>

Set max QP [51]

qp_step=<1-50> (ABR or two pass)

maximum  value  by which the quantizer may be incremented/decre-

mented between frames (default: 4)

--qpstep <integer>

Set max QP step [4]

ratetol=<0.1-100.0> (ABR or two pass)

allowed variance in average bitrate (no particular  units) 

(de-fault: 1.0)

--ratetol <float>

Allowed variance of average bitrate [1.0]

vbv_maxrate=<value> (ABR or two pass)

maximum local bitrate, in kbits/second (default: disabled)

--vbv-maxrate <integer>

Max local bitrate (kbit/s) [0]

vbv_bufsize=<value> (ABR or two pass)

averaging     period for vbv_maxrate, in kbits (default: none, must

be specified if vbv_maxrate is enabled)

--vbv-bufsize <integer>

Enable CBR and set size of the VBV buffer (kbit) [0]

vbv_init=<0.0-1.0> (ABR or two pass)

initial buffer occupancy, as a fraction of vbv_bufsize (default:

0.9)

--vbv-init <float>

Initial VBV buffer occupancy [0.9]

ip_factor=<value>

quantizer factor between I- and P-frames (default: 1.4)

--ipratio <float>

QP factor between I and P [1.40]

pb_factor=<value>

quantizer factor between P- and B-frames (default: 1.3)

--pbratio <float>

QP factor between P and B [1.30]

qcomp=<0-1> (ABR or two pass)

quantizer     compression  (default: 0.6).  A lower value makes the

bitrate more constant, while a higher value makes the  quantiza-

tion parameter more constant.

--qcomp <float>

QP curve compression: 0.0 => CBR, 1.0 => CQP [0.60]

cplx_blur=<0-999> (two pass only)

Temporal    blur  of  the estimated frame complexity, before curve

compression (default: 20).  Lower     values     allow    the  quantizer

value  to     jump around more, higher values force it to vary more

smoothly.     cplx_blur ensures that each I-frame has quality  com-

parable  to the following P-frames, and ensures that alternating

high and low complexity frames (e.g. low fps animation)  do  not

waste bits on fluctuating quantizer.

--cplxblur <float>

Reduce fluctuations in QP (before curve compression) [20.0]

qblur=<0-99> (two pass only)

Temporal    blur  of  the quantization parameter, after curve com-

pression (default: 0.5).    Lower values allow the quantizer value

to jump around more, higher values force it to vary more smooth-

ly.

--qblur <float>

Reduce fluctuations in QP (after curve compression) [0.5]

zones=<zone0>[/<zone1>[/...]]

User specified quality  for  specific  parts  (ending,  credits,

...).  Each zone is <start-frame>,<end-frame>,<option> where op-

tion may be

   q=<0-51>

    quantizer

   b=<0.01-100.0>

    bitrate multiplier

NOTE: The quantizer option is not strictly enforced.  It affects

only  the planning stage of ratecontrol, and is still subject to

overflow compensation and qp_min/qp_max.

--zones <zone0>/<zone1>/...

Tweak the bitrate of some regions of the video

Each zone is of the form

    <start frame>,<end frame>,<option>

    where <option> is either

  q=<integer> (force QP)

    or  b=<float> (bitrate multiplier)

direct_pred=<name>

Determines the type of motion prediction used  for direct macroblocks in B-frames.

　none

　　Direct macroblocks are not used.

　spatial

　　Motion vectors are extrapolated from neighboring blocks.(default)

　temporal

　　Motion vectors are interpolated from the following  P-frame.

　auto

　　The  codec selects between spatial and temporal for each frame.

Spatial and temporal are approximately the same speed and PSNR,

the  choice  between them depends on the video content.  Auto is

slightly better, but slower.  Auto is most effective  when  com-

bined with multipass.  direct_pred=none is both slower and lower

quality.

