I have posted several thing about AVCHD in the past (search AVCHD) and have essentially given up on PD7's AVCHD rendering. I render to MPEG2 (Blu-ray template) and this is almost indistinguishable from the original AVCHD files but of course lose the benefits of file size.

pjc, just in case you didn't read it:

Quote: I would welcome either pjc, Keyan or Juan posting their enquiry and comments under a NEW topic of "AVCHD rendering" or some such. I'd also request the three of you provide a "pure" 1 to 5 second unedited avchd video and then render the same video in PD7, making both avchd's available to me to look at by uploading to an ftp and I'll get them looked at by R&D etc at CyberLink.

Let me know when you can and have these available - I'll send you the ftp info. Make sure all file names of the video relate to YOUR NAME and can be identified easily.

Let's see if this issue can be resolved.

Dafydd

In reference to the poor "AVCHD final render quality" using PD7:

- IMHO the best solution to this issue is to fix Smart Video Rendering Technology (SVRT) for AVCHD files. This is the only way to preserve the quality of the original videos.
- pjc, rendering to mpeg-2 present several disvantages. First, increase in file size. Second, the re-encoding always results in lost of quality. Third, longer rendering time. fourth, the re-encoding always results in lost of quality (yeah, it's bad, inadequate, unsatisfactory, deficient, malignant, carcinogenic, etc), sorry I can't stress this enough.

Dafydd, to reproduce the AVCHD issue you need at least two videos (or one video twice), because SVRT does NOT work after a transition or effect is applied. If only one video is used, then the video rendered in PD7 would be the same as the original (smart rendered)

I will produce a 5 sec video, I don't have one right now.

Juan

Quote:
- pjc, rendering to mpeg-2 present several disvantages. First, increase in file size. Second, the re-encoding always results in lost of quality. Third, longer rendering time. fourth, the re-encoding always results in lost of quality (yeah, it's bad, inadequate, unsatisfactory, deficient, malignant, carcinogenic, etc), sorry I can't stress this enough.
Juan

Juan, I can't agree more but in my workflow MPEG2 is the only real option at the moment.
SVRT works as you say for the first clip for my Canon but SVRT doesn't work at all for my Pana SD9 due to the slight quirks Pana have used when using their own interpretation of the AVCHD codec.
So at the moment I am stuck until they sort it out.

As for the clips to Dafydd - I cannot believe Cyberlink wouldn't already have access to clips of all the main players in AVCHD cameras. But I shall PM him and send a clip from both cameras.

To keep people informed of the development of this issue, as requested I have sent the following PM to Dafydd:

In relation to the poor "AVCHD final render quality" using PD7.

As requested, I have ready to be uploaded:
- A "pure" 5 second unedited avchd video from my Canon HF-10 camcorder (11 MB)
- The same video rendered twice in PD7 (17 MB), using the AVC.MPEG-4 default profile type: AVCHD 1920 X 1080
Please note that to reproduce the AVCHD issue you need at least two videos, because SVRT does NOT work after a transition or effect is applied.

please send the ftp info.

Thank you for your time,

Juan

Quote: SVRT works as you say for the first clip for my Canon but SVRT doesn't work at all for my Pana SD9 due to the slight quirks Pana have used when using their own interpretation of the AVCHD codec.
So at the moment I am stuck until they sort it out.

This is good information, what canon camcorder do you have?

HF100, not technically mine but I use it a lot - I like to have two angles for lots of stuff.

Hi Juan,

The number of avchd examples you provide to show the issue is NOT a problem - when I can get the ftp location to function! IF you can upload a longer clip that would be even better. However you have to give a BEFORE and AFTER of each clip - basically upload 2 clips each time, ALL clips must be identifiable to you as the source of the material. The main problem will be the size of clip and your time. A 30 second clip(s) would be excellent. I appreciated the need to add a transition to an edited clip to show the issue - please do so.

I have PM'd everyone involved here, all know what's expected and the present situation.

Dafydd
[Moderator]

My experience:
Rendering AVCHD files from a Sony HDR-SR11 camera provides mixed results. This using the max settings of 14.800 bps for the average bitrate and 20.000 bps for the max bitrate.
Files with an average of 15.8 Mbps are trimmed and combined with SVRT without loss off quality. However files with a bitrate above 16 Mbps are not recognized by SVRT and the result when trimming and combining clips is a bitrate of 14.5 Mbps with much longer rendering times and subsequent loss of quality.

