Sony PSP transcoding target question

[satsuke]

Just a quick question for anyone.

What are your thoughts on a secondary encoding target specifically for portable devices?

Allow me to explain.

There has been an explosion of portable devices capable of mobile video. The largest of which is the Sony PSP and Ipod video, but includes most recent Palm models, Windows Mobile devices, in the future Nintendo DS, etc.

What is required is transcoding or encoding current "production" run shows to a smaller target file size and resolution.

For example, I have a Palm T/X handheld with a screen of 480/320. It is possible to get a standard 22 minute episode down to ~50Megs per episode.

Does anyone think this is a distribution target worth persuing?

For clarification, this is proposing either a subsection of an existing tracker or just a new type of target file. If we do it, I'd only plan on distributing my groups (CA) files this way.

This all has the potential to increase where and how fansubs are seen (think waiting for cosplay at kon or on train).

Any ideas

Satsuke



edit:

For clarification, the subs are indeed readable at reduced resolution if handled properly. For the encoders reading this thread, what is required to encode is simply to reduce the resolution to 480/320 or 320/240. (possibly) reducing frame rate to 15fps, using 64kb OGG, ACC or mp3, limiting data rates to ~450kb/s video and turning off some of the xvid encoding tweaks (trellis encoding).

[Soluzar]

Fansubbing should be primarily for you, so if you want these encodes you should make them. I don't want them, but there's no reason why that should influence you.

[DryFire]

technically all avc ipod encodes should be able to play on a psp assuming you don't need to run it through atom changer as the ipod only supports baseline profile while the psp supposrt main profile (I forget the levels).

As far as readability, subs are ledigble on my psp for both 4:3 and 16:9 material. However I wouldn't dare try and make some widescreen encodes for the ipod.

I'd say it'd simply be easier for the end user to run the subs through pspvideo9, or if they are more advanced to use somehting like megui + the ipod or psp profiles sharktooth provides. I don't really see a reason to distribute them as another release.

[Zero1]

Originally Posted by satsuke

Just a quick question for anyone.

What are your thoughts on a secondary encoding target specifically for portable devices?

Allow me to explain.

There has been an explosion of portable devices capable of mobile video. The largest of which is the Sony PSP and Ipod video, but includes most recent Palm models, Windows Mobile devices, in the future Nintendo DS, etc.

I think it has potential, providing people follow it through as sort of a side service, and do the job properly, for example spec compliant encodes that people will be able to play without hassle, or files that are reasonably small in filesize. It would be a bit of fun for those people with portable devices, and maybe even dialup users would show interest in these smaller files (well hey, a "low quality" fansub is better than no fansub, I can vouch for that since I'm stuck on GPRS right now). I dare say it may attract interest from the more impatient of viewers who simply want to see it NOW, or something to tide them over until the regular release has finished downloading, particularly if it has few or no seeds.

There is one thing we would need to clarify though. The dreaded "Profiles@Levels", rather than calling a release or encode "PSP version". I would advise people find out what Profile@Level these devices support if you intend to aim for a specific device, and include that profile in the filename. Many encoders (including myself somewhat) hate being confined to a particular profile or level, but they are important for interoperability between portable devices (and not limited to either, MPEG-4 DVD players will make use of levels), especially since they are low on CPU and battery power.

So why levels? Well calling an encode the "PSP version" implies that it may not work correctly on other hardware, or if it will, you are unsure of what other hardware it will play on until you try. Calling an encode by the profile and level it complies to, means that any hardware supporting at least this profile and level will play the encode back fine. This means you can now base your encoding decisions on the lowest powered device you wish to support. For instance the iPod supports H264 encodes for Baseline @ Level 1.3 (BP@L1.3), whereas the PSP supports Main @ Level 3 (MP@L3). So if you wish your encodes to be playable on iPod, encode it at Baseline@L1.3, and it will also be playable on PSP (since PSP supports a higher level and profile). Obviously, you will be able to play both of these encodes on your PC also which is where the 56k'ers might pick up these "portable" encodes. People are getting used to H.264, AAC and MP4 now, the overall idea is becoming more feasible.

So what do these Profiles and Levels mean, or actually do? Well a simplified explanation is that the Profiles define what encoding features can or cannot be used. For example, with the Baseline profile, you cannot use CABAC (this is a bad thing ) or B-frames, but with High profile, you can use custom quantisation matrices, or 8x8 intra prediction. The levels define the characteristics, or complexity of the video, such things as resolution, or number of maximum macroblocks per second.

At this moment in time, there's isn't a whole lot of reason to put a Profile@Level in a filename for regular encodes since PC's are so different there is no set level for them, but what with some shows being broadcast in HD these days it might be a good way of telling the complexity of the file, rather than putting framerates, resolutions and codec types in the file name. It might also come in handy for the leechers of those who release with the intention of these files being playable on hardware at some point. Anyhow, this is just consideration at this point, who knows which way it will go.

There is also the convienience factor, being able to pick up a small encode of a fansub rather than spending the time recompressing the released version (yet more generation loss), or how about picking up the latest fansub on the move, when you are at work or away from home? I will elaborate.

I have a mobile phone running the Symbian Series 60 OS, on it I have:
-Opera (web browser)
-WirelessIRC (IRC client)
-Agile Messenger (multi network IM client (AIM, MSN Messenger, Yahoo, ICQ))
-ZipMan (ZIP/RAR/TAR/GZIP utility)
-yEdit (text editor, think notepad)
-yFTP (FTP client)
-Psiloc Extended File Manager (file manager, really!)
-Realplayer (self explanatory, built in. Handles MP4 with ASP+AAC)
-Smartmovie & XviD decoder (third party video player. Handles AVI with ASP+MP3, larger resolutions too)

Now this isn't "oh look how pwn my phone is", but you can see where this is going. The only thing it lacks it bittorrent Add to that a recent model (which I don't think is released yet) called the Nokia N91, which has an internal 4GB harddrive (I currently use a 512MB RS-MMC). Allowing that your phone provider has some sort of unmetered internet access (3G for example, which I believe is 384kbps+), then fansubs on the move are feasible, even if it is something low quality that you are going to delete after a view or two, it's just a bit of fun, entertainment isn't it?

As I said before in the old H.264 thread before the unfortunate happened, the savings you will make switching from ASP (XviD, DivX) to AVC/H.264 means that you can distro these portable versions and still be saving on the previous filesizes.

Also mods, this might be good merged to the H.264 thread, no?

[satsuke]

I may not have been clear, this was more aimed at a secondary tracker / section of animesuki or otherwise a resource for portable video.

I agree, marking something as PSP only would be a bad idea, maybe marked
"kawaii neko eps 3 - portable.avi" or some such.

As far as h.264 encoding, that isn't really relevant. Encoding to a certain profile would make sense, but would be up to the individual group.

Transcoding other groups releases, I already do that to cut space on my SD card already, but I'm not about to release them as my own.

[Zero1]

Originally Posted by satsuke

I may not have been clear, this was more aimed at a secondary tracker / section of animesuki or otherwise a resource for portable video.

I agree, marking something as PSP only would be a bad idea, maybe marked
"kawaii neko eps 3 - portable.avi" or some such.

As far as h.264 encoding, that isn't really relevant. Encoding to a certain profile would make sense, but would be up to the individual group.

Transcoding other groups releases, I already do that to cut space on my SD card already, but I'm not about to release them as my own.

No, I do tend to jump the gun sometimes, your post was perfectly clear now I check through it again.

