Please help! QuEnc max GOP size 600

[Menco]

Hi all,

I'm using QuEnc to create some special MPEG streams. I noticed that QuEnc can create GOP's of maximum 600 frames. Although this is already much larger than normal, I actually would like to produce streams with much larger GOP's. Is there some way to make this larger?

I really need this. Can someone help?

[dragongodz]

I really need this.
MY GOD!!! you NEED a GOP bigger than 600 frames ? what the frel for ?

[Menco]

Well, I try to study some encoder/decoder behaviour.

Does enyone know if there is an encoder where i can have more control over the size of the GOP. I mean, it shouldn't be that hard to change the limit of 600 in the source code of the encoder.

[Mug Funky]

you might want to try ffmpeg (not sure if that has hard-coded GOP limits).

how large a GOP were you thinking of? are you trying to measure cumulative decoder inaccuracies or something? because it's likely on a sane video sample that even on a very long GOP, the whole image will be completely rebuilt with intra-coded blocks many times over before the next I-frame.

unless you're doing a static test pattern...

[dragongodz]

I try to study some encoder/decoder behaviour.
what study requires a GOP size bigger than 600 ? if something hasnt happened by that point the chances of it happening a few more frames on is rather slim.

so no unluss someone gave me a really good reason i wont change the GOP size limit in QuEnc and i have doubts Nic would see any sense in changing it aswell.

[Menco]

Well for all the curious members. I'm testing some decoders that can do an P-frame start of the movie. They claim to use some P-frame start with error concealment to make pictures look quite nice. I want to see what theyr're doing. For this matter a GOP of 600 is quite small. Especially because my set-up prevents me from starting the video and starting the decoder afterwards very quickly. So 600 frames are gone in 20-30 seconds. Not so much after all for me.

[Mug Funky]

as a wild guess, with access to the first few frames, it could work backwards and build frame 1 from the I-blocks present in the first few frames. any remaining holes could be filled via interpolation of neighbors.

so it's a streaming thing, yeah?

can you say a little more about your setup? btw, it might be possible to simply cut the i-frame off the start of an m2v.