Encoding with Lossless Formats

[Mr. Fanservice]

Happy New Year, everyone. Now that out of the way...

If you encode videos (e.g. burning in subtitles, not cutting at keyframes, frame interpolation, deinterlacing) with lossless formats, does that mean the result will also be lossless?

[poisondeathray]

If you encode videos (e.g. burning in subtitles, not cutting at keyframes, frame interpolation, deinterlacing) with lossless formats, does that mean the result will also be lossless?

It would be lossless compared to the filtered output - ie. there would be no compression losses .

But it would not be considered lossless compared to the original because of the hardcoded subs. Also, the deinterlacing - would need to retain the original fields (some deinterlacers have an option to retain original fields but the temporal visual quality is usually worse), and the interpolation would need to retain every original frame among the interpolated ones. An encode is considered lossless compared to the original, only if you are able to get the original uncompressed data back

Cutting on keyframes is irrelevant when you re-encode directly through an avs or vpy script, because the video is decompressed to uncompressed (every frame is an I frame) before recompression

[Mr. Fanservice]

If you encode videos (e.g. burning in subtitles, not cutting at keyframes, frame interpolation, deinterlacing) with lossless formats, does that mean the result will also be lossless?

It would be lossless compared to the filtered output - ie. there would be no compression losses .

But it would not be considered lossless compared to the original because of the hardcoded subs. Also, the deinterlacing - would need to retain the original fields (some deinterlacers have an option to retain original fields but the temporal visual quality is usually worse), and the interpolation would need to retain every original frame among the interpolated ones. An encode is considered lossless compared to the original, only if you are able to get the original uncompressed data back

Cutting on keyframes is irrelevant when you re-encode directly through an avs or vpy script, because the video is decompressed to uncompressed (every frame is an I frame) before recompression

So if I deinterlace a video and choose a lossless output, the resulting video would still be interlaced?

[poisondeathray]

So if I deinterlace a video and choose a lossless output, the resulting video would still be interlaced?

No , it would be deinterlaced, but the original fields would be kept in double rate (bob) mode

This means you could discard the interpolated fields, and get back the original fields, and thus the original uncompressed interlaced video if you used lossless encoding (and didn't do other stuff like hardcoding)

Lossless deinterlacing mode with something like QTGMC usually looks much worse - more flicker and temporal inconsistencies. Yadif is considered a lossless deinterlacer in double rate mode, because it retains the original fields - but the yadif deinterlacing quality is low with many artifacts and flicker

[Mr. Fanservice]

So if I deinterlace a video and choose a lossless output, the resulting video would still be interlaced?

No , it would be deinterlaced, but the original fields would be kept in double rate (bob) mode

This means you could discard the interpolated fields, and get back the original fields, and thus the original uncompressed interlaced video if you used lossless encoding (and didn't do other stuff like hardcoding)

Lossless deinterlacing mode with something like QTGMC usually looks much worse - more flicker and temporal inconsistencies. Yadif is considered a lossless deinterlacer in double rate mode, because it retains the original fields - but the yadif deinterlacing quality is low with many artifacts and flicker

So which is better for deinterlacing normally and losslessly? Will doubling the frame rate help?

[poisondeathray]

So which is better for deinterlacing normally and losslessly? Will doubling the frame rate help?

In general, QTGMC's non lossless mode "looks" better and is smoother . But it depends on what scenario this is for, what criteria you are using

Normally you would double rate deinterlace (bob) always, otherwise you lose 1/2 the temporal samples, and motion would be less smooth

Test them out and see what works better for your scenario

[Mr. Fanservice]

So which is better for deinterlacing normally and losslessly? Will doubling the frame rate help?

In general, QTGMC's non lossless mode "looks" better and is smoother . But it depends on what scenario this is for, what criteria you are using

Normally you would double rate deinterlace (bob) always, otherwise you lose 1/2 the temporal samples, and motion would be less smooth

Test them out and see what works better for your scenario

I just want to deinterlace videos in the best and highest quality possible (or maintain the same quality if that's better). I also want to deinterlace multiple videos at once.

[poisondeathray]

I just want to deinterlace videos in the best and highest quality possible (or maintain the same quality if that's better). I also want to deinterlace multiple videos at once.

People have different perceptions and measurements for "quality" .

If you want the same "quality" as the original, don't do anything. Use the original , or a copy of the original. Filesize will be much smaller too if it used lossy compression. You should always keep the originals anyways - maybe some better deinterlacer comes out in the future. Some machine learning deinterlacers

You would have to do some tests and adjust the settings for each specific video

If you use the same settings for every video, you definitely won't get the "best quality" in every case - no matter how you're defining "quality" . You wouldn't necessarily use the same settings for different genres and sources .

[Mr. Fanservice]

I just want to deinterlace videos in the best and highest quality possible (or maintain the same quality if that's better). I also want to deinterlace multiple videos at once.

People have different perceptions and measurements for "quality" .

If you want the same "quality" as the original, don't do anything. Use the original , or a copy of the original. Filesize will be much smaller too if it used lossy compression. You should always keep the originals anyways - maybe some better deinterlacer comes out in the future. Some machine learning deinterlacers

You would have to do some tests and adjust the settings for each specific video

If you use the same settings for every video, you definitely won't get the "best quality" in every case - no matter how you're defining "quality" . You wouldn't necessarily use the same settings for different genres and sources .

But is there a batch deinterlacing tool though?

[poisondeathray]

But is there a batch deinterlacing tool though?

Not sure if there is a GUI, but you can create batch scripts, batch encodes with CLI tools - there are many forum posts on that topic, you can use search


Different sources require different settings and filters . "Default" anything must be customized to your specific source, otherwise you won't get ideal results

eg. QTGMC default settings might be appropriate for a SD , soft, lower quality, slightly noisy source.

But those same settings would oversharpen and excessive remove excessive details for a "clean" HD source

[lollo]

In general, QTGMC's non lossless mode "looks" better and is smoother . But it depends on what scenario this is for, what criteria you are using

I agree. QTGMC lossless mode is useful only when there is a need to create a "temporary" deinterlaced video to apply, for instance,a spatial-temporal denoiser and then discharge the interpolate field and then interlace back.

You would have to do some tests and adjust the settings for each specific video

So true!

If you want the same "quality" as the original, don't do anything.

Yes. If I can add something, the original need to be deinterlaced to be displayed nowdays, so it cannot be watched in its native form anyhow. Then let's do the deinterlacing the best possible.

[Mr. Fanservice]

But is there a batch deinterlacing tool though?

Not sure if there is a GUI, but you can create batch scripts, batch encodes with CLI tools - there are many forum posts on that topic, you can use search


Different sources require different settings and filters . "Default" anything must be customized to your specific source, otherwise you won't get ideal results

eg. QTGMC default settings might be appropriate for a SD , soft, lower quality, slightly noisy source.

But those same settings would oversharpen and excessive remove excessive details for a "clean" HD source

I have 1080i videos. And what should the search term be?

[poisondeathray]

"avisynth batch script"

"ffmpeg batch encode" using for loops