Yes, you are on the right lines with marking them something other than platform specific, but portable, in my opinion is too vague, since the abilities of different portable hardware varies greatly, for example the iPod and PSP. If you throw a mobile phone into the equation, that would also be classed as portable. I'm not implying that I assume my method is correct or right, but it seems logical to me. The alternative would be to make your own grading system, something like Q1 - Q5, say a Q5 encode requires hardware of a similar spec to the PSP, or Q1 is the bottom end of the scale with mobile phones and iPods at Q2 or 3, but that's essentially a dumbed down Profile@Level system, so I'd be interesting in hearing other feasible systems

The easiest alternative, which we didn't seem to want go the way of was marking encodes hardware specific, but I guess that would save a lot of headache in the longrun, "You need this device to play it, it might work on others, if it does good, if it doesn't then we warned you"

As for a seperate tracker or something... Hmm, it's hard to say really. In some respects you could say lump them in with the current trackers like the dual XviD/x264 encodes are doing, in others you might say that the difference is more significan than a different codec or container, and therefore should be seperated. It's hard to decide, but I would be leaning towards keeping it as-is, and treating it like a dual release.

I hear you with the transcoding, though it's not something I've bothered with. The only low complexity encodes I have done recently was a DVB dump of Enter the Dragon (love Bruce Lee ) which I encoded to 208x176 for my mobile phone, just to watch during my lunchtime at work. But yes, it would be nice if these "portable" encodes were readily available so you could just chuck a load on a memory card or disc for convienience.

[gumbaloom]

There is a place for such things but rather than individual groups working on it themselves why doesnt someone or a group of people with similar interests form a group to do the releases ?. That way encoding experience / distro resources of like minded people are shared.

At least in L-E the attitude when someone tried a PSP release was laissez faire by the fandom but if another group wants to re-encode L-E releases for Ipod Video / PSP and as long as tech support queries go to them then I say go for it. it's a worthwhile venture.

I think another requirement for encoders who encode is that they test their encodes on their OWN devices before passing em for release.

Regards


-gumbaloom

[DryFire]

Well there is one question that keeps coming to mind. What would be the baseline for portable? Plays on a psp? an ipod? a palm pilot? a cell phone?

I suppose the goal is to find one setting that fits as many scenarios as possible.

As far as low filesize low overhead perhaps adding a main profile encode would be worth a shot. Even though reducing teh resolution has the largest effect on playback I don't know how many people would want to watch 320x240 video on thier computer.

[satsuke]

Originally Posted by DryFire

Well there is one question that keeps coming to mind. What would be the baseline for portable? Plays on a psp? an ipod? a palm pilot? a cell phone?

I suppose the goal is to find one setting that fits as many scenarios as possible.

As far as low filesize low overhead perhaps adding a main profile encode would be worth a shot. Even though reducing the resolution has the largest effect on playback I don't know how many people would want to watch 320x240 video on thier computer.

It would be almost impossible to baseline low enough for phones and still be usable on bigger devices. I'd have to say an Ipod would be the best choice for "entry level" on this sort of thing.

Depending on the phone, it might be able to ignore every x line or scale the image down.

As far as watching on a computer, I wouldn't want to, but 320/240 is approx the same quality as VHS tape playback and the old days of VCD playback.

[Enragin_Angel]

I guess you could do two or three profiles.

1) Cellphone
2) QVGA pocket pc and palm pilot
3) iPod and PSP

You could also have groups sign up or endorse your group for transcoding their work to a mobile format. And for groups that already encode their fansubs to a mobile format, you can host their work on your tracker.

[TheFluff]

Too many different profiles/levels and too bad quality. If people want this kind of stuff, they can encode it themselves - I certainly won't do it for them.

[Sylf]

I was trying to encode a 15-minute show so it's watchable on my phone. I still haven't found the spec it can handle. So I reverse-guessed from the type of video it captures with the built-in camera - MPEG-4, Simple Profile @ Level 1, 176x132, 15fps *bleh*

So I encoded the video in AVC Baseline Profile. (I didn't pay attention to level...) Converted the audio to aac 48kbps. Transcoding took 10 minutes. That's not too bad, is it?

The bad thing was, even at such small profile, the video stuttered in panning scenes. I need to find the setting that the phone can handle...

Maybe someone can start building a reference chart that shows what different divice can handle what kind of profile etc.

