Once again, a panel of experts advised me on this article (see below). Instead of relying on grading influencers, who are dependent on publishing new content at regular intervals, I have once again
content at regular intervals, I am once again relying on the experience of four senior colourists with decades of experience. Added to this are several weeks of intensive research, expensive further training (Resolve Colourist Advanced) and paid online platforms (Lowepost). In the end, I take on the role of a curator with the aim of collecting sensible solutions for frequently occurring problems.

In commercials, skintone treatments and spot adjustments are the order of the day. However, I also learnt from the experts that beyond this, for example with the rejuvenation of people by a decade, the limits of what is technically feasible are always being scratched. Of course, colourists always hope for the stroke of luck that complex shots are handed over to retouching experts.

The professionals also report that the time requirements in other areas, such as feature films, documentaries and image films, are now so high that professional colourists can often do little or nothing in the beauty sector. Of course, they are also less likely to work on complex cases: A production whose budget can afford external grading often also has at least one professional make-up artist on set. As a result, any abnormalities that are difficult to correct are moved to post. The extent to which the fact that such retouching is almost exclusively carried out on female actresses is socially questionable is something we unfortunately have to leave out of this article.

During my research, I learnt that some actresses, for example, have certain retouching steps contractually guaranteed. In addition to the rest of the grading of the feature film, between half a day and a whole day is then estimated for these edits. In order to ensure that this person does not stand out as having been retouched, their counterpart in the scenes in question often has to be retouched as well.

Even today, there are still clear differences in how much effort is put into male and female actors. For male actors, skin retouching, such as red discolouration or spots, is often retouched. For female actors, however, the adjustments are often more extensive. In this case, the adjustments go far beyond retouching the face, such as skin retouching or changing the colour of lipstick. For example, the hands are also worked on by adjusting the skin tone (often too reddish because there is no make-up) or rejuvenated using various techniques. Sometimes even the entire body contour is shaped with a grid (the Warper effect in Resolve). In commercials, the entire image is sometimes simply scaled down by 4 per cent to make the actress appear slimmer.

Nevertheless, editing steps in the beauty area can also be worthwhile for smaller productions, for example in the image film sector. They can be used to remove temporary blemishes from the protagonist, increase the production value or simply draw attention to other aspects by reducing details in a part of the image.

Face refinement

The often-discussed internal Face Refinement plugin is actually (as I have learnt from our experts) also used time and again in a professional environment. Both for skin retouching and for dark circles under the eyes or forehead wrinkles.
However, if the tracking or placement of the outline does not go well, the beauty plugin included in the delivery can also be used. For example, if the camera moves too much or the head is turned too far, the face refinement plugin gets confused again and again. The short shot length naturally benefits graders in the advertising sector, as the automatic analysis function of the plugin is less likely to fail. If it is wrong, very little can be adjusted manually. Whether the outline can be rotoscoped at some point is still up in the air. As a result, this tool is extremely polarising: it is often an enrichment, but in some cases it is also completely useless. In the following, we therefore use tools that can be used as alternatives and that allow a high degree of manual customisation.

Skin Tone

As Alexis van Hurkman impressively explains in his “Colour Correction Handbook” (Peachpit Press – is.gd/cc_handbook), the skin tone is a so-called memory colour. In other words, a certain hue range that is determined by the blood circulation and the oxygen content in the blood and is therefore very similar in all healthy people within a certain spectrum (see image).
The experts agreed that a healthy skin tone is an extremely important element that should always be kept in mind. But exactly how this is defined is a matter of taste. The preferences of the director and camera department are therefore clarified in preliminary discussions. There are also often personal preferences when it comes to saturation, i.e. how pastel the skin tones should be. The Skin Tone Indicator therefore definitely makes sense as an aid and general reference.
However, whether the skin tone is processed separately from the rest of the grading depends more on the type of production. In commercials, it is almost always treated separately with qualifiers and masks – in feature films it is never actually treated separately. There, you always try to grade largely with primaries, looking at the whole picture and the skin tone, and only use secondaries if it is either not possible otherwise or if two actors have very different skin tones. Professionals see the advantages of primaries in terms of speed and, above all, consistency within the image.

