Experiences with HC Encoder

[m3taPT]

Originally posted by edeus
perhaps it is still necessary to cut out extras if they are too large as i am finding 50% quality drop is still taking up valuable space even though they are basically unwatchable.

You might want to cut out extras first (retaining everything else), THEN doing the entire reencode with dvd-rb.

[gizzin]

Did anyone do a encode of Saving Private Ryan With HC, does it do a better job than CCE? I did a encode with 2.66 and 2.70 both were 6 passes but they both came out pixelated. Even though 2.70 was a little better. With a average bitrate of 2855.

[onesoul]

Originally posted by gizzin
Did anyone do a encode of Saving Private Ryan With HC, does it do a better job than CCE? I did a encode with 2.66 and 2.70 both were 6 passes but they both came out pixelated. Even though 2.70 was a little better. With a average bitrate of 2855. Personally I prefer hc or quenc, go ahead and try it, don't forget to check back your results.

[Boulder]

SPR is a pain with any encoder as the ultra-shaky camera makes a very bitrate hungry video.

I'd try Bach1 as the quant matrix, it can give a nice compressibility boost but will smooth the video a bit. It could be that you don't notice anything due to the almost constantly moving camera anyway. Using RemoveGrain(mode=2) + Deen() could also help. LRemoveDust(4,2) often works wonders (use it by itself only), the function can be found in the RemoveGrain thread at Avisynth usage forum.

[Morte66]

Originally posted by hank315
Best is to enable it.
On a scene change a new GOP will start, that's better than have a intra coded P or B frame in a GOP or there must be a special reason to have a fixed GOP length.

Thanks. I ran a full backup of Aliens R2 PAL (movie only) with DVD-RB/HC on Best profile and QLB matrix. It looks very good, slightly better than my usual FreeEnc (which gives pretty much identical results to using QuEnc in normal mode but multitasks more amiably). It took 961 minutes on my XP1900+ with some other concurrent usage; I'd expect about 360 minutes with FreeEnc.

I guess maybe I'll have to run QuEnc in high quality mode with trellis, which should take about the same time as HC, and see how that compares.

Now I have multiple installs of DVD-RB: 0.69 for FreeEnc which dislikes some current QuEnc switches, and 0.79 for HC.

[SAPSTAR]

Originally posted by hank315
Best is to enable it.
On a scene change a new GOP will start, that's better than have a intra coded P or B frame in a GOP or there must be a special reason to have a fixed GOP length.

For people having trouble downloading HC, it's also in the Doom9 download section.

First hank315 :thanks:, congratulations for your encoder, I did a lot experiment with it, and you know what, I tend to like it more and more, I am/was a CCE believer....well I may change my mind, yes your encoder is slower, but I found the quality is better with HC. :)

Anyway, I was wondering if you plan to correct the bitrate control pb on HCbatch in a close future ? That's the only thing that's preventing me to use it for all my encoding jobs.

[TheSeeker]

I use dvd rebuilder 0.80, avisynth 2.55, dgdecode 1.10, hc_batch 0.12. On a completely unpreprocessed source. HC_batch is in same directory as rebuilder.exe and named HC_Batch.exe. Anyways, when I go through the prepare phase, all is well. But the instant I hit the encode button, no matter what I do, I get an Runtime Error 5: Invalid Procedure Call or Argument. I know HC_Batch on its own works on my machine because I have taken the hc.ini file that dvd rb creates and thrown it into the same directory as hc_batch and the encoder works to rebuild that particular segment. SO it will work fantastically if I kick every segment off manually. Its just the interation with DVD REbuilder that is pissing it off or something. Not sure what to do about this. Let me know if you need any more info Jdobbs. Im at work right now but I can give you whatever you need when I get home.

[cmsoliveira]

@TheSeeker

I believe the problem you're having is because of the name...

Quoting Jdobbs:

The HC encoder's .EXE file must be loaded in the same directory as DVD Rebuilder and the name must start with "HCBATCH" (e.g. "HCBATCH_012.EXE").

In your case you have HC_BATCH.Exe

Hope it helps

[TheSeeker]

hmmm.. very interesting. Your right. I will confirm the file name when i get home. If your right, i will jump for joy that I can now use this fantastic encoder with rebuilder. but it would be a stupid mistake on my part.

[johnhamler1]

hello,

is there a way, to increase the tyargetsize of HC? it miss around 200 mb on my DVD-5?
this new HC encoder is great.

[Fishman0919]

Yes, this is from the "Sticky: DVD ReBuilder: Settings".

