Here’s the process for an improved capture of Time At The Top (1999) I recently did using a recently serviced prosumer S-VHS editing deck with TBC (Time Base Corrector)- the Panasonic AG-1980, and a newer capture card- a Hauppauge HVR-1150. The original cap I did in 2009 and is mostly noise and I can’t believe I used to consider that quality ok…
This was done before VEAI 1.8.0 was released. Artemis MQv9 might find some more horizontal detail, but as is it looks great to me, so I’m not revisiting this one. VHS spec is 250 lines horizontal which I think means 500 pixels. The noise levels were surprisingly low and only QTGMC in the post VEAI script was necessary. If I were to do it again, I might omit Warpsharp and instead try vsMSharpen. I really don’t like the way Warpsharp changes the video. I’m glad I used only a light touch of it.
These are my notes for the process and every decision was made by first asking, what will this do to the quality? Modify to suite your particular case:
Capture at 720x480 29.97fps with Virtualdub 1.8.7. Use HVR-1150 built in audio input (not PC sound card). Format and reinstall OS (XP SP3) and everything if any dropped or added frames… This took 10 tries before it worked properly. Use Huffyuv 2.1.1.
Scripts using Avisynth 2.6 32bit opened in Virtualdub2 b44282 32bit (on XP SP3 or Win7 64bit).
Convert to 23.976fps from 29.97fps. Crop beginning and end in Virtualdub2. Full process audio from stream 1 (tatt10.avi), Compress with PCM. Use Huffyuv 2.1.1.:
Use MkvMerge 9.6 (9.7-10.0 have data corruption issues) to create a .mkv file from 01-tatt10-cropBE-23.976-huffy-pcm.avi. Demux with MkvExtract to get .wav. Open .wav in Audition and fix any issues. Convert to 1648 before saving. Save as .flac or save as .wav and convert to .flac with Flac Frontend (or equivalent). By creating the final .flac this way (from the cropped huffy-pcm.avi), good AV sync is assured.
Crop without changing framesize (important per several guides, makes no sense to me). Deinterlace with QTGMC 3.368s. No audio:
QTGMC( Preset=“Slow”, EZKeepGrain=0.1, SourceMatch=2, Lossless=2, Sharpness=0.1, Sbb=0 )
Use Lagarith compression.
Add horizontal resolution (in theory) with Topaz Video Enhance AI 1.6.1. Gaia-CG model at 200%. Uncheck crop. Output .png images to a folder (run9, I actually did ~15 runs of VEAI before I found the best settings).
Post-VEAI in Virtualdub2:
Warpsharp filter at 53, 3. Sharpen at 8. Avisynth has equiv functions, but I couldn’t get them to work as well as VDub’s filters. In Vdub, no audio, compress video with x264-8bit ABR 5742Kbps Placebo, Film, L3.1, YUV 4:2:0, SAR W=1 H=1, FD and ZL unchecked. Color space Rec. 601, range limited. In the following script, I used QTGMC a second time to remove noise as the settings in the pre script don’t denoise. I used Bicubic as Lanczos over-sharpened. CropVD is a function I created to have VDub style cropping (posted on doom9).
ImageSource(“f:\w1\run9%06d.png”, fps=23.976, start=000000, end=138319, use_DevIL=true, pixel_type=“RGB24”)
QTGMC( Preset=“Fast”, InputType=1 )
Save video as MP4 Part14:
Create subtitles with SubtitleEdit v3.5.1b.
Use MkvMerge 9.6 to merge .mp4, .flac, .srt to .mkv. Make sure to set video to 24000/1001p. Rename appropriately.
Here are two 15 seconds clips of the new and the old captures: