Improving badly shot interview footage?

Any suggestions as to how to improve this interview footage?

The color can be adjusted, but it's out of focus which essentially cannot be fixed in post. You can try to hide a little focus problem with sharpening, but if you could truly fix out-of-focus material in post, NASA would not have sent astronauts to the Hubble Space Telescope to add corrective optics.

The amount of color & exposure latitude will vary based on the codec. If it was shot with a 10-bit 4:2:2 low compression codec, you can do more. If it is a transcoded or captured 8-bit H264 file, you are very limited on what is possible.

But it looks almost faded.  If was originally film, that can sometimes be re-scanned by a professional company, provided the film is still available. E.g. this place can do "wet gate" scanning and they are experts at restoration:  www.posthouse.com

The problem is it's nearly monochromatic. I think it's beyond regular correction & grading. See attached attempt just using FCP color wheels and masks. It really needs colorization via some of the new AI tools. I have never used those. Some of the potential can be seen in Peter Jackson's WWI documentary: 

However some new colorization tools are now available for regular users. Here is an example of one person who used an old computer (not even a Mac) to restore some 1945 B&W film. I don't know what software he used: : 

Joema said it all.
You can you use B-roll to insert images  when possible.

Those were some awfully grim examples there but maybe that was before they finally discovered bright & vivid primary colors.

Note everything has the same drab tones and hues and brilliance. Everyones skin is the same narrow range. That limited range of colors is an embarrassing disaster to anyone who deals in film footage.

This damaged film I recovered from 1939 has better and more believable colors than that colored garbage from 1945.

Skin tones and garment colors are all universally different by mere simple nuances but what's shown in those examples is one shade of gray, green, or brown, or blue for EVERYTHING. 1941 wasn't so monotone as those colorists would have you believe. (All these images are direct from 8mm frames. No post processing or grading was applied.)

Some film remains brilliant while others experience dye shift towards blue or magenta due to aging or the quality of the original processing facility. (Bride / Sun Hat) The problems are correctable providing the transfer collects sufficient data to tolerate the pushing and pulling (I run 1080 ProRes 442 minimum and will at times go 4K ProRes HQ if warranted) and FCP can handle the simple corrections but for those more involved situations I rely on Davinci.

"If you can't improve on the silence of the desert, then remain silent." (Bedouin expression.)

"If you can't improve the color of three generations of women, then leave them alone" (VTC expression.)
 - - - - -
The more observant reader will notice the overscan area on the 8mm bride frame. Some cameras captured a greater field of view which is eventually cropped out by the standard 4x3 projector gate. My transfer station captures the entire image which is fairly close to 16x9 displays of today.

Samples for clients who had their stuff run by other companies and didn't believe there was anything that could be done that would improve their footage. (This helps get a refund from the people that did the lousy job.)

Thanks for the replies everyone. The interview was shot on MiniDV back in 2007, but all we now have to work from is a rip from DVD. Those color restoration images above look impressive - how would I gain access to any of those color tools?

The interview makes up part of a featurette which cuts together several other nicely shot interviews.

I have spent most of today trying to run the video clip through DeOldify , but I keep getting errors - can anyone help me to understand what is happening? Has anybody else had any success with this tool?

colorizer = get_video_colorizer()
then I get:
EOFError: Ran out of input

I also had some (limited) success with neural.love ai.

I think this is about as good as I'll get it. I converted it to black and white, exported as frames, then colorized via Photoshop neural filters/actions.

I can live with wonky colors it's the brick courses and floor levels being askew that grates my nerves. Correct that. (I did.)

I've added a vignette to it, any other suggestions which would help?

BeatGirl wrote:

...any other suggestions which would help?

Don't neglect the audio, which can have a perceptual impact equal or greater than video. Oftentimes lower quality audio can be improved using tools like iZotope RX8. I'm not an audio expert but if you can upload an approx 5-10 sec audio clip I could try a few things. Ideally the clip should include a pause or delay when he's not talking, which facilitates sampling "room tone" to perform spectral noise reduction.