Hidden Settings

These are settings not available via the GUI, but can be set in the 'Rebuilder.ini' file in the your DVD ReBuilder program directory. (Note: this is not the same as the file created in the WORKING directory named 'Rebuilder.ini'). These settings are for advanced users, otherwise they wouldn't be hidden.

All optional settings MUST be placed in the '[Options]' section of the .ini file, otherwise they will be ignored. Any settings changed here will not be activated until the next time you run DVD-RB.



Target Output Size

TargetSectors= nnnn
CCETargetSectors=nnnn
ReJigTargetSectors=nnnn
QuEncTargetSectors=nnnn

Use Sectors, 1 sector = 2Kb.
Example: CCETargetSectors=2236400

As a reference the default for DVD-RB is 2236400 sectors and the maximum for a DVD-5 is 2297888. You should always leave a some room for error.

TargetSectors will be the default for all methods. Other, mode specific, target sector settings will override it when using that mode.

-------------------
Use TargetSectors= nnnn and set "nnnn" to something a little higher then the default (2236400), I use 2252000 with HC Encoder and it gets me a DVD about 4.35g

[Vanderlow]

Of course the person who said they missed by 200mb is probably talking about the non-Pro version (as that is what I get) and changing the rebuilder.ini settings my be ignored like the 8000 bit rate limit was. I'm just saying maybe. Report back if it works. If the Pro version with HC is coming at 4.32Gb, as JDobbs reports, that tells me the Pro version's settings for HC seem to have the settings right for targeting the size at 4.32Gb and maybe Hank should see what the Pro version is using in this regard and adjust the non-Pro version accordingly.

[TCrowe]

Originally posted by Vanderlow
Of course the person who said they missed by 200mb is probably talking about the non-Pro version (as that is what I get) and changing the rebuilder.ini settings my be ignored like the 8000 bit rate limit was. I'm just saying maybe. Report back if it works. If the Pro version with HC is coming at 4.32Gb, as JDobbs reports, that tells me the Pro version's settings for HC seem to have the settings right for targeting the size at 4.32Gb and maybe Hank should see what the Pro version is using in this regard and adjust the non-Pro version accordingly.

I came up short, 3.87 GB, using .80 Pro in my initial (only) test of HC. So I don't know if Pro Vs. Non-Pro will resolve this problem. I have only done one test however. Somebody recommended setting the GOP to 12 but I have not tried this yet.

[Oldeman]

I just ran DVD-RB ver.80 Pro with HC in one pass, GOP-12 and came up with a output size of 3.94Gb. Target sectors were set at 2266000. I am running again with target sectors raised to 2277000. HC has good quality but has consistantly undersized for me.

I kinda remember someone said that sizing was better i

[ernstblaauw]

Target Output Size

TargetSectors= nnnn
CCETargetSectors=nnnn
ReJigTargetSectors=nnnn
QuEncTargetSectors=nnnn

Use Sectors, 1 sector = 2Kb.
Example: CCETargetSectors=2236400
Is there also a 'HCTargetSector'-setting?

[jdobbs]

It will use the "TargetSectors" or "QuEncTargetSectors" with the QuEnc one having priority. I just forgot to add it and will put an HCTargetSectors in the next version. Also, I don't think I documented anywhere, but there is also a NoEncodeTargetSectors= value that can be entered...

[hank315]

I know about the undersizing issue, tests I did also showed some undersizing.
If the source can be compressed very easily the bitrate control as implemented now has trouble to keep the desired bitrate.
ATM I'm working on a complete new encoding engine, one of the things that will be changed is the bitrate control.

Just wondering what kind of DVD's undersize under 4 GB, smallest backup I did was 4.25 GB...

[jdobbs]

I've never gotten any that small either... It's possible it may not be your encoder at all... I haven't found it yet, but I have a suspicion that I may have a bug related to recovering sectors when "Steal space from Extras" is used.

[quantum]

So far my HC encodes have been right on the money also. My last Video_ts directory was 4,658,577,408.

I'd also like to take this opportunity to say that Hanks encoder using the 'best' option has given me noticably better encodes than CCE under heavy compression, around 2,100 Kbs.

I finished 3 episodic disks using both CCE and HC and compared them in Vdub and HC has less macro blocking and tends to be closer to the original. The difference is not earth shattering, but it is noticeable.

For higher bitrate encodes, both will do very well at 3000, so CCE is probably the better choice since it's faster. I haven't done any tests using HC's faster options so they may be worth exploring for higher bitrates for those interested in using only HC.