Skin retouching

Midtone Detail is usually used for general skin retouching. This tool increases the contrast at edges, but only in the medium and light brightness areas. This at least does not increase the noise in dark areas and is therefore also suitable as a general sharpening method. In combination with a coarse mask, for example around the head, and a qualifier on the skin, you can often achieve good results with Midtone Detail – the professionals agree. With the qualifier, however, care should be taken (for example with the saturation parameter) not to include the eyes or lips in the key. If you also want to give the skin more shine or a porcelain-like appearance, you can also use the key you have just created directly for this and simply raise the centres (gamma) with the wheels. This makes the skin a little softer in terms of contrast and at the same time usually reduces the saturation with the increase, making everything a little more pastel and porcelain-like.
If the focus is less on the general appearance of the skin and more on specific areas of skin, other tools are used. Sometimes a tracked mask with a blur on the area in question can be enough.

If this is not enough, a reduction in contrast (using Pivot), a slight blur and colouring with the wheels (most likely Offset) often helps to beautify the masked area and adapt it to the surrounding skin tone. The next stage, which is also used by our professionals, is node sizing. To do this, we first mask an area that is “flawless” and matches the skin structure and illumination of the area we want to enhance as closely as possible. In the Sizing area, we select the small Node Sizing tab and use Pan and Tilt to move this beautiful area to the unsightly area.

Now all we have to do is track the mask and adjust it to the size and, with a bit of luck, we’re done. Another option is called Patch Replacer and is also a supplied effect. It works like the clone tool in Photoshop by defining a source and a target. The many setting options would go beyond the scope of this article, but here is another important tip for tracking: switch from window tracking to FX tracking. Only then can a manually set tracking point be tracked. Then simply drag the target circle of the plug-in to the correct position and it should move correctly. Patch Replacer is only slightly different from Node Sizing, but for some it offers a more pleasant UI and a few more options for the settings. If none of this helps, there is probably no way around a compositing programme.

Circles under the eyes

Personally, I have always achieved very good results with dark circles using two very similar techniques. Both first require a tracked, soft mask around the dark circles under the eyes. Then it is basically a matter of increasing the brightness of the dark skin area, reducing the contrast in order to visually flatten the skin and adapt the skin tone to the surrounding skin.
In the first method, we reduce the contrast considerably and use Pivot to adjust the brightness to the surroundings. We then mainly use Offset (and Saturation if necessary) to give the area the right colour.
The second method is similar, but usually results in slightly more realistic skin tones with less reduction in the darkening of the eyes. First lighten the shadows with Lift until a light-coloured area is created. Now the 32 bit floating point calculation of Resolve comes into play: With Gamma we now lower the centres until a suitable brightness with a view of the surroundings is created again. If you now use the Lift and Gamma wheels to give the area the right colour, you often get very nice results.
Of course, both methods require good tracking and masking. If you have to render with handles or if shots are added, adjustments in this area are of course difficult, but you should keep this in mind with all the techniques in this article. Speaking of..

Eyes

First of all, it must be mentioned that adjustments to the eyes themselves are very difficult to transfer consistently to many settings. Either the tracking runs away from you or you have to adapt your editing to the lighting conditions, which is sometimes not always easy over many shots. As this is “only” a question of (rotoscoping) effort, we still don’t want you to miss out on the tricks.

If, for example, digital eyeshadow is to be applied, this is very straightforward with short shots. Simply apply a mask around the eyelid area, set softening, track and now adjust both the colour and the brightness with the gamma wheel. This creates both smokey eyes and different colours in a very natural way. When it comes to the iris and the white eye skin, the opinions of the experts differ, as an eye skin that is too light and desaturated looks extremely unnatural. One suggestion from the professionals was to increase the sharpness with a negative blur, with which I have often achieved beautiful results myself. This makes the eyesight a little more prominent, which makes the eye appear more reflective and therefore shinier, and the additional sharpness of the iris also draws the eye more to this part of the image. With the same intention, you can also enlarge the mask and “put on” the subject a mask that resembles diving goggles in shape and size. To lighten the whites of the eyes and the iris, Contrast together with Pivot is much more suitable than working with the Wheels, as this creates unnatural contrasts more quickly
unnatural contrasts are created more quickly. But as with everything in this area, less is often more. Enhancing the colour of the iris, on the other hand, is difficult as the surrounding eye skin and the eyelid are often difficult to keep out and are therefore also enhanced in saturation, which creates unwanted colour clashes.

Hair and beard

According to our experts, adjustments to long hair (for example in shampoo adverts) are often passed on to dedicated specialists or tackled during and before the shoot with lighting, styling and choice of model. However, if only minor retouching is required, this is already done by the grading department.

