Music Videos = Fuzzy Video

[Red Hat]

Whenever I capture a music video, my video looks really fuzzy in any program I capture in. I prefer VirtualDub with YUY2 color space and Huffyuy codec. I also capture in uncompressed audio.

[junkmalle]

Capture higher res? Don't deinterlace?

[Red Hat]

I capture in 640x480 and always use deinterlace when I encode finally files. It doesn't look any better without it.

[GreyDeath]

I make Music Video discs too from my old tapes and/or off the PVR. Some samples of the raw footage and what it comes out looking like would help better describe the problem. What is your source material, what are you going to be viewing it on?

[junkmalle]

Try downloading and using one of the Smart Deinterlace filters for VirtualDub. The blend algorithm of VirtualDub's built in filter just blurs the two fields together.

If your final destination is DVD then capture interlaced at 720x480, 704x480 or 352x480, ignore the comb lines, and convert to interlaced MPEG2 for DVD. Just be sure the program you're using to convert gets the source field order right. And use a high enough bitrate to maintain quality.

If you're final destination is a Divx AVI file you may want to deinterlace with a smart deinterlace filter.

What you need to understand is that each frame you capture contains two pictures, half of each one (every other scanline, called a "field"), taken at two different times (1/60 second apart). If nothing moved during that time interval the two pictures make one clean image with no comb lines. But if anything moves you get comb lines where the motion is.

You don't see this on TV because you only see one field at a time. By the time you see the next field the first one has faded away. If you burn the interlaced video to a DVD correctly the DVD player will play them back the same way, one field at a time, in the correct order.

Deinterlacing is an attempt to make one picture out of the two half pictures. The methods of doing this involve blurring the two pictures together (blend deinterlace in Virtualdub), throwing away one picture and duplicating (or interpolating) the scanlines of the first to fill in the space (duplicate field N in VirtualDub, or discard field N followed by a resize), or some smart procedure where the deinterlacer tries to deinterlace only the areas where there is motion (with more or less success).

[Red Hat]

I use deinterlace, temporal smoother, and resize fliters and get the fuzzy video and it looks kinda fuzzy on the uncompressed file but if it's 320x240 then no it's not fuzzy.

[lordsmurf]

Quit deinterlacing.
You've not once said what the goal of these is. Burn to DVD? VCD? What?

[Red Hat]

My goal is just to have a DivX AVI with 512x384 resolution, 23.976fps, 950kbps video bitrate. The audio I use is MP3 128kbps 44KHz CBR audio.

[junkmalle]

Try:

Smart Deinterlace (http://neuron2.net/smart/smart.html)
Temporal Smoother (not more than 7)
Resize (Lanczos3)
Divx of Xvid (2-pass VBR)

[Red Hat]

How do I do the 2-pass VBR?

[junkmalle]

How do I do the 2-pass VBR?

Set everything up the way you want. Set Xvid to "Two pass, first pass". Save as AVI. When it's done go back to the Xvid config dialog and change to "Two pass, second pass". Set the Target Size to the desired size (remember that's the size for the video only, the audio will make it larger). Save as AVI again.

I think Divx calls it "multi pass, first pass" and "Multi pass, N'th pass". You can perform more than two passes but the quality doesn't improve much beyond the second pass.

Note the AVI file from the first pass may not contain any video.

[Red Hat]

Music Videos come out fuzzy, everything elese comes out perfect.

[junkmalle]

Can you post a short sample of your source? Or a before and after frame exported from VirtualDubMod?

Does your video look fuzzy in VirtualDub's output window before compression? or is only fuzzy after compression?

[SatStorm]

Music Videos have lot's of action and movement. So, more bitrate is needed

[junkmalle]

Another thing that kills video compression is strobe lights.

[GreyDeath]

If your source is a video tape, it's always going to look fuzzy on the computer. Just the nature of the source material. It can be cleaned up a bit, but there's only so far you can push it.

I agree, more bitrate is probably needed too. Video Noise is more work the compressor has to keep track of.

[Red Hat]

Music Videos have lot's of action and movement. So, more bitrate is needed

It's a show theme song and I don't want to give a high bitrate just so the theme song wil look good but everything elese will look great and I tried it in 3454kbps bitrate and it still looks fuzzy. Btw are these the correct Huffyuy settings I should be capping in?

[GreyDeath]

Oops, just noticed this is under the Capture Topic and not Authoring... My Bad.

I don't think there's alot you can adjust in the Huffy Codec, I stopped using VDub for capture when I picked up the ADVC-100. What card or item are you using to capture with? Are there settings in there you can adjust?

[Red Hat]

ATI Tv Wonder Pro.

[junkmalle]

HuffYUV settings aren't your problem.

Have you tried the Smart Deinterlace filter I suggested earlier?

Does your video look fuzzy in VirtualDub's input window? In the output window? Is it only fuzzy after compression?

I still recommend you post before-and-after frames or short video samples.

[Red Hat]

It is fuzzy before compression but not as much.
http://s60.yousendit.com/d.aspx?id=39U78VEAN5MZB1BR3BYFOQFA07

[junkmalle]

You are definitely suffering from too little bitrate -- there are lots of macroblocks all over the place (4x enlargement):



Notice how there is much more action in your "music video" clip (things are moving, the camera is panning and zooming, lots of quick scene changes, etc). Hence the need for a higher bitrate. Try using Quantization mode with a Q value around 3 or 4. You'll probably find that acceptable.

Other things to consider:

If your source is from satellite or digital cable it may have lots of macroblocks to start with (because they're transmitted as low bitrate MPEG). You can try using a deblocking filter like MSU Deblocking Filter to clean them up a bit.

Too much temporal filtering may be responsible for some of the blur.

Some parts of the video are blurry even on the master film or video tape because of the high motion.

[Red Hat]

A friend of mine encoded and capped the same theme song and it looked perfect so it can't be that. He also had a all-in-wonder card.

[GreyDeath]

Wow! Picasso would be proud. Or is that Monet?

I suppose the question would be what are your setting for the TV Wonder at, or could it be something with your system. I know when I had a Radeon AIW, I really wasn't satified with the results, but that was years ago (2-3? :P) I don't know what their products are like now. :/

[junkmalle]

I noticed your music video had the fourcc DIVX (Open Divx 4) whereas the other one was DX50 (commercial Divx 5 or 6).

[Red Hat]

I get good results with everything elese except the music video one.

[junkmalle]

I recommend you use the latest Xvid, single pass, Quantization mode. If you want the output to look exactly like the input (the output pane in Virtualdub) use Q=2. If you can live with a little macroblocking use Q=3. If you can live with a more macroblocking use Q=4 or more.

[junkmalle]

I noticed your video files were 23.976 fps. I suspect you did an inverse telecine that did not work well (I noticed duplicate frames and missing frames) and then followed with a smart deinterlace to remove remaining comb artifacts. This may be responsible for some of your blurriness.

[Red Hat]

I always do a 3:2 pulldown and I will downloaded the lastest XviD now.

[junkmalle]

I always do a 3:2 pulldown

You have the terminology backwards. Creating 29.97 fps interlaced video from film is 3:2 pulldown, or telecining. Recreating the original progressive film frames from 29.97 fps interlaced video is called Inverse Telecine.

TV shows are often shot on film, telecined then edited. This leads to lots of breaks in the telecine pattern -- especially in the intro since it has lots of very quick cuts.

You are probably using the "Reconstruct from fields, adaptive" option to IVTC in VirtualDub. It doesn't work all that well. That's why you're seeing comb lines after IVTC, and jerky video.