PyTivo was very kindly modified by its author and I was able to transfer very high change Hidef files to the S3. For example:
was ragged as it started up, and as it ended, but there were no blockey breakups during any of the fast motions in this short hidef clip.
has two files:
HV10-05, which is a very low datarate 1440x1080 file
HV10-12, a train passing by at high speed, also 1440x1080.
Amazingly, PyTivo's mpeg decoder (FFMPEG) actually handled the train quite well. The low datarate flower was a great hidef picture.
The downsides I found- I had to lower the datarate setting down to "6Mi" which I assume to be average VBR rate. If not, I can't understand how the picture would be even vaguely as crisp as it was. Even if this is an average rate, from what little I know, I would think this is way too low for a genuine Hidef picture.
What FFMPEG appeared to do on fast motion was two things- on the passing train, the detail of the foreground portion of the train was grainey, while the stationary objects were razor sharp. Secondly, during some motion on a camcorder shot of my daughter's classroom, the frame rate appeared to go down slightly- like maybe to 20fps- as if they just gave up and started repeating frames.
I also saw blockiness at the very beginning of the Crawford test file, and at the very end- I chalked that one up to some transition problem- either the encoder, or the S3, but from my armchair, I'd wager much easier to deal with.
The largest problem occured while I was playing back the train file before it had completely transfered. I got pretty rough with it, FF'ing up to the point it was just delivering, then triple REW back. Anyway, something happenned to the S3's ability to play back audio from any show from the Now playing list. Most of the time everything was just silent, but on test files, audio would play very intermittently. Video was unaffected. I tried a few things to get it back to normal, but gave up after not too long and restarted. Everything was back to normal after that, and I could not repro by simply FF'ing and REW'ing in an already transfered show. Something about an in progress transfer had some bugs. The workaround for such a bug is pretty easy- just start the transfer 10 minutes early and you'll never get up to the "just loaded" bits where the problem seems to occur.
In files I first encoded using a ulead media studio, I could get 20mbps (max VBR) rates on the daughter camcorder as well as CGould's charm video. The Train one I never did very well at- I couldn't even get 15mbps to work, but it wasn't like I knew what I was doing either.
Anyway, for all I know the S3 will be tuned for such transfers and FFMPEG will be able to transfer at much greater rates.
As of this point in time, what I have seen in these tests is that workable functionality for even Hidef files could be expected for TivoToGoBack transfers. Certainly, what is workable for gearheads like us is nowhere close to where products have to be for civilians. I have no way of telling how close they are. I have related the data I have collected. It's probably not enough to draw conclusions from, but what the heck- this is the blogosphere.
If Photos would be helpful on the graininess/ what this blockiness is that I am talking about, I can shoot a few more illustrations. They may well have more to do with the encoder than they do the S3, so I hesitate to document these artifacts. I could also take some shots of the Crawford sequence so that comparison could be made with the file played on a computer.