If, for example, the shine on the hair needs to be emphasised more, our experts mainly help themselves by increasing the contrast in these areas and sometimes adding a little sharpening. If there is discolouration in the hair, simple but very precise qualifiers are combined with tracked masks to precisely address the area in question.
Beards or light-coloured hair are rarely processed, according to our experts. In most cases, however, this is done with tracked masks and a little darkening. In my projects, I have already achieved really beautiful results when darkening areas of the beard or the entire beard, as they appear much thicker afterwards. All in all, a lot can be achieved with simple means in grading without having to go straight to the big compositing stage.

The experts

Sebastian Göhs
Sebastian Göhs is a freelance senior colourist from Berlin and works mainly on feature films, TV series and commercials. Most recently he worked on the Amazon show “Der Greif” and İlker Çatak’s feature film “Das Lehrerzimmer”. Sebastian began training as a film and video editor at Geyer Berlin in 2002. After graduating, he moved to Digilab and advanced to Senior Colourist. This was followed by the development and management
of the image department at Rotor Film and finally became self-employed. www.sebastian-goehs.com

Stefan Andermann
After training as a film and video editor at ARRI Film and TV Services, which lasted three years, Stefan Andermann worked in the dailies colour grading department for around two years, but was quickly given the opportunity to take on fine corrections. In the following 12 years, he worked in commercial grading at ARRI. He has specialised in feature films for around seven years and works at Pharos on cinema productions, for streaming services and TV productions. www.pharos.de

Andreas Brückl
Andreas Brückl started as a colourist at Bavariafilm in 2005 and worked as a freelancer from 2010. After four years in the advertising industry in Istanbul and a stopover in Malaysia, he moved to India in 2017. After four years of chaos and colour saturation in Bollywood blockbusters, he moved to Dubai for two years. He has been back in India with DNEG as a supervising colourist since 2023. www.baselight-colorist.com www.vimeo.com/colorgrading

Florian Wolf
In 2003, Florian Wolf began his career as a career changer in post-production and telecine at Pictorion das werk. He worked there as a junior colourist until 2009. From 2009 to 2011, he moved to ARRI Commercial Munich and was promoted to Senior Colourist. After a brief intermezzo at das werk Munich, he took the plunge into self-employment and has been working as a freelance senior colourist since 2011. In addition to colour grading, he offers the production of commercials, image films and music videos from concept to shooting including the entire post-production. www.flowolf.info

But now the “Scary Fast” event for the release of Apple’s M3 computers was filmed entirely on the iPhone, as the credits alone reveal. On YouTube you can find the “Behind the
Scenes…”, where it also becomes clear that the filming and technical effort was actually the same as always(is.gd/iphone_bts). But the dollies, cranes and the drone were actually just sitting on a new iPhone 15 Pro Max. In addition, the predominant visual theme was “black”, matching one of the new laptop colours.

The EU is to blame

This doesn’t mean that this iPhone is a low-light wonder (it isn’t), but that you can set the light to high contrast and then grade it to black. The decisive factor is that recording in ProRes with Log is now available. Apple would certainly have liked to hold on to an in-house connection for longer. But the EU has demanded that all smartphones that come onto the market here in 2024 must have USB-C. Apple has taken the plunge and has already installed the corresponding socket in the new series this year.

Data transfer

In the early productions with iPhones, the employees must have had incredible patience. Recordings had to be transferred via WiFi or the interface with the euphemistic name “Lightning” (or its 30-pin predecessor), which is very slow by today’s standards. Two short test clips of around 750 MB took over ten minutes via “AirDrop”, via USB-C the process was
the process was barely measurable in seconds.

Initially, there were still rumours floating around the web that, apart from charging the battery via USB-C, Apple might restrict the use of external devices to its own offerings or only allow its own cables at pharmacy, sorry, Apple prices.
None of this has materialised: A standard SSD connected via a normal data cable (not just a charging cable, of course) works without any restrictions. With the iPad, we still had to complain that FCP-X cannot work directly from connected storage devices (DP 23:05). On the iPhone, on the other hand, Apple does not force internal recording, as already available apps show. The Apple event was obviously recorded with the help of the free Camera App from Blackmagic (BM for short), Apple even thanks us in the credits.

Storage media

It is clear that Apple also wants to satisfy its shareholders: the new options are limited to the two top models. With the “Pro”, you also have to pay an extra 130 euros for 256 GB of internal memory to get full functionality with the memory-intensive ProRes even without an SSD. This is the basic configuration for the Pro Max. The internal memory is really fast, with 1.6 gigabytes per second (GB/s) when writing and reading. Measured with Diskio by Sergii Khliustin, a free app with witty comments – is.gd/diskio.