--direct <string>

Direct MV prediction mode ["spatial"]

    - none, spatial, temporal, auto

(no)weight_b

Use weighted prediction in B-frames.  Without this option, bidi-

rectionally predicted macroblocks give equal weight to each ref-

erence frame.  With this option, the weights are    determined  by

the temporal position of the B-frame relative to the references.

Requires bframes > 1.

-w, --weightb

Weighted prediction for B-frames

partitions=<list>

Enable some optional macroblock types (default:p8x8,b8x8,i8x8,i4x4).

p8x8

　Enable types p16x8, p8x16, p8x8.

p4x4

　Enable types p8x4, p4x8, p4x4.

　p4x4 is recommended only with subq >= 5, and 

　only at low resolutions.

b8x8

　Enable types b16x8, b8x16, b8x8.

i8x8

　Enable type i8x8. i8x8 has no effect unless  8x8dct  is enabled.

i4x4

　Enable type i4x4.

all

　Enable all of the above types.

none

　Disable all of the above types.

Regardless  of this option, macroblock types p16x16, b16x16, and i16x16 are always enabled.

The idea is to find the type and size that best describe a  certain  area  of the picture.  For example, a global pan is better represented by 16x16 blocks, while small moving objects are better represented by smaller blocks.

-A, --partitions <string>

Partitions to consider ["p8x8,b8x8,i8x8,i4x4"]

    - p8x8, p4x4, b8x8, i8x8, i4x4

    - none, all

    (p4x4 requires p8x8. i8x8 requires --8x8dct.)

※オレンジ部、MEncoder側に無い。

※旧mv系はこっちか。

--direct-8x8 <-1|0|1>   Direct prediction size [-1]
    -  0: 4x4（※(no)4x4mv か？）
    -  1: 8x8（※(no)8x8mv, (no)b8x8mvか？）
    - -1: smallest possible according to level

(no)8x8dct

Adaptive    spatial     transform  size: allows macroblocks to choose

between 4x4 and 8x8 DCT.    Also allows the i8x8 macroblock     type.

Without this option, only 4x4 DCT is used.

-8, --8x8dct

Adaptive spatial transform size

me=<name>

Select fullpixel motion estimation algorithm.

   dia  diamond search, radius 1 (fast)

   hex  hexagon search, radius 2 (default)

   umh  uneven multi-hexagon search (slow)

   esa  exhaustive search (very slow, and no better than umh)

--me <string>

Integer pixel motion estimation method ["hex"]

    - dia: diamond search, radius 1 (fast)

    - hex: hexagonal search, radius 2

    - umh: uneven multi-hexagon search

    - esa: exhaustive search (slow)

me_range=<4-64>

radius  of  exhaustive  or multi-hexagon motion search (default:16)

--merange <integer>

Maximum motion vector search range [16]

subq=<1-7>

Adjust subpel refinement quality. This parameter controls qual-

ity versus speed tradeoffs involved in the motion estimation de-

cision process.  subq=5 can compress up to 10% better than subq=1.

1    Runs fullpixel precision motion estimation on all åcandi-

    date macroblock types.   Then  selects  the  best     type.

    Then refines the motion of that type to fast quarterpix-

    el precision (fastest).

2    Runs halfpixel precision motion estimation on all candi-

    date  macroblock    types.     Then  selects    the best type.

    Then refines the motion of that type to fast quarterpix-

    el precision.

3    As 2, but uses a slower quarterpixel refinement.

4    Runs  fast  quarterpixel    precision motion estimation on

    all candidate macroblock types.  Then selects  the  best

    type.   Then  finishes  the  quarterpixel refinement for

    that type.

5    Runs best quality quarterpixel precision motion  estima-

    tion on all candidate macroblock types, before selecting

    the best type (default).

6    Enables rate-distortion optimization of macroblock types

    in I- and P-frames.

7    Enables  rate-distortion    optimization of motion vectors

    and intra modes. (best)

In the above, "all candidates" does not exactly mean all enabled

types: 4x4, 4x8, 8x4 are tried only if 8x8 is better than 16x16.

