Filmomat Smart Convert

[Tom Kershaw]

I've put a sample of colour negatives through the demo version and it does seem to work well, and interestingly the software will convert B&W without anything odd occurring. I'm intrigued as to whether the 'Filmomat Smart Convert' is taking a similar approach to ColorPerfect.

With Smart Convert, I have needed to dial back the contrast and increase density generally, but the auto white-balance seems to work well.

 

[Tom Kershaw]

My initial impressions having made a few practical direct comparisons between ColorPerfect and Filmomat Smart Convert is that the two pieces of software are not performing equivalent operations on the negative image.

 

[PhilBurton]

My initial impressions having made a few practical direct comparisons between ColorPerfect and Filmomat Smart Convert is that the two pieces of software are not performing equivalent operations on the negative image.

Sure. What are the results? Your impressions or conclusions?

 

[Romanko]

I intend to write a blog about this soon

Please do.

 

[Tom Kershaw]

Sure. What are the results? Your impressions or conclusions?

ColorPerfect seems to be more consistent or effective in terms of maintaining the colour and tonal qualities of the original scene, although I suspect many comparisons would be needed in order to get a more representative impression. Attempting to direct match the results between the software is difficult from initial impressions as the style of conversion is different.

One example (ColorPerfect):

CP Example

Filmomat SmartConvert:


Filmomat SmartConvert example

 

[Tom Kershaw]

ColorPerfect has trouble with this negative, but the SmartConverter software seems to understand the scene better. I suspect CP gets confused by the contrast range and density of the cloud highlights.

example

 

[Romanko]

I tried Smart Convert on 20 rolls of film that I shot recently (mostly Ektar, Gold200 and Portra 160 in medium format). Ignoring two or three cases of "catastrophic failure" the results are not bad. The "Flat" preset is probably the most useful as a starting point for processing; their default "Normal" settings give too much contrast and saturation; the "High contrast" is unusable. I would replace "High contrast" parameters with "Normal" and make default settings half-way between current Flat and Normal.
I tested a Windows version on my laptop with 1920x1080 display resolution. The conversion seems to be stable while the GUI is full of bugs. The layout changes from one image to another, the controls are overlapping so I have no means of changing contrast, density, saturation and other parameters. I did not like the interface. The controls are unusual for photo editing software, awkward to use and occupy too much space.
I did not have much luck with the auto-cropping function. The manual crop is useful though as it saves some disk space.
Unfortunately, on my laptop the software is unusable. If the authors fix the bugs I would probably buy it. I don't use Adobe products or other "rented" software, so I am really glad that applications like Smart Convert exist.

 

[benderscameracorner]

I've been trialling both Negative Lab Pro and Smart Convert, trying to figure out which one to actually buy.

Gotta admit that Smart Convert seems to come up with a more natural and usable inversion with a lot less stuffing around, especially when the negatives are lacking density. I've found NLP can get rather "crunchy" with some ugly artifacts, and you spend a lot of time faffing through all the options trying to deal with them. SmartConvert has fewer - though more logical/straightforward - controls, and just seems more 'one and done'.

I also like the fact it doesn't need to you crop out the film edges etc in order to achieve a proper conversion, as NLP does.

Being standalone is great and would have actually suited me a year ago, but the lack of integration with Lightroom is actually a negative now (for me anyway). Having to manually export TIFFs etc is a bit annoying. Would be nice if it worked both stand-alone and as a Lightroom/Photoshop plugin.

Just a shame they wanna charge me 19% VAT despite being as far away from the EU as possible without entering orbit....

 

[250swb]

ColorPerfect seems to be more consistent or effective in terms of maintaining the colour and tonal qualities of the original scene, although I suspect many comparisons would be needed in order to get a more representative impression. Attempting to direct match the results between the software is difficult from initial impressions as the style of conversion is different.