I don't know why the original SVRT for AVCHD thread is no where to be found, beside the original link that I bookmarked (http://forum.cyberlink.com/forum/posts/list/30/2993.page)

So here is my work around to get it to work with files encoded at more than 16Mbps.
Change the Profile.ini (in App data\Cyberlink\PowerDirector\7.0\) to change average bit rate to 16000000, min to 12000000 (or 8000000) and max to 20000000 (or 24000000).
All my clips (recorded with a Sony HDR-SR11) can be rendered successfully with SVRT.

However, if you change the clip in anyway e.g. start with a transition effect, trimming, etc, SVRT will only works to the first point of edit of the clip. That is if a clip has a transition at the end, SVRT will work all the way until the transition starts (i.e. if you use a transition at the start of the clip, the WHOLE clip will NOT be rendered with SVRT). Trimming, even 1 frame, at the start will cause the WHOLE clip to be rendered WITHOUT SVRT.

Example: you have 6 clips A B C D E F in the timeline like this:
A - Fade transition - B - C - D - E (trimmed first x frames off) - F (fade out)
* A and F will be rendered with SVRT all the way till the transition kicks in
* C and D will be rendered totally with SVRT
* B and E will NOT be rendered with SVRT at all

I think the problem with SVRT is that it doesn't recognize that trimming/transitioning at the beginner of a clip will only change several frames (not the whole clip!) so the rest of the clip can be copy over to the new vid wholesale!

I believe improving this is as easy as
1. If (whole clip can be rendered with SVRT) then SVRT=ok
2. Detect the frames that are edited by the user (e.g. transition, trimming, etc) NOT whether the bit-rate or any non-user-changeable craps fit certain criteria
3. Render the frames that are changed
4. Copy the rest wholesale

So easy, anyone with basic programming knowledge + the source code can fix it! And I thought that's how SVRT should have worked to begin with!

My original clips are from Sony HDR-SR11 - 1920x1080 - Dolby Digital 5.1, 448Kbps - Average bitrate 16.15 Mbps

Normally SVRT does not work - everything is re-encoded which degrades the quality and results in an average bitrate of 14.35 Mbps.

The above mentioned solution works for me if I do not tick the Dolby Digital 5.1 box. But then the sound of the output file is degraded from DD5.1 with 448Kbps to Stereo 384Kbps.

If the DD5.1 box is checked then all segments are re-encoded

I can upload some files before and after rendering if needed

H
Please amend your Profile to display a (fuller) contact name.
Dafydd
[Moderator]

I have a Canon Powershot TX1 which produces motion jpeg .avi files. These are pretty large so I purchased power DVD and started using it a few hours ago to encode these to AVCHD files to place on a DVD and play in a blu ray player. The trial software seemed to work fine on the shorter clips I used, especially as I did not really use any complex special effects beyond subtitles.

I was surprised when actually editing for real on my purchased version having applied (begrudgingly because of the slight quality loss) the video stabiliser to correct camera shake in one of the clips. The segment was 11 minutes long and comprised about 10 clips. Interestingly the clip just prior to the anti-shake function and the clip with the anti-shake function exhibited these artifacts in about 70% of the footage (click to enlarge).



Even after removing all the other clips apart from the 2 exhibiting these artifacts, and removing the anti-shake filter, the problem still remained. Opening up a new project and bringing in exactly the same clips, in the same order and rendering again to AVCHD resulted in no artifacts. Then adding the anti-shake filter to the same clip and rendering once more to AVCHD resulted in no artifacts as you can see (click to enlarge).



The only thing that was different with the bad render result is that I applied and removed the anti-shake filter a few times, plus changed the strength back and forth in edit mode, and used smart music....so a number of different operations before rendering to AVCHD, much like you would do when editing. With the good render, I simply bought the clips in and applied the anti-shake, then rendered to AVCHD

The difference between the two clips is exactly the same as I am seeing in the full motion video. The AVCHD quality is 1920x1080 and dolby 5.1 selected. My input files are 1280 x 720 (720p) motion jpegs in NTSC 30 fps.

The point of my post is to see if other people have had similar/same errors that don't necessarily easily "reproduce", as these are the hardest errors to fix and have you found ways of avoiding them?

I guess the other reason for posting is.....check your output files thoroughly, just because it worked on one set of edited material doesn't mean it will next time, especially when effects are being added and taken away (obvious I suppose).