I realize 99% of the masses would rather spend months reading messages about other peoples work rather than do actual testing themselves, so I'll feed the sloth by posting my results.

Yes, Quantum has deemed HC to be cool. Swarm to it and give Hank our thanks ;)

[Oldeman]

After raising target sectors to 2277000, I got exactly the same size 3.94Gb as before. Movie was preprocessed with DvdReMake to 5.4Gb.
I selected the french audio stream (364mb)to remove with with Dvd-Rebuilder.

DVD-RB v,80, HCbatch12a.
Options, 1 pass, GOP 12, DC precision 9, Dynamic bit rates, did not select steal space from extras. Did not select dynamic bit rates.

[Oldeman]

correction did select Dynamic bit rates, one pass CQ vbr...

[jdobbs]

Originally posted by Oldeman
After raising target sectors to 2277000, I got exactly the same size 3.94Gb as before. Movie was preprocessed with DvdReMake to 5.4Gb.
I selected the french audio stream (364mb)to remove with with Dvd-Rebuilder.

DVD-RB v,80, HCbatch12a.
Options, 1 pass, GOP 12, DC precision 9, Dynamic bit rates, did not select steal space from extras. Did not select dynamic bit rates. Could you post the [Status] section of the REBUILDER.INF file?

[onesoul]

Originally posted by jdobbs
I've never gotten any that small either... It's possible it may not be your encoder at all... I haven't found it yet, but I have a suspicion that I may have a bug related to recovering sectors when "Steal space from Extras" is used. I didn't use any other option like steal space from extras. As I have posted on hc thread, the encoder is undersizing comparing to hc batch 0.01.
With hc batch 0.12 I got ~3920mb (autogop on) and ~4030mb (12 2 gop), with hc 0.01 got ~4320mb (12 2).

I analysed the log file and I could verify the undersizing. The source was interlaced dvd pal.

[onesoul]

Originally posted by Oldeman
After raising target sectors to 2277000, I got exactly the same size 3.94Gb as before. Movie was preprocessed with DvdReMake to 5.4Gb.
I selected the french audio stream (364mb)to remove with with Dvd-Rebuilder.

DVD-RB v,80, HCbatch12a.
Options, 1 pass, GOP 12, DC precision 9, Dynamic bit rates, did not select steal space from extras. Did not select dynamic bit rates. This is a long shot, could you check that the french audio is deselected again with dvd-rb?

For example, I verified that when disabling audio streams with dvdshrink, if I don't deselect that same audio stream with dvd-rb, it will take in account that stream space which would lead to a undersized output (because that audio stream wasn't actually there).

The undersizing I mentioned before exists although.

[Oldeman]

Sorry, I have already deleted the working directory, but here is the rebuilder log for what its worth.

I can rerun the whole run in about 2.5 hours, if you need it.

- Detected ANGLE and/or INTERLEAVING on source.
[11:21:00] One Click encoding activated...
-----------------
[11:21:00] Phase I, PREPARATION started.
- HC encoder selected
- VTS_01: 2,831,420 sectors.
-- ANGLE and/or INTERLEAVING is present.
-- Scanning and writing .D2V file
-- Processed 187,955 frames.
-- Building .AVS and .ECL files
- Reduction Level for DVD-5: 79.7%
- Overall Bitrate : 3,394Kbs
- Space for Video : 3,247,930KB
- Analyzing VTS_01 for optimal CQ factor.
-- TargetSize (sectors):1,948,758
-- Sampling 1884 of 187955 frames.
-- Predicted size (sectors) at CQ=2: 4,961,536
-- Predicted size (sectors) at CQ=6: 1,623,209
-- Predicted size (sectors) at CQ=4: 2,482,863
-- Predicted size (sectors) at CQ=5: 1,978,964
- CQ Value selected: 6
- HIGH/LOW/AVERAGE Cell Bitrates: 4,612/917/3,394 Kbs
[11:30:52] Phase I, PREPARATION completed in 9 minutes.
[11:30:52] Phase II ENCODING started
- Creating M2V for VTS_01 segment 0
- Creating M2V for VTS_01 segment 20
[12:59:20] Phase II ENCODING completed in 89 minutes.
[12:59:20] Phase III, REBUILD started.
- Copying IFO, BUP, and unaltered files...
- Processing VTS_01
- Rebuilding ILVU Area VOBID: 1 CELLID: 1
- Rebuilding segment 2 VOBID: 3 CELLID: 3
- Updating NAVPACKS for VOBID_03
- Rebuilding ILVU Area VOBID: 4 CELLID: 1
- Rebuilding segment 19 VOBID: 6 CELLID: 17
- Updating NAVPACKS for VOBID_06
- Rebuilding ILVU Area VOBID: 7 CELLID: 1
- Rebuilding segment 20 VOBID: 9 CELLID: 1
- Updating NAVPACKS for VOBID_09
- Updated VTS_C_ADT.
- Updated VTS_VOBU_ADMAP.
- Updated IFO: VTS_01_0.IFO
Correcting VTS Sectors...
[13:10:27] Phase III, REBUILD completed in 11 minutes.