One example (ColorPerfect):

CP Example

Filmomat SmartConvert:


Filmomat SmartConvert example

It's a very interesting comparison. Both are in need of further colour correction but when that's done by using 'Auto Color' in Photoshop two different photos are revealed. The ColorPerfect version appears to show the scene as low sunset warm light especially with tints in the white of the hull, but the Filmomat version seems like a starkly neutral dull day? I wonder if the Filmomat inversion is overly neutralising whites? I prefer the ColorPerfect one myself but what was the light like when you made the exposure? I use ColorPerfect as my jumping off point with inversion and despite it's convoluted old school interface just pressing 'OK' seems to work most of the time.

 

[Tom Kershaw]

The ColorPerfect version appears to show the scene as low sunset warm light especially with tints in the white of the hull

Yes, the CP version is more accurate. The negative is a fairly recent one and I can remember the scene well.

 

[Romanko]

I tried Smart Convert on a high-resolution screen connected to the same laptop and it worked as intended. I hope they fix this bug in the next release.

 

[radialMelt]

I have also been trying out SmartConvert as an alternative to NLP and Vuescan inversions. First impressions are that SmartConvert gets me to a usable image much faster than NLP or Vuescan, with more neutral and consistent output as well. Having the controls laid out similar to a lab scanner, making it so that you can very quickly tweak things and export JPGs/TIFFs is great.

I've used NLP for years now but lately I am finding it quite buggy, and I really dislike having to go through an additional UI in LR to access its features - it's clunky, slow, and marries one to the software permanently if you ever want to go back and tweak conversions. I'm much happier doing a quick inversion externally and importing that into my catalogue and making further tweaks from there.

One thing I am noticing however - and I was obviously unable to test thing until actually purchasing the software - is that the JPG/TIFF output seems significantly less malleable than the JPGs and TIFs that I would get from NLP or Vuescan. Relatively small tweaks start to exhibit posterization in the files. I can push around the output from NLP and/or Vuescan much more before artifacts begin to show.

Has anyone else noticed this?

 

[Romanko]

Relatively small tweaks start to exhibit posterization in the files.

What is the resolution of the output TIFF file (8 or 16 bit)? Is the histogram of an image produced by SmartConvert smooth or does it look upsampled?

 

[radialMelt]

What is the resolution of the output TIFF file (8 or 16 bit)? Is the histogram of an image produced by SmartConvert smooth or does it look upsampled?

There's no options to specify this when outputting from SmartConvert. However, Bridge tells me its a 16bit file.

To be honest I don't know what the histogram of an upsampled image would look like. I attached a screenshot of the histogram of one of the TIFFs as per Photoshop. What do you think?

 

[_T_]

When you upsample from 8 bit to 16 bit you get a comb filter effect with a bunch of little valleys in the signal at regularly spaced intervals it looks like this:

edit: I just noticed you have quite a lot of clipping in the highlights of all three channels of your file

 

[radialMelt]

When you upsample from 8 bit to 16 bit you get a comb filter effect with a bunch of little valleys in the signal at regularly spaced intervals it looks like this:

edit: I just noticed you have quite a lot of clipping in the highlights of all three channels of your file

Thanks for the explanation! And yes, it was a quick and dirty conversion so not surprised in the clipping. Also the image itself is of a subject in a foggy but sunny environment so lots of bright stuff.

On that topic, it would be nice to see a histogram of some kind within SmartConvert for this reason.

EDIT: yeah, comparing more inversions side by side, the NLP JPGs seem to have far more latitude when it comes to say, pushing color luminance values (HSL) versus the JPGs and TIFFs out of SmartConvert... wish I knew why... all generated from the same Vuescan DNGs

 

[IMoL]

Thanks for the explanation! And yes, it was a quick and dirty conversion so not surprised in the clipping. Also the image itself is of a subject in a foggy but sunny environment so lots of bright stuff.

On that topic, it would be nice to see a histogram of some kind within SmartConvert for this reason.