Oh and hello to everyone on the forum. Intel Quad Core, NV 8800GT, Vista (32bit), Canon Powershot TX1. Editing 1280x720 (p) NTSC 30 fps motion jpeg in .avi container....to AVCHD DVDs for Blu Ray playback

Hi David & welcome,

The clips you present seem to exhibit interlacing artifacts probably because you are converting progressive footage to interlaced. We all know there shouldn't be a problem with changing progressive to interlaced but it looks superficially that the field order has been switched and the first field from one progressive frame has been combined with the second field from the previous frame.

What the rest of this thread is mostly about is introducing blocking and compression artifact with resultant loss of "definition".

The artifacts I get are reproducable whether the clip is an individual with no "editing" or in a large project with losts of manipulation. Thanks for adding some more interest to this topic. : Nice camera BTW

PJC: Thanks, yes the TX1 is nice, it meets my main criteria for a video camera, very small and nice quality output. A few irritations such as internal reflections in the lens assembly, so purple halo effects on very bright areas and into sun. However the (massive) motion jpeg files more than make up for that...I am so glad I don't (yet) have an AVCHD camera. The TX1 files are huge and horrible to handle because of their size, but they do render down nicely (very nicely!), even having to render/upscale the damm things to 1920x1080.....gbut I digress:

The issue I raised, may have a bearing/be related to on the quality issues. Obviously with an advanced filter on the PC media player I can remove these occasional artifacts and watch "normal" looking playback. Also, as you would expect playback of the AVCDH DVD on a blue ray player, also shows no "combing" artifacts.

However, the actual rendered, interlaced output at these specific areas is slightly degraded, hence the need for a more advanced deinterlacer to view the output with no obvious combing on the PC. When viewing on the blu ray player, I can see the slight degradations at these spots.

There is possibly something (I don't know what) wrong with the AVCHD renderer, this problem seems to come out more in longer clips, almost as there is some sort of relationship. Certainly the more effects or transitions that are included, the more likely the problems are to happen and I don't mean specifically at the effect or transition areas.

I would think that if this can happen when re-rendering AVCHD output the results could be truly poor, luckily I have motion jpeg, so it's momentary a few seconds or so.

Reading the comments about smart rendering and trimming existing AVCHD files, I think Cyberlink might do well to recheck some of the basic coding.

P.S. It is irritating that it doesn't have the ability to produce progressive output for the AVCHD files (which I believe is in the AVCHD specification), especially as the 1280x720 (p) is an option in the produce tab (grrr ). Having the files interlaced is irritating. Intel Quad Core, NV 8800GT, Vista (32bit), Canon Powershot TX1. Editing 1280x720 (p) NTSC 30 fps motion jpeg in .avi container....to AVCHD DVDs for Blu Ray playback

Quote:
Reading the comments about smart rendering and trimming existing AVCHD files, I think Cyberlink might do well to recheck some of the basic coding.

I bought a S1IS as soon as they arrived because of the great MJPEG video resolution. Canon certainly put a lot of effort into the video side of their still cameras.
I must agree with all of your post bit especially the encoding engine. I could do without SVRT if the AVCHD rendering was sorted (best to sort both really).
Sorry to assume you were not talking about the same problem - the stills didn't show the effect as well as the interlacing problem .
Anyway the more people who point it out the better so thanks for joining the cause.

I think I've said my piece on AVCHD editing in a half dozen previous threads, but I thought I would add my name to the list. If you need more clips I can see what I can pull together. Intel Core 2 Quad, 4 GB DDR2 RAM, Vista 64 bit, ATI Radeon HD 3850, Canon HF100

Quote: I think I've said my piece on AVCHD editing in a half dozen previous threads, but I thought I would add my name to the list. If you need more clips I can see what I can pull together.

The more the merrier,
the more the point is made!

I have today sent requests to you William and David. Please complete the report form, provide all the details requested and supply two avchd video clips.

Thank you.

Dafydd

Dafydd,

I have an AVCHD camera so I can send clips as well. Just let me know exactly what you want and what settings to take the video at.

Jim Q9300 2.5 GHz
4 GB Ram
Nvidia 9800 GT

Quote: Dafydd,

I have an AVCHD camera so I can send clips as well. Just let me know exactly what you want and what settings to take the video at.

Jim

Thanks Jim, details have been sent to you.

Dafydd