-m, --subme <integer>

Subpixel motion estimation and partition decision quality: 1=fast, 7=best. [5]

(no)chroma_me

Takes into account chroma     information  during  subpixel    motion

search (default: enabled).  Requires subq>=5.

--no-chroma-me

Ignore chroma in motion estimation

(no)mixed_refs

Allows each 8x8 or 16x8 motion partition to independently select

a reference frame.  Without this option, a whole macroblock must

use the same reference.  Requires frameref>1.

--mixed-refs

Decide references on a per partition basis

(no)brdo

Enables  rate-distortion    optimization of macroblock types in B-

frames.  Requires subq>=6.

--b-rdo

RD based mode decision for B-frames. Requires subme 6.

(no)bime

Refine the two motion vectors used in bidirectional macroblocks,

rather  than  re-using  vectors  from  the  forward and backward

searches.     This option has no effect without B-frames.

--bime

Jointly optimize both MVs in B-frames

trellis=<0-2>

rate-distortion optimal quantization

   0    disabled (default)

   1    enabled only for the final encode

   2    enabled  during  all  mode  decisions  (slow,   requires subq>=6)

-t, --trellis <integer>

Trellis RD quantization. Requires CABAC. [0]

    - 0: disabled

    - 1: enabled only on the final encode of a MB

    - 2: enabled on all mode decisions

deadzone_inter=<0-32>

Set  the    size  of the inter luma quantization deadzone for non-

trellis quantization (default: 21).

--deadzone-inter <int>

Set the size of the inter luma quantization deadzone [21]

deadzone_intra=<0-32>

Set the size of the intra luma quantization  deadzone  for  non-

trellis quantization (default: 11).

--deadzone-intra <int>

Set the size of the intra luma quantization deadzone [11]

    Deadzones should be in the range 0 - 32.

(no)fast_pskip

Performs    early  skip  detection in P-frames (default: enabled).

This usually improves speed at no cost,  but  it    can  sometimes

produce artifacts in areas with no details, like sky.

--no-fast-pskip

Disables early SKIP detection on P-frames

(no)dct_decimate

Eliminate     dct blocks in P-frames containing only a small single

coefficient (default: enabled).  This will remove some  details,

so  it  will  save bits that can be spent again on other frames,

hopefully raising overall subjective quality.  If you  are  com-

pressing    non-anime  content with a high target bitrate, you may

want to disable this to preserve as much detail as possible.

--no-dct-decimate

Disables coefficient thresholding on P-frames

nr=<0-100000>

Noise reduction, 0 means disabled.  100-1000 is a     useful     range

for  typical  content, but you may want to turn it up a bit more

for very noisy content (default: 0).  Given its small impact  on

speed, you might want to prefer to use this over filtering noise

away with video filters like denoise3d or hqdn3d.

--nr <integer>

Noise reduction [0]

chroma_qp_offset=<-12-12>

Use a different quantizer for chroma as compared to luma.      Use-

ful values are in the range <-2-2> (default: 0).

--chroma-qp-offset <integer>

QP difference between chroma and luma [0]

cqm=<flat|jvt|<filename>>

Either  uses  a predefined custom quantization matrix or loads a

JM format matrix file.

   flat

    Use the predefined flat 16 matrix (default).

   jvt

    Use the predefined JVT matrix.

   <filename>

    Use the provided JM format matrix file.

NOTE: Windows CMD.EXE users may experience problems with parsing

the command line if they attempt to use all the CQM lists.  This

is due to a command line length limitation.  In this case it  is

recommended the lists be put into a JM format CQM file and load-

ed as specified above.

--cqm <string>

Preset quant matrices ["flat"]

    - jvt, flat

--cqmfile <string>

Read custom quant matrices from a JM-compatible file Overrides any other --cqm* options.

cqm4iy=<list> (also see cqm)

Custom 4x4 intra luminance matrix, given as a list of  16     comma

separated values in the 1-255 range.

--cqm4 <list>

Set all 4x4 quant matrices

    Takes a comma-separated list of 16 integers.