EDIT: yeah, comparing more inversions side by side, the NLP JPGs seem to have far more latitude when it comes to say, pushing color luminance values (HSL) versus the JPGs and TIFFs out of SmartConvert... wish I knew why... all generated from the same Vuescan DNGs

In my testing of SmartConvert, I find I get the best results by exporting using the "Flat" profile (or a tweaked version thereof) and getting contrast etc where I want it in photoshop afterwards. I am still not convinced that I get results with SmartConvert that I can't achieve almost as easily doing manual inversions and colour corrections in photoshop. It's another tool in the toolbox though and I will continue to test with it.

One thing I really wish it had was the ability to drag and drop files to start the conversion process (or use "open in" in Bridge), having to open SmartConvert first and then navigate to my working directory and find the files is a pain.

 

[albireo]

@runswithsizzers I am interpreting this as a jab at NLP which bends RGB curves absolutely arbitrarily trying to generate a pleasing image. Perhaps their inversion is based on a density and gamma characteristics of an RA4 paper with automatic white balance.

I just downloaded the demo and I am super impressed with the results. In fact, I am tempted to say this is the best converter I've tried to date. And it's the only one which doesn't require Adobe.

I've played a little with it and while I get nice inversions out of the box, the tools seems to insist on clipping the highlights of many contrasty shots. This is also happening using the 'flat' profile.

Is there a way to tinker with the histogram tails in Filmomat?

 

[albireo]

I have been having some issues with black and white conversions. It sometimes blows out highlights (e.g. in cloudy skies) that were not blown out on either the negative or the original scan - you can reduce the contrast so it doesn't clip to pure white but the data does not come back. I have given this feedback to the developer.

Just noticed your comment. I found the same, though with C41 starting material. Quite disappointing. I'd be interested in knowing if the developer shares their thoughts with you on this.

 

[Filmomat]

Hi there, this is Lukas from Filmomat, I’m the developer of SmartConvert. I’m glad to see that the software is already being used by a lot of people! I’m always open for feedback and improvements. There shouldn’t be a lot of clipping in the highlight / shadows happening. If you encounter issues like this, the best way is to send me an email at info@filmomat.eu, ideally with example files which show the issue. This way I can take a more in depth look and see what is causing the issues

 

[IMoL]

Just noticed your comment. I found the same, though with C41 starting material. Quite disappointing. I'd be interested in knowing if the developer shares their thoughts with you on this.

I did feed that back to the developer and the situation was much improved in version 1.35

 

[radialMelt]

Hi there, this is Lukas from Filmomat, I’m the developer of SmartConvert. I’m glad to see that the software is already being used by a lot of people! I’m always open for feedback and improvements. There shouldn’t be a lot of clipping in the highlight / shadows happening. If you encounter issues like this, the best way is to send me an email at info@filmomat.eu, ideally with example files which show the issue. This way I can take a more in depth look and see what is causing the issues

Thanks for chiming in, Lukas! Can you comment on why the TIFFs and JPGs from SmartConvert seem to have far less headroom for tweaks vs JPGs from NLP? Both produced by the same DNGs out of Vuescan

 

[Romanko]

Hi there, this is Lukas from Filmomat, I’m the developer of SmartConvert.

Hi Lukas. Great to see you on this forum. Could you please have a look at your GUI on low-resolution displays? See my post #32. I can send you screenshots if you are interested.

 

[radialMelt]

I am still not convinced that I get results with SmartConvert that I can't achieve almost as easily doing manual inversions and colour corrections in photoshop. It's another tool in the toolbox though and I will continue to test with it.

I agree, the results are not really far off from any other conversions I've done with other software, manual or otherwise. However I will say that SmartConvert gets me there much faster, with a workflow that is more to my liking!

 

[ant!]

Interesting! I use since a few years Darktable + negadoctor (all open source, on Linux), but I guess I should boot to Windows to try this and a few other standalone solutions mentioned in this thread to see how they compare. While negadoctor/Darktable needs a bit to get used to the controls, I find I get mostly nice results with some tweaking. If I can get a faster equally good or better result with this, I might consider switching this process to Windows.... I had tried NLP some years ago when I had an Adobe subscription through work, but I found the results and workflow not worth the cost for me (NLP + Adobe)...