chilledinsanity
22nd July 2011, 23:54
This topic could be under a couple sections, so I'll try here and it can get moved if it needs to. I'm doing some video creation from videogame footage and it undergoes several steps. I tend to lose color through each of them from having to convert back and forth. While I think my process makes it impossible to completely preserve color, I would appreciate some advice for trying to make it as good as possible given the circumstances. Here is my process:
1. Output the videogame footage directly with the engine via Xvid. While the game video itself is RGB, due to AVI limitations of the engine (I can't have files bigger than 2GB and I'm recording at 180fps without dropping any frames), I pretty much HAVE to record with Xvid. I can't avoid this.
2. Once recorded in Xvid, I run an AVIsynth script to downsample the footage to 30fps, which adds motion blur in the process. I usually use "ConvertToYV12" to go with this, though now I'm not sure that's the best option anymore
3. I encode the AVS script via Virtualdub. I've recently started using Lagarith as I couldn't get 64-bit HuffyUV working on 64-bit windows with 64-bit Virtualdub. I'm not entirely what settings I should be using to encode with for Lagarith, it seems to all produce similar results after step 4.
4. I edit the footage in Adobe Premiere, which I'm assuming is converting everything to RGB.
5. I output the footage uncompressed (since I don't trust Premiere's encoding abilities at all).
6. I convert the footage back to Xvid for distribution, or perhaps in the future x264.
As you can tell, this is something of a nightmare of colorspace conversion and some color fidelity is definitely being lost along the way, so I would appreciate any help or recommendations I can try. I'm not expecting magic, just something to maybe improve it a little.
EDIT:
IF it helps at all I can use fraps or something to get a video sample of the exact same scene with all the color before all the encoding processes I go through if there's a way to correct the values against the final footage.
1. Output the videogame footage directly with the engine via Xvid. While the game video itself is RGB, due to AVI limitations of the engine (I can't have files bigger than 2GB and I'm recording at 180fps without dropping any frames), I pretty much HAVE to record with Xvid. I can't avoid this.
2. Once recorded in Xvid, I run an AVIsynth script to downsample the footage to 30fps, which adds motion blur in the process. I usually use "ConvertToYV12" to go with this, though now I'm not sure that's the best option anymore
3. I encode the AVS script via Virtualdub. I've recently started using Lagarith as I couldn't get 64-bit HuffyUV working on 64-bit windows with 64-bit Virtualdub. I'm not entirely what settings I should be using to encode with for Lagarith, it seems to all produce similar results after step 4.
4. I edit the footage in Adobe Premiere, which I'm assuming is converting everything to RGB.
5. I output the footage uncompressed (since I don't trust Premiere's encoding abilities at all).
6. I convert the footage back to Xvid for distribution, or perhaps in the future x264.
As you can tell, this is something of a nightmare of colorspace conversion and some color fidelity is definitely being lost along the way, so I would appreciate any help or recommendations I can try. I'm not expecting magic, just something to maybe improve it a little.
EDIT:
IF it helps at all I can use fraps or something to get a video sample of the exact same scene with all the color before all the encoding processes I go through if there's a way to correct the values against the final footage.