--cqm4i, --cqm4p

Set both luma and chroma quant matrices

--cqm4iy, --cqm4ic, --cqm4py, --cqm4pc

Set individual quant matrices

cqm4ic=<list> (also see cqm)

Custom 4x4 intra chrominance matrix, given as a list of 16 comma

separated values in the 1-255 range.

cqm4py=<list> (also see cqm)

Custom 4x4 inter luminance matrix, given as a list of  16     comma

separated values in the 1-255 range.

cqm4pc=<list> (also see cqm)

Custom 4x4 inter chrominance matrix, given as a list of 16 comma

separated values in the 1-255 range.

cqm8iy=<list> (also see cqm)

Custom 8x8 intra luminance matrix, given as a list of  64     comma

separated values in the 1-255 range.

--cqm8 <list>

Set all 8x8 quant matrices

    Takes a comma-separated list of 64 integers.

--cqm8i, --cqm8p

Set both luma and chroma quant matrices

cqm8py=<list> (also see cqm)

Custom  8x8  inter luminance matrix, given as a list of 64 comma

separated values in the 1-255 range.

level_idc=<10-51>

Set the bitstream's level as defined by annex  A    of  the     H.264

standard    (default:  51  - Level 5.1).  This is used for telling

the decoder what capabilities it needs to support.  Use this pa-

rameter  only  if you know what it means, and you have a need to

set it.

--level <string>

Specify level (as defined by Annex A)

threads=<1-4>

Split each frame into slices and encode them  in    parallel  (de-

fault:  1).   Also  allows multithreaded decoding if the decoder

supports it (lavc does not).  This has a slight penalty to  com-

pression.      Requires that libx264 was compiled with pthread sup-

port; if not, this option will produce  a     warning  and  enables

slices but not multithreading.

--threads <integer>

Parallel encoding (uses slices)

(no)global_header

Causes  SPS and PPS to appear only once, at the beginning of the

bitstream (default: disabled).  Some players, such as  the  Sony

PSP, require the use of this option.  The default behavior caus-

es SPS and PPS to repeat prior to each IDR frame.

該 当無し

(no)interlaced

Treat the video content as interlaced.

--interlaced

Enable pure-interlaced mode

log=<-1-3>

Adjust the amount of logging info printed to the screen.

-1   none

0   Print errors only.

1   warnings

2   PSNR and other analysis statistics when the encode  fin-

    ishes (default)

3   PSNR,  QP, frametype, size, and other statistics for ev-

    ery frame

該 当無し

(no)psnr

Print signal-to-noise ratio statistics.

NOTE: The 'Y', 'U', 'V', and 'Avg' PSNR fields  in  the  summary

are  not    mathematically    sound  (they are simply the average of

per-frame PSNRs).     They are kept only for comparison to  the  JM

reference     codec.     For all other purposes, please use either the

'Global' PSNR, or the per-frame PSNRs printed by log=3.

--no-psnr

Disable PSNR computation

(no)ssim

Print the Structural Similarity Metric results.  This is an  al-

ternative     to  PSNR,  and may be better correlated with the per-

ceived quality of the compressed video.

--no-ssim

Disable SSIM computation

(no)visualize

Enable x264 visualizations during encoding.  If the x264 on your

system  supports    it, a new window will be opened during the en-

coding process, in which x264 will attempt to present  an     over-

view of how each frame gets encoded.  Each block type on the vi-

sualized movie will be colored as follows:

red/pink

    intra block

blue

    inter block

green

    skip block

yellow

    B-block

This feature can    be  considered    experimental  and  subject  to

change.    In  particular, it depends on x264 being compiled with

visualizations enabled.  Note that  as  of  writing  this,  x264

pauses  after  encoding  and visualizing each frame, waiting for

the user to press a key, at which point the next frame  will  be

encoded.

--visualize

Show MB types overlayed on the encoded video

該 当不詳

灰：-x264encopts以外にある（と思し き）もの

青：-x264encoptsにあってしかるべきもの