Done.
[13:10:29] PREPARE/ENCODE/REBUILD completed in 109 min.
[15:05:00] One Click encoding activated...
-----------------
[15:05:00] Phase I, PREPARATION started.
- HC encoder selected
- VTS_01: 2,831,420 sectors.
-- ANGLE and/or INTERLEAVING is present.
-- Scanning and writing .D2V file
-- Processed 187,955 frames.
-- Building .AVS and .ECL files
- Reduction Level for DVD-5: 80.3%
- Overall Bitrate : 3,417Kbs
- Space for Video : 3,269,930KB
- Analyzing VTS_01 for optimal CQ factor.
-- TargetSize (sectors):1,961,958
-- Sampling 1884 of 187955 frames.
-- Predicted size (sectors) at CQ=2: 4,961,536
-- Predicted size (sectors) at CQ=6: 1,623,209
-- Predicted size (sectors) at CQ=4: 2,482,863
-- Predicted size (sectors) at CQ=5: 1,978,964
- CQ Value selected: 6
- HIGH/LOW/AVERAGE Cell Bitrates: 4,643/917/3,417 Kbs
[15:15:10] Phase I, PREPARATION completed in 10 minutes.
[15:15:11] Phase II ENCODING started
- Creating M2V for VTS_01 segment 0
- Creating M2V for VTS_01 segment 20
[16:43:00] Phase II ENCODING completed in 88 minutes.
[16:43:00] Phase III, REBUILD started.
- Copying IFO, BUP, and unaltered files...
- Processing VTS_01
- Rebuilding ILVU Area VOBID: 1 CELLID: 1
- Rebuilding segment 2 VOBID: 3 CELLID: 3
- Updating NAVPACKS for VOBID_03
- Rebuilding ILVU Area VOBID: 4 CELLID: 1
- Rebuilding segment 19 VOBID: 6 CELLID: 17
- Updating NAVPACKS for VOBID_06
- Rebuilding ILVU Area VOBID: 7 CELLID: 1
- Rebuilding segment 20 VOBID: 9 CELLID: 1
- Updating NAVPACKS for VOBID_09
- Updated VTS_C_ADT.
- Updated VTS_VOBU_ADMAP.
- Updated IFO: VTS_01_0.IFO
Correcting VTS Sectors...
[16:54:07] Phase III, REBUILD completed in 11 minutes.

Done.
[16:54:08] PREPARE/ENCODE/REBUILD completed in 109 min.

[Oldeman]

I did not remove the French audio track in the preprocess, I selected it for removal in DVD-RB.

[jdobbs]

The problem is that you selected one pass CQ encoding... That has a very limited prediction capability... and often the size you are seeing will be as close as it can get. Changing CQ by one click can often make a huge size jump.

If you want to get accurate sizing you have to do two passes. Limited sizing is the price you pay for one pass encoding. (109 minute encoding is nice, though...)

[DK]

i have had one slightly oversized title so far but the rest was almost perfectly on spot using 012a

[TCrowe]

Originally posted by hank315
I know about the undersizing issue, tests I did also showed some undersizing.
If the source can be compressed very easily the bitrate control as implemented now has trouble to keep the desired bitrate.
ATM I'm working on a complete new encoding engine, one of the things that will be changed is the bitrate control.

Just wondering what kind of DVD's undersize under 4 GB, smallest backup I did was 4.25 GB...

It was an NTSC/DVD9, preprocessed with VobBlanker.

jdobbs: I did have "Steal Space" on when I did my test. I am going to run another test tonight.

[Oldeman]

Just re-ran movie about with DVD-RB .81 and HCBatch12a. This time used 2-pass fast option instead of 1 pass with Gop-12 DC prec-9.

Removed french audio track as above.

Final size was 4.38Gb...

Now, can we take advantage of hyper-threading on the P4s and get some speed out of this mother.