[TheFluff]

That's the point I was trying to make. There's 21526457 different ways to encode stuff to work on specific devices - pick one and you'll piss everyone off except the people who own a target device. Hence I won't ever do any official "portable" encodes (plus, butchering stuff like that is not my favorite occupation...). If people want that stuff they can figure out what their device supports and encode according to that.
However, I won't complain if a third party starts reencoding my encodes, or demuxes softsubs from releases I've worked on and uses those to encode from fresh raws. (not that I would do that anyway...)

[Sylf]

Originally Posted by TheFluff

That's the point I was trying to make. There's 21526457 different ways to encode stuff to work on specific devices - pick one and you'll piss everyone off except the people who own a target device. Hence I won't ever do any official "portable" encodes (plus, butchering stuff like that is not my favorite occupation...). If people want that stuff they can figure out what their device supports and encode according to that.

I guess I was trying to make couple of points - one was what you stated.

The other part is the "10 minutes" part - it doesn't take long at all to do it yourself.

[Zero1]

Originally Posted by gumbaloom

I think another requirement for encoders who encode is that they test their encodes on their OWN devices before passing em for release.

Agreed, this is mandatory, no questions asked. Be it for hardware players, PSP, iPod, PC, or whatever it should always be checked by at least the encoder and QC team (this is pretty redundant though since it's so obvious).

I'm not enthralled about the thought of someone transcoding my files for distro, though I don't really care if people are doing it for personal use. I can't really explain why I wouldn't particularly like someone transcoding my stuff for distro, I guess encoding is just pretty subjective and people hold themselves to different standards. If it was some of the encoders I know though, I don't think I would mind.

Originally Posted by DryFire

Well there is one question that keeps coming to mind. What would be the baseline for portable? Plays on a psp? an ipod? a palm pilot? a cell phone?

I suppose the goal is to find one setting that fits as many scenarios as possible.

As far as low filesize low overhead perhaps adding a main profile encode would be worth a shot. Even though reducing teh resolution has the largest effect on playback I don't know how many people would want to watch 320x240 video on thier computer.

This is the thing. What do you use as the lowest supported device? I would like to think the PSP for a number of reasons. First of from what I gather it's pretty versatile (I don't actually own one though, yet) and it's capabilities are good. It's also pretty popular.

In an ideal world you would find a "universal" level, but as you know the capabilities vary to widely to be able to choose a "one size fits all" profile or group of settings. Perhaps if this thing was to be taken seriously, it would require a specialist group of encoders to put out the multiple formats, but that would be really ugly

As far as encodes I do for online distro are concerned (for PC playback) I don't care about profiles or levels, I'll use whatever settings I deem useful or viable. With regards to 320x240 video, I still see that resolution a lot these days. Web trailers (for obvious reasons) and some AMVs (of which are mostly 352x240 MPEG-1, but hey). Yes, given the choice I doubt people would ever choose 320x240 over 640x480, but there is a very small minority that may benefit from the "PSP encode", those people with pre-historic CPU's that cannot handle 640x480 ASP, but just may be able to handle 320x240 H.264 with the help of CoreAVC or something.

This is kind of amusing, we talk of 320x240 as if we couldn't live without 640x480, but now the same thing is beginning to happen with HDTV & HD-DVD, we complain now that the resolution is overwhelming, but if and when it becomes the defacto standard for subbing (or at least HD resized to PAL or slightly larger resolutions as I suggest to offset filesize requirements), we will wonder how we ever lived with 640x480. Well maybe

I needn't say that my stance is to upgrade (as you will know from the H.264 threads), part of the cost of being a fan IMO, as much as buying a VCR and video tapes was when subs were on VHS, but if someone is seriously going to do a "portable" encode then at least make it decent so more than one type of user will benefit. What we should bear in mind is that technology is constantly improving and while iPod cannot handle the same complexity of video as the PSP now, it's just a matter of time. No point purposely crippling the encodes, gun for high quality now and reap the benefits a little later.

Originally Posted by Enragin_Angel

I guess you could do two or three profiles.

1) Cellphone
2) QVGA pocket pc and palm pilot
3) iPod and PSP

You could also have groups sign up or endorse your group for transcoding their work to a mobile format. And for groups that already encode their fansubs to a mobile format, you can host their work on your tracker.

