Reply to topic

Rendering with sameq / vcodec copy option

gefazetec
Registered Member
Posts
5
Karma
0
Hello,

I'm using Kdenlive since 0.7.8 and i am amazed by the great capabilities of the soft and particularly the ffmpeg integration. Most of the project i've managed were conducting DV to mpeg2/4 rendering, so transcoding and losing image quality was part of the game without any trouble.

From now on, i'm considering mpeg2 to mpeg2 rendering because original files are mpeg2. So now the idea is to keep image quality as close as possible to the source file. So there are my questions :

First issue : Mpeg2 lossless preset/profile may keep closest quality from source file when rendering. Is there another possibilty to get a fast rendering closest to original quality by adjusting ffmpeg options and particularly sameq?

Second issue (a development plan) : A "classic" timeline contain at least 70% of unchanged video (i.e. no effects, no transition, no sound mod, NOTHING AT ALL). Is it possible in this case to automatically detect (or manually set) those parts of unchanged video in the KDNLIV's timeline to get the render really faster, equal to a cat command in the terminal, considering that the destination file is the same codec/quality. The rendering process would be made of successive rendering phase (image vectors redefinition) when effects detected and catenate phase when nothing and hopefully faster without quality matters. This is actually possible inside and outside the soft by catenate/vcodec copy but would be great only in the soft (even by hand) : is it there anyway to tune this up in the current version by a magic trick i would be very glad to know.

Thank you very much.
capslock
Registered Member
Posts
290
Karma
0
OS
First question: Working with mpeg2 sources, you could use the original material and then render with the same bitrate as the original clip has. There might be some finetuning parameters - I don't know them, sorry. You can try to render to h.264 which gives quite good results when using presets (see ffmpeg manual). But you will always have a loss in quality - more or less visible.

Second question: kdenlive always renders the complete timeline, regardless of changes to the source. Rendering only the parts that changed is called smart rendering and has the downside that you sometimes see a quality difference when coming from copies parts to rendered parts. So, having smart rendering, some of the users (of cinelerra for example) put an effect over the complete timeline to get rid of the quality difference. I did that, too and never saw a downside in kdenlive's rendering the whole timeline - it even saved me an additional editing step.
gefazetec
Registered Member
Posts
5
Karma
0
Thank you for your reply.

For the first question, i render mpeg2 to mpeg2 project with the mpeg2 lossless preset and then output file as the same bitrate as input. The results are good and i can't even detect a sensible quality loss. I should precise Input files bitrate are over 12000kbits/s. And i'm force to agree there will always be quality loss across successive rendering. But i was wondering another possible way, you confirm my beliefs.

Your second statement keep my attention. Do you know a way to proceed smart rendering in kdenlive? It would be great.

In fact my background idea is to batch encode DV files which are desperate space eaters in mpeg2 file near 12000kbits/s. The space gain is about 3 times lighter and quality in just slighty lower. Then i come to kdenlive and render in lossless preset as fast as possible. The final destination file is then requantize to whatever bitrate considering the final project size and media destination. The idea was to get rid of long transcoding/rendering time and introduce requantize in the process which is far way faster process.
capslock
Registered Member
Posts
290
Karma
0
OS
Sorry if I was misleading: There is definitely no way for smart rendering in kdenlive.
gefazetec
Registered Member
Posts
5
Karma
0
Ok, i thank you for your reply. From now on, i'm minding about requantizing, it's so bad that a great tool such as ffmpeg doesn't assume this algorithm. I would be very interested in having this possibility in the rendering profiles of KDNLV.

Thank you very much.

 
Reply to topic

Bookmarks



Who is online

Registered users: ab4bd, Baidu [Spider], BeS, Bing [Bot], Exabot [Bot], GeekQuack, ggael, ghevan, Google [Bot], google01103, igorm, jensreuterberg, kainz.a, maarten, mmistretta, Nuc!eoN, orbmiser, plfiorini, scummos, searchfgold6789, SecretCode, Steve T, TheraHedwig, Uri_Herrera, vblazquez, Yahoo [Bot]