Well done, JDobbs and Hank.:);)

[johnhamler1]

did a back up of king arthur, pal version, integral version.

quality best, matrix, 2 pass, 1 click,steal space 50%-D1...(with DVDrebuild0.79&HC)(take around 8.5 hours)
came out 500 mb less than expected.(the other PAL movie(Below) I did with HC was 200mb short, quality is fantastic on my projector, so I burned it).

sound like the one click option compress too much. strange???I dont see why this setting will affect compressibility???
will try again tonight, and I will post my results...

you say I can change target size, is that for the pro version only???, I use the non-pro version)

[jdobbs]

No -- it is in both versions.

[johnhamler1]

oh good u are here on line...

so if i understand , I have to change my setting like that:
f*maxbitrate 8000
*matrix qlb
*profile best
*targetsectors= 2252000

and it should resolve my size problem, right???:D

[jdobbs]

You are confusing the HC.INI with REBUILDER.INI.

The TargetSectors= goes in REBUILDER.INI under the [Options] area.

BUT!!!! If you change it, don't EVER complain to me about oversizing.... the default setting has been tested and is the most reliable.

[johnhamler1]

ok jdobbs , I will check that when at home.

why should I complain?, if it oversizes , not a big deal!!!. even undersized, it is still the best (free)encoder and much better than a transcoder.

I did a 3x zoom on my projector (3 meters diagonal) with starwar2 last night , with DVD shrink, it is all blurry and I have plenty of artifacts. with HC, it is watchable (at 3x zoom)...and I have not seen any artifacts.

I am very excited about this new HC....

(ps : found the file to "correct", thank you)

[rendez2k]

Did my first test with HC 0.12 last night using the 'good' profile with DVD-RB 0.81 Pro. Source DVD was 7.22gb with 3 hours of programs plus about 15 mins of extras. I just used the one click option and turned off any steel space from extras options. Took just under 8 hours to complete, but I had HT turned on so HC was only running at 50% processor time (I have a 3.2gig P4). Heres two questions though. When for instance video fades in from black, it looks a little pixilated, is this something to do with low bitrates (more noticeable on the extras)? Can it be avoided? Also, it might be me (and I will check further tonight) but it looks like theres very occasionally tiny jumps or stutters in the video - has anybody else had this?

[jdobbs]

You can clear up some of the low bitrate pixelation with a custom matrix for low and very low bitrates. I haven't noticed any jitters.

[rendez2k]

I'll check the jitters again tonight, I was very tired this morning! As for custom matrix's, is there somewhere I can read up on this?

[Rockas]

I've developed a MatrixEditor for Rebuilder check:
http://forum.doom9.org/showthread.php?s=&threadid=91929

On the editor you'll find descriptions of the matrices included that may help you.

[rendez2k]

The jitters i mentioned are quite noticeable at one chapter point on my Sony standalone. However, on my other standalone it plays through perfectly, although there is some slight pixilation at that point and the camera is moving quite quickly too. Any further ideas?

[cmsoliveira]

The jitters i mentioned are quite noticeable at one chapter point on my Sony standalone.

Hi, do you mean, between chapters or during a specific chapter? I'm asking this because, i finished one backup in order to test HC, and the output is great, but between chapters i notice a little jitter. Minimum but it's there. The sound plays perfectly.

I own a sony too.

I need to make other tests because the movie was post processed so the fault might be there.

[Fishman0919]

Same here on my Sony, but play fine on 2 Pioneer.

[JohnG]

No big surprise. I haven't bought a Sony product in 10 years. Nothing but problems.

[jdobbs]

Does it only happen when using HC?

[rendez2k]

One specific chapter was worse than others, but I think there something minor at the chapter change in most or all of them. The DVD wasn't pre-processed in any way either. I've never noticed this before with CCE but I guess I would need to run the same DVD through CCE to see if its a DVD or HC related problem.

[hank315]

There's a difference how CCE and HC ends a sequence, ATM HC always writes a sequence end code.
In the ecl file I see this is disabled for CCE (seq_endcode=0), maybe this could cause it?

[jdobbs]

[Corrected] The sequence end code is only used on still frames. Is there a way you could add it as an option? It's definitely a problem, as I had the same result with CCE early in DVD-RB development.

[jdobbs]

Yep. In fact, here's a quote from the change log:

- Corrected output of seq_endcode in CCE encoding. This value was incorrectly set to "1" when it should have been "0" -- the result was an end code being inserted at the end of every cell and may have caused glitches on some players. It may be related to possible "stutter" and audio dropouts that has been reported. Thanks to RB for pointing this out. (v0.29)

[hank315]

No problem, shall create a new command which will disable the end code output, will PM you about it.

[jdobbs]

Thanks.