There was a group that transcoded some fansubs for the PSP as the target device, I believe they used ASP (XviD) and AAC (don't know the encoder). Can't for the life of me remember the name, but I think they did He Is My Master.

Originally Posted by TheFluff

Too many different profiles/levels and too bad quality. If people want this kind of stuff, they can encode it themselves - I certainly won't do it for them.

Age for truth
I agree that there is too much of a difference between capabilities of devices, I mean look at the iPod and PSP, quite a difference there. The iPod cannot even make use of nice features such as CABAC or B-frames, screw encoding for that
Yes, it's not something I would particularly go out of my way to do, but if for instance I was encoding a MP@L3 version for my own personal use on a PSP (whether I decide to get one is up in the air), I would be willing to distro it, seperate from the regular episode.
I don't think I would purposely encode for iPod, PSP and various mobile phones, too much variation, adds to encoding time and few people benefit. This might be where the transcoding group comes into it, but I don't hold a great deal of hope. Potential yes, likely to execute well, no.

But you guys know me now, I like to theorise and waffle, it's also an excuse to do some non important encoding and mess with gadgets

Originally Posted by Sylf

I was trying to encode a 15-minute show so it's watchable on my phone. I still haven't found the spec it can handle. So I reverse-guessed from the type of video it captures with the built-in camera - MPEG-4, Simple Profile @ Level 1, 176x132, 15fps *bleh*

So I encoded the video in AVC Baseline Profile. (I didn't pay attention to level...) Converted the audio to aac 48kbps. Transcoding took 10 minutes. That's not too bad, is it?

The bad thing was, even at such small profile, the video stuttered in panning scenes. I need to find the setting that the phone can handle...

Maybe someone can start building a reference chart that shows what different divice can handle what kind of profile etc.

I haven't even considered trying H.264 on my phone yet, it has a hard time dealing with ASP

Actually, I don't know if it was intentional or not, but MPEG-4 Simple Profile is not H.264/AVC. The kind of encoder you want to be using to output MPEG-4 Simple Profile is MSMPEG4/DivX 3.11 (which should in theory work, but hacks = no guarantee) or XviD with restricted options (ie no B-frames). Speaking of XviD's profiles, they don't actually limit the encode, but can be used as a guide as to what settings you should or shouldn't use.

Well if it captures at SP@L1, it's safe to assume it can probably handle ASP@L1/2 or SP@L2/3, since when you consider it, encoding is a lot more complex than decoding, and it is also grabbing images from the camera.

Recently I resonded to a similar thread on Doom9, and it's still pretty relevant.

Originally Posted by Zero1 @ [url

http://forums.doom9.org]I[/url] unfortunately don't have any concrete specs for you, I can tell you of a few observations though.

For Symbian series 60 OS' there is a player called Smartmovie which can handle MPEG-4 ASP (ie XviD) and MP3 contained in AVI. This is obviously very convienient. Decoding speed is very good. A video will play quite smoothly at 208x176@15fps (approx 140kbps ABR) with Mono MP3 @ 16/22.05/24KHz @ 32kbps. This isn't really standard and requires a 3rd party player.

Recently however, I've been playing around with a more standard format. This time I am using XviD again, but at 176x144 (which is a common resolution for mobile phones, another is 128x96). The internal Realone player didn't seem to want to play my MPEG-1 Layer 2 stream, and nor did some media players. VLC handled it fine. In view of this I tried LC and HE AAC, both worked fine when contained in MP4. Playback was not satisfactory though, it was sluggish and seemed to skip frames. This might be from me using AAC now instead of MP3, but I didn't think it would have such a large bearing. The only other thing I could think of is that the player is just poorly optimised.

To summarise, the third party player using the XviD decoder can handle 208x176@12/15FPS with 2 B-frames + MP3 audio, but the supplied RealOne player chokes on 176x144 with 2 B-frames + AAC audio. I don't consider the AAC audio to be so much more complex that it would cause the lower resolution video to lag, where the larger video is fine. Reducing or disabling B-frames will probably be sufficient to fix it.

This shows you that although profiles can be used as a guide, that there is an amount of flexibilty in them. What I tend to do is start with a basic encode, add features as I see fit.

Well here is the 176x144 encode, using ASP (1 or 2 B-frames, I don't remember which (I could be unlazy and check in FFDShow...)) & HE-AAC. When I tested with LC-AAC, I didn't see a noticable improvement in playback speed, so it wasn't worth sacrificing the audio quality. The video looks ok on a mobile phone, but obviously at fullscreen on a PC becomes "looooool blockfest GET!" (so you don't need to remind me that I suck at encoding, I've given my excuse already )
I uploaded this on my shaky GPRS connection at the lightening speed of 1.55KB/s! so check it out
Hmm, I just noticed the first line of pixels is fugly, probably an AVIsynth filter bork.

Standards can be a farce sometimes, but I like to experiment. Grab the file here.

Originally Posted by TheFluff

That's the point I was trying to make. There's 21526457 different ways to encode stuff to work on specific devices - pick one and you'll piss everyone off except the people who own a target device. Hence I won't ever do any official "portable" encodes (plus, butchering stuff like that is not my favorite occupation...). If people want that stuff they can figure out what their device supports and encode according to that.
However, I won't complain if a third party starts reencoding my encodes, or demuxes softsubs from releases I've worked on and uses those to encode from fresh raws. (not that I would do that anyway...)

Again true. I'll only encode stuff if it suits me to do so. It would suit me to encode stuff for mobile phones but:

a) The resolution is shitty (hard to read subs, well barely readable)
b) CPU power is low, which means crippled encodes
c) Not a whole lot of people would be bothered about watching it on a phone anyway, not until they improve significantly.

Actually "mobile phone culture" is a big thing here in the UK, so if anime was more popular here (it's still pretty underground I guess, owing to us being in PAL land, and not getting any good releases. There are some fans crawling out of the woodwork though) it might be something that would slowly pick up. I see requests for "PSP video" on the animemusicvideos forum a bit recently, also.

Anyway, it's the fasion statement to have the latest, smallest phone out with MP3 ringtones of some popular music or something. A bad thing that has stemmed from it though is "happy slapping" which is basically beating up random people and filming it on the phone's camera, and sharing it via bluetooth. A lot of "happy slapping" incidents have resulted in serious injury or death.

[DryFire]

Subs are borderline readable on my PSP (a lot of signs tend not to be). I wouldn't even bother for anything less then a psp.

As far as doing it officially... I too don't tink it's anywhere worth the effort.

[gumbaloom]

Like I said.

PSP / Ipod releases are a relatively niche market compared to mainstream fansubs. So rather than trying to make multiple fansub groups waste time on re-encoding stuff, whoevers interested in this sort of thing should setup a channel on rizon and let people join and work something out.

If someone does setup a re-encoding group I don't think you'll get any hassle from the majority of groups as long as you don't
a) portray yourself as the group who did the original subs
b) let any tech support queries land on the doorstep of the original subbers and deal with all of that in house.
c) give credit to the right subbing group where multiple groups are working on the same show.

-gumbaloom

[mangatron]

Eh, and here I thought I was the only one who watched anime on my Palm/PPC devices. I agree on the part about naming the file as "portable", seeing as how I wont be using a PSP for that

Well, I don't watch subs so I can't comment about the subtitles, but I make my own files and encode it to WMV (for my PPC). Some background text in anime are readable at 320x240, so I think that size is best on quite a number of devices.

[[darkfire]]

I encode to my psp all the time. If you want profesional quality i suggest you resize the video in vituraldub first before you encode. Resize to 320x240 with lancoz3. Xvid at quantization level 1. You dont want any compression untill the last step. get pspenc. Its a gui interface to ffmpeg it allows for 2 pass encoding and target size or bit rate.

http://jonny.leffe.dnsalias.com/pspenc/index.htm