Tag Archives: Frame rate

The Hands of Orlac (Eureka)

Eureka’s BD of The Hands of Orlac (Orlacs Hände, 1924) has one of the most aggressive and inept restorations I’ve seen. Nearly every shot has restoration artifacts.

Be sure to click on the GIFs to see them full size.

Interpolation

The first and last frames of most shots are interpolated. Sometimes it looks like a duplicate pulldown frame. Other times, it introduces some serious distortions. Nearly every example in this section has frames that are completely computer-generated. Faked.

Conrad Veidt’s head stretches, Alexandra Sorina’s fingers split, grain and background warp, plus frozen temporal cloning at the top edge:

The pen is broken up and the rest of the frame is nearly frozen:

Stuck grain and warping throughout:

Hideous computer-generated teeth:

Not a freeze-frame, but interpolation:

Interpolation makes scratches stick to the image:

Here, interpolation is combined with spot removal. At first glance, it’s impressive. Look closer, and you can see that it leaves smudges in place of damage AND it leaves remnants behind:

Notice how the sheets of paper distort:

Here it’s combined with frame blending and frozen temporal cloning:

Interpolation utterly fails when it’s called upon to generate background, such as behind Veidt’s head. And check out his smashed head and hand:

Veidt’s hand becomes putty and merges with the sleeve. The sofa pulsates:

You can see magnetic scratches on the wall behind the now-breathing sofa:

I admit that it’s not always immediately apparent recognizing restoration artifacts on a small screen, but look how obvious it is here when displayed at a fraction of its full size! The man gets smashed. And that background…

Interpolation is combined with temporal cloning on the ceiling lamp and chairs on the left. They now bob up and down:

His body compresses and arm stretches, creating a second wrist. Yikes!

Veidt’s head smashes into an invisible curved ceiling:

Our villain’s fingertips disappear. The disembodied hand on the left is completely frozen:

Interpolation destroys the head:

An example of interpolation that is nearly a freeze-frame, plus some negligible spot removal:

The foot disappears:

Continue reading

Mother / The End of St. Petersburg (Flicker Alley)

Flicker Alley’s presentation of Mother (1926) is a bit of a throwback to when silents were released on home video with whatever print the distributor happened to own. Nowadays, with so many movies getting processed to death, it’s jarring to see a major silent presented as-is from a well-worn print. Still, the transfer is nice. Natural grain, no major compression issues, and most importantly, no restoration artifacts.

Clipped directly from the BD.

As you can see, it’s rough, but it’s nicely encoded. I suppose it could be processed with new grading and stabilization, but it’s so easy to be heavy-handed, that I think I’d prefer this. Besides, it’s clearly not first-generation material, so what’s the point?

The End of St. Petersburg (1927), also well-worn, derives from Mosfilm’s 1969 restoration. It runs at a variable frame rate, primarily 24fps (most of the time) and 18fps, step-printed in 1969. I wonder about the accuracy of this. 18fps makes sense for slower scenes in the first half, but the dips in speed are drastic and inconsistently applied as the action and editing get faster in the climax.

FA’s transfer has further stretching that results in random single frames being duplicated, mostly in the second half. Many of these are easy to miss amid the frantic action. However, it’s another nice encode.

Clipped directly from the BD.

When Mosfilm step-printed in 1969, they introduced a lot of debris. You can see this when frames get repeated; you end up with the same image getting completely different layers of dirt and different exposures. They also introduced gate weaving and occasional out-of-focus frames (at splices). By undoing their step-printing, we can get non-destructive repair:

Left: 1969 step-printing. Right: Step-printing removed and corrected to 60fps.

The difference is slight, but the right side is a little cleaner and gate weave is more natural and less…slippery. It may not be noticeable in this short clip, but over the course of the movie, and on a big screen, the 1969 step-printing defects are visible.

I am very curious what original material actually exists and its condition. Alas, Ukraine. Oh, well…

Mother is transferred at 20fps and encoded at 24fps. The End of St. Petersburg is transferred with a variable frame rate, 8-24fps, and encoded at 24fps.

Ashkan, the Charmed Ring and Other Stories (Deaf Crocodile)

[Ashkan] was shot on video at 25 frames per second. [This version was converted to 24fps] by removing 1 frame per second from the movie.

I initially misread that as “one frame per shot,” which would have been no big deal. So, to see such choppiness really caught me off guard:

Clipped directly from the BD.

It’s really noticeable and distracting throughout the movie, though the hand-held shaky cam helped hide some of it.

If this is the best version, I have some questions. Why couldn’t the 25fps original be used? Does it still exist? The video that remains is basically slowed down; slowing the original down to 24fps would be effectively the same. Why couldn’t it be a 25fps->1080i60 conversion? What about the sound? Is it slowed or pitch-corrected?

I’m not thrilled with the video compression, but it could be worse:

Clipped directly from the BD.

A poor presentation. Whatever the price is for this four-disc set, it’s 25% too much.

Blood and Sand (Kino)

Kino’s BD of Blood and Sand (1922) has relatively little computer repair work done on it, but what there is, is sloppy.

The most common restoration artifact here is what I call “quick-reverse” (QR) frames. It is when, instead of the film advancing as a normal ABCDEFGH, frames repeat like ABCDCDEFGH. It gives a feeling of the movie seizing up and is ugly to see in motion. Here, it’s limited to some titles.

Above, you can see Kino’s original with QR frames compared with those frames removed (and encoded at 60fps). Kino also added an additional repeating four frames to pad the length, but in this case they are far enough away from their initial appearance than it doesn’t register as looping frames. If you must pad out the length by repeating frames, spacing them out to break our persistence of vision works ok. In my corrected example, I removed these, too.

However, video is capable of arbitrary frame rates. Just stretch the playback speed in the timeline.

Some ugly interpolation:

Above brightened to see detail

The remaining errors are mostly frozen temporal cloning. What follows are some of the worst.

Top third of frame is frozen.
Top of frame is frozen and has an obvious seam.
Retouched area is a frozen splotch that doesn’t match.
Detail of above splotch.
Frozen splotch at top right.
Detail of above splotch.

Notice how the frozen temporal cloning above is misaligned and has a blurry edge. The technician needs to align it and use a harder brush and choose an indistinct source frame far enough away that it doesn’t result in either frozen or quick-reverse temporal cloning.

Top third of frame is frozen and misaligned.

Two thirds of the above frame are frozen, including the boy, which is the one part of the frame that shouldn’t be frozen. This spot also coincides with the 24fps pulldown. The same frame is shown for a duration of three frames, which means this spot runs for the equivalent of six frames per second!

Deleting that mostly-frozen frame results in a tiny jump cut, but the technician created that jump when cloning the motion out of existence.

And an editing error at 1:02:51:

The source print has lots of damage, though it still looks good. The damage is not particularly bad, but it’s there and occasionally significant. I find it hard to believe that the damage covered by these sloppy repairs was worse than and intolerable in relation to the damage that remains.

Blood and Sand is transferred at 19fps and encoded at 24fps.

PS: I would love a better word to describe quick-reverse frames.