Externally, only regular USB-C is supported, no Thunderbolt. However, with a good SSD, this is completely sufficient to record ProRes 422 HQ in UHD with up to 60 fps directly, for which you only need around 200-225 MB/s. Initially, the app even listed ProRes 4444, but that was probably too much for the system. Even on an Arri Alexa, this codec is not always selected, and 422 HQ is completely sufficient as a source for the iPhone. In UHD at 60 fps, it needs just under 800 GB of space for one hour.

A Samsung T5 SSD, for example, as external storage shows around 373 MB/s throughput when writing, depending on the app being measured. Even a T7, which is not recommended for BM cameras, ran for two hours on the iPhone at 25 fps. The free OWC Drive Speed app can actually directly measure the suitability for various codecs, resolutions and the achievable frame rate. However, it already warns of picture dropouts,
without us being able to detect these in practice for the respective media, so this app is probably not yet fully developed.

Even a T5 is still relatively large for an iPhone, although thanks to MagSafe it can be easily attached to the back with a magnetic ring. Some people wonder whether a smaller flash drive with USB-C would also work. Samsung has relatively fast models for this, currently with a maximum of 256 GB. According to the company (we didn’t have one to test), these should be able to read 400 MB/s, but only write 110. This information is always given with the small addition “up to” and only refers to the 256 GB model. This should still be sufficient for ProRes 422 HQ with 25 fps, but ProRes 422 LT could also go up to 60 fps. Unfortunately, such sticks are often not compatible with the mobile phone case.

We have also not tested the idea of connecting a really fast SDXC card with a small USB-C adapter. But a Sony “Tough” or comparable cards with 300 MB/s should work with a really fast adapter. A card reader can of course also be useful for backups on the iPhone when working with conventional cameras. If you want to edit the material on the PC, you should pay attention to the formatting of the media: The iPhone does not recognise NTFS, but exFAT works up to 2 TB.

Be careful, there is a risk of data loss! The “classic” HFS or the new APFS, which the iPhone naturally supports, are better. One of these formats would therefore also be advisable for transferring to the PC. Under Windows there are programmes such as “HFS for Windows” or “APFS for Windows” from Paragon for using these formats. Of course, the iPhone also uses these two formats.

Power and heat

A fully charged Pro Max still showed 66 per cent charge remaining after one hour at 60 fps, and 42 per cent after two hours of recording in ProRes 422 HQ at 25 fps in UHD, with the T5 SSD also being supplied with power. The SSD absorbed the greater proportion at 25 fps, at 60 fps it is the other way round. The iPhone 15 Pro still had 58 per cent after one hour at 60 fps, the camera app needed the larger share, the SSD about 2/3 of it. This was the app from Blackmagic. This value is not the same for all apps, as we will see in their tests. These are amazing runtimes, but the iPhone does get very warm, despite iOS 17.1.1, which has essentially eliminated the initial heat problems. You should definitely remove any silicone case, we got overload messages at 60 fps on the Pro Max after a few minutes at an ambient temperature of 27 degrees. Another tip from BM was to switch off the overlays when recording, i.e. to “wipe them away”. This allowed the Pro Max to run smoothly even in slightly warmer conditions.

The iPhone 15 Pro may seem more attractive for video with its 3x lens and slightly lower price. However, its battery is smaller; in stand-alone mode with SSD, 20 per cent charge was still left after two hours. Unfortunately, the smaller housing also cools less well, and despite the above cooling measures, 60 fps was over after a few minutes. At 25 or 30 fps, the Pro also lasted without overheating and the app consumed less power in relation to the SSD. Very fast SSDs, such as a Samsung 980 Pro in an NVMe enclosure, require too much power and only run on both iPhones with an external power supply via USB-C hub. You don’t necessarily have to take a hub with you to externally power the iPhone itself, as there are already power banks for MagSafe.

Frame rate and synchronisation

With a runtime limited only by power and storage medium, another question arises: What about maintaining the frame rate and thus the picture-sound synchronisation over a longer period of time? In Apple’s film, no statement was longer than 20 seconds without a cut, which any smartphone can actually manage without becoming asynchronous. But unfortunately, practically all of today’s smartphones do not provide a fixed frame rate, but a variable one (VFR = Variable Frame Rate). Although DaVinci Resolve (DR for short) simply reports a highly precise 60,000 under “Shot Frame Rate”, the iPhone is unfortunately no different. Before recording, we switched to aeroplane mode and closed all other apps to rule out any interference.

However, a further measurement showed that this was not necessary. In the free MediaInfo, the information is slightly more correct, showing a frame rate between 59.940 and 60.181 fps – that’s a deviation of far less than one per cent! A longer recording with a film gate showed that the variable frame rate in DR is relatively unproblematic. The sound ran continuously one frame early for over 30 minutes (at 60 fps!), only at one point did we have the impression that it tended towards two frames. Since in DR the picture occasionally moved one frame ahead of the TC flap, which we kept continuously in the picture, the editing software probably does react to VFR. This is usually manageable, but the iPhone is probably not quite suitable for uninterrupted concert recordings lasting several hours.

Cameras and lenses

A few years ago, the statement “My smartphone has three cameras and a LiDAR scanner” would probably have been met with incredulous smiles. But there are actually three lenses here, each with its own sensor, just like before. In fact, most of it is even identical to the iPhone 14, only the signal processing has been further developed. The ultra wide-angle lens with 13 mm, the regular 24 mm and the 77 mm in the iPhone Pro are identical to their predecessors.

Only the 120 mm in the Pro Max is really new and a very clever design with a multiple prism that basically folds a relatively long focal length into the flat body. To us, such a design seems potentially more durable than a miniaturised zoom lens like the competition (usually referred to as a periscope lens), as there are no moving parts other than the sensor. It is therefore not an optical zoom at all, as is sometimes incorrectly written. Please note that in the following we use the term KB full frame for the photographic full frame of 24×36 mm, because experienced film cameramen tend to mean the S-35 full frame. However, its diagonal is not always the same, but depends on the respective widescreen format. The information provided by Apple and in the apps refers to the equivalent focal length of KB full frame. The main lens 1× has a real 6.86 mm with f 1.78 and the ultra wide-angle 0.5 has a real 2.22 mm at f 2.2, the 3× has 9 mm with f 2.8 and the 5× has a real 15.66 mm, also at f 2.8.

The iPhones also offer a zoom across the lenses, but this is based on continuous digital scaling. Unfortunately, this shows a disturbing jump when switching to telephoto due to the parallax of the separate cameras. In addition, even with sophisticated algorithms, the sharpness drops slightly when the resolution of the sensor falls below 48 megapixels (mpx), which is particularly noticeable on the way to 5x. This is a disadvantage that longer focal lengths also have with conventional cameras: The 77 mm only focuses up to 60 cm at close range, the 120 mm only up to 1.3 m.

You have to be careful here: The iPhone should be set to “Keep setting” for macro control. Otherwise, only the small flower indicates that the macro range is switched to the ultra-wide angle camera, although the original lens is still selected. However, this only has 12 mpx, unlike the standard camera. This automatic function can be deactivated. With the default setting, the photos from the 24 mm lens are downscaled to 24 mpx and recorded in HEIC, the full 48 mpx are only available in RAW (DNG format). The main camera also delivers the best quality for video thanks to downsampling. And for the selfies: The front camera has 2.69 mm at f 1.9, so it looks like a 15 mm, but only has just under 8 megapixels (actually photo cells = Sensel).

Photography

The DP is not explicitly aimed at photographers, but in view of the different resolutions of the sensors, a few comments seem appropriate. In our test of the Sony A7SIII (DP 21:05), we already proved that megapixels aren’t everything. Just as the Sony “only” offers 12 mpx, images from the iPhone would also be perfectly suitable for an A4 page. For large posters, however, we would generally recommend professional cameras, even if Apple sees it differently in its own advertising.

It should be noted that the sensor has an aspect ratio of 4:3, which is even closer to the square than the usual 35mm with 3:2. The HEIC photos are very compact at around 2.8 MB, whereas DNG requires over 67 MB for the same subject. A section in 2,400×2,400 pixels from the iPhone at 120 mm (equivalence) with the same section from the Sony A7 IV with approximately the same focal length differs primarily, of course, in the real bokeh, which is not simulated here on the iPhone either. The slightly different colours can of course be adjusted, as both are raw images with plenty of colour depth.
The resolution of the test chart in 48 mpx from the 24 mm of the iPhone is excellent and shows no moiré, but slight colour noise even at ISO 320.

The counterpart in standard setting as HEIC with 24 mpx shows less noise, but colour moiré and clear edge sharpening. The modern HEIC compression is very high quality in itself, but unfortunately Apple applies all the “image enhancements” to these photos that the fast A17 processor is capable of, such as smoothing and sharpening contours or local colour tone mapping. Edge sharpening in particular looks just as distracting as it does on amateur-class drones. Of course, Apple also knows that the sharpening of details in faces is usually undesirable. Accordingly, skin tones tend to be smoothed out, which often gives them the infamous “plastic look”.

The design of these functions is certainly based on comprehensive market analyses, as the vast majority of users will simply take photos with their mobile phones and presumably find all this attractive. Professionals who don’t want to be deprived of the opportunity to tweak their pictures themselves will therefore have to work with RAW Max (as it is called in the menu) and plan the necessary space for it, either conveniently and expensively in the iPhone or more cheaply on an external storage device. However, raising your eyebrows at “computational photography” in general is not entirely appropriate. After all, most professional cameras have been correcting various lens errors in the background for years, primarily vignetting and distortion. This means that lenses can be optimised in other respects without being sinfully expensive.

The iPhone goes beyond this and achieves amazing things in two areas: bokeh and low-light. The ability to use the (actually relatively coarse) information from the LiDAR to subsequently influence and target the depth of field was already available on a number of predecessor models. But Apple has significantly improved the algorithms and, at least in photography, the results can now only be recognised as fake with intensive pixel peeping, even with hair and other difficult edge zones.

The results in RAW (DNG) are impressively good, while artefacts can be seen in HEIC. However, we also photographed the test panel at 320 ISO for the wide angle and 500 for the telephoto. Vignetting or chromatic aberrations are hardly recognisable and the distortion correction for 24 mm is also better than with the uncorrected professional lens. Even moiré is hardly noticeable at 48 mpx from 24 mm and the resolution is even slightly higher, as the A7IV only has 33. At 77 mm or 120 mm, the A7IV has a recognisable advantage over the 12 mpx of the iPhones, although it was at a slight disadvantage in terms of speed each time. A 13 mm was not available to us for the full frame, so no comparison was made.

The Sony’s HIF images look cleaner because, apart from subtle noise filtering in extremely low light, everything else can be deactivated. In low light, the iPhone does all by itself what, in principle, any camera on a tripod can do: A series of photos that are then combined to reduce noise. Red has been offering something similar for 10 years (see DP 13:03). With the iPhone, this also works with good stabilisation from reasonably steady hands, but you can get even more out of it with a tripod. Some even recommend this for astrophotography; there is even a free app for this, AstroShader, which can also save raw images.

Now, we have too much light pollution near a city for a great starry sky, but you can focus on the lights of the city itself instead. It’s impressive what such a tiny chip can do with the help of a computer. The tree in the foreground was barely distinguishable from the night sky with the naked eye. Apple is clearly making a virtue of the A17’s computing power out of the necessity of the tiny sensor and calculating far better images than you would expect from a mobile phone. However, these calculations take time and cannot be used for video. So how do the 15s perform when filming? We use the Sony A7 IV (see DP 22:05) as a reference here, even if some people seriously compare the images with those from an Arri Alexa in grading(is.gd/quazi_iphone_arri).

Film: Resolution

When filming from the high-resolution 24 mm sensor in UHD, the iPhone not only scales down the 48 million photocells (aka Sensel), but also crops slightly from the 4032 horizontal pixels available per photo to 3840. So if Apple had only added a few more Sensel, the iPhone could even deliver the 4096 pixels of the DCI cinema format. There would be more than enough vertical pixels anyway. Other formats are possible with third-party apps; we will go into this in more detail in the respective apps. The same applies to the Ultra-WW and the telephoto lenses, which logically have less oversampling, but are also slightly cropped.

The downsampling of the 48 mpx at least offers the possibility of noise reduction by integrating the neighbouring pixels or a digital 2× zoom without loss of resolution. However, when zooming through to 5× telephoto, not only does the parallax error mentioned above occur, but the resolution also collapses beyond 2×. Up to 3× on the iPhone 15 Pro, these effects are correspondingly less noticeable. A 28 mm and a 35 mm are still simulated quite well by both iPhones. This is comparable to Sony’s digital zoom in the camera, which works with similarly good algorithms, but cannot deliver the full resolution beyond around 1.7.

The resolution of the iPhones is also reduced at the higher levels of stabilisation. “Standard” only uses optical stabilisation by mechanically tracking the sensor, while “Cinematic” and “Extreme” additionally track a reduced area of the sensor surface. Unfortunately, “Extreme” in particular also shows overshoots – or is this intended to simulate the inertia of a Steadicam? Be that as it may, a small gimbal can do this better. Another disadvantage of the two higher stabilisations are massive deviations in the frame rate in the first second; the iPhone is obviously “stressed” here (thanks to Kurt Friis Hansen for these measurements).

The other sensors with 12 mpx naturally also offer less resolution reserve when filming, but are completely sufficient for UHD. For comparison: On S-35, the 5× lens of the Pro Max would correspond to about 75 mm, the 3× to about 48 mm. Because a suitable prime lens was not available for the Sony, we used a very high-quality zoom lens. However, this actually put the A7IV at a slight disadvantage in terms of speed. In all shots, the film images of the iPhones on the test chart show slight moiré and some loss of resolution due to scaling; the Sony obviously has better downsampling. The differences are small at 24, 77 and 120 mm and the moiré can practically only be seen on the test chart. Other cameras only deliver better results with very good downsampling of images in 6K or more.
It is obvious that a lot of calculations are also carried out on the images in log, as the charts show no vignetting and hardly any distortion, just like in photography. This is not to be expected with such lenses without software correction. The very high-quality Zeiss C/Y zoom on the Sony shows more optical errors due to the lack of software correction. It should be understandable that a lens with a real focal length of just under 16 mm (the 5×) without software tricks does not offer such a differentiated depth of field as a real 120 mm for KB full frame (see above). All lenses definitely show some breathing during focus shifts – if Apple doesn’t simulate this as well

Film: Compression

Here, too, the comparison with the Sony A7IV, which only records 10 bit with 4:2:2 as H.265 for maximum quality, ProRes is not available without an external recorder. Our endurance test was a water surface in the rain, which can really torture any GOP codec (such as H.264/265). The iPhone only recorded a good 7 MB/s at 60 fps in UHD, while the same motif was recorded in ProRes 422 HQ at 213 MB/s. The Sony A7 IV records this codec at 50 fps with 25 MB/s. Nevertheless, the typical artefacts, such as block formation, can hardly be seen in the correctly exposed motif on the iPhone. However, this does not only speak in favour of the quality of the encoder alone, as the finest details are already reduced before compression.

This becomes clear with a shot of the same subject that is underexposed by just one f-stop and corrected in DR. Despite 10 bit and 4:2:2, the signal in H.265 already shows signs of banding and is overall much less “full” and more coarsely structured than the version in ProRes. The same applies to noise, which is also filtered more heavily before compression and then “muddied” by the codec. The noise remains much finer and more “organic” in ProRes. It is therefore less disturbing in itself and is also more suitable for careful filtering in post-production. Consequence: If you don’t have enough storage space on location, H.265 is perfectly usable in Log, but should be exposed as correctly as possible. ProRes is considerably more tolerant, while Cinema P3 can record better H.265 at higher data rates (see p. 128).

Film: Motion display

We can be brief here and take the liberty of quoting the values from CineD (is.gd/cined_test) because our reference fan was not available. For UHD, our colleagues measured 5.3 milliseconds for the sensor at 24 mm, and even 5 ms at 77 mm and 120 mm. Our subjective impression from our own tests agrees with this. These are excellent values, also in comparison to conventional film cameras. The jelly images of older iPhones are definitely a thing of the past and action is the order of the day.
However, the Bentley film mentioned at the beginning was still noticeable with the jerking of fast movements at only 24 fps. A look at the individual images shows that they lack any motion blur. Even these tiny sensors are too light-sensitive today to comply with the filmic rule for sufficient motion blur in daylight. After all, mobile phones primarily use the time for exposure control, because the aperture is fixed. Without an ND filter, the only option is to regulate the ISO value, which the iPhones also make intensive use of. But this has consequences for the usable contrast range.

Film: Dynamic range and colours

The HDR shots from an iPhone with the corresponding screen are very impressive at first glance, but on closer inspection the sharpening becomes unpleasant. Added to this is the local tone mapping, which is particularly noticeable with intensely coloured light sources and soft colour gradients in the moving image. Only recording in Apple Log avoids these annoyances, which is why we limited ourselves to it for dynamic measurement. Apple Log is excellently tuned, it does better justice to the iPhone than an A7 IV with its S-Log3, because even for 10 bit it is still very flat. Incidentally, in the current version 18.6.4, DR understands the Apple Log with the colours in Rec 2020 with automatic colour management.

The amazing night vision in photos ends for the iPhone when filming, as there is no time for the exposure tricks from photography. We also used the Sony A7 IV for comparison. It is not quite as sensitive to light as the A7S and also works with noise reduction in low light. Compared to the iPhone, however, it shows the richness of detail that a full-frame sensor can achieve when collecting light.

However, we came across an interesting phenomenon that suggests multiple exposures in the highlights. When shooting at 50 fps in Log with a longer exposure time of up to 1/80 second, the highlights are consistently limited to around 70 per cent. If you now go to at least 1/96, the values expand into the space above. Hardly any further information can be found about this function, only CinemaP3 refers to it as EDR (Enhanced Dynamic Range).

In the documentation for developers, the term is primarily used for playback on HDR displays. Red has been offering something similar for recording under the name “HDRx” for some time, but probably does not hold a patent on this (as with the compression of RAW in the camera). Obviously, two (perhaps even more) readouts of the sensor are combined here as soon as the time until the next single image is sufficient. 7.5 milliseconds is obviously not enough, but 10 ms is fine. This is less noticeable at 25 fps, as the usual exposure time of 1/50 with 20 ms is easily sufficient. Here the effect already starts at 1/40, at 1/30 the remaining time is still too short – which mathematically corresponds to the behaviour at 50 fps with just under 7 ms. The resulting distribution across the log curve can be easily controlled using ND and ISO.

At the other end of the scale, it is noticeable that there is practically no recognisable noise. You can simply cover the lens with manual exposure and the line drops to zero. So there is obviously a lot of filtering going on in the camera, because such a small sensor cannot be so low-noise despite all the technical progress. This becomes quite clear when shooting at high ISO values with short exposure times, as the lower mid-tones are already very noisy. Medium ISO with longer exposure times using ND looks much better.

So there is no “correct”, native ISO value that always fits. The iPhone makes too many adjustments for this, even with Log. It all depends on where the important details are: If you want to bring out the highlights, e.g. add structure to a cloudy sky, a higher ISO value is appropriate. EDR reduces the contrast in the highlights, but the details are there. If the shadow areas are the most important thing, on the other hand, you should go for a low ISO in order to provide the sensor with enough light, then the highlights will be compressed, but largely without clipping.

In our exposure series with EDR, 2 f-stops overexposure and 3 f-stops underexposure were still easily correctable, the colours become weaker, but do not derail. Incidentally, they are excellent: especially in the critical area between red and yellow and thus with the skin tones, the iPhone shows a precision landing. Noise is visible in dark areas, but is fine-grained in ProRes and can be controlled in post. We worked with DR’s colour management and only corrected brightness and contrast. We measured an extremely high-contrast motif with a cloudy sky and molleton in the shadowy foreground with 9.5 f-stops – this would still be usable in post (the blue tone on the blacks is scattered sky blue).

After all these observations, I must also note that Gerald Undone and Patrick Tommaso’s categorisation of ISO 1,100-1,400 as the native value should be treated with caution. The two did not explain whether they used ND filters or varied the exposure time(is.gd/undone_iphone). But DXOmark also assumes 1,250 to 1,480. I tend to favour ISO 200-800 for noise reduction and use EDR with sufficiently long exposure times. Of course, ND filters are the order of the day in bright light, otherwise the typical staccato films from action cams will result.

Film: Equipment

Basically, the only indispensable accessory for good pictures is an ND filter. But a variable ND consists of two polarising filters and has the corresponding side effects, from the infamous “X” to altered colours (warmer ones in our case). Conventional ND filters are better. We were unable to observe any noticeable IR contamination with an inexpensive set, even at 10 f-stops; Apple probably has this well under control. Filters can be attached with a simple clamp mount, but this should not cover too much of the interface on the side of the screen. In addition, such a mount does not fit well with the usual transparent silicone covers, as stray light can then enter from behind. Such covers are also not conducive to cooling in warm environments.

If it’s not too warm, a thin, black cover would be useful. If you want it more perfect, you’ll be back on the road to “Frankenrig” at some point, because the accessories industry has immediately adapted to this. But there are alternatives that can still be put in your pocket as a protective frame without all the accessories. The company Beastgrip has been involved in the development since the Bentley advert and also produced the universal cage that was used for Apple’s event. Their “Beastcage” will soon be released in specific versions for the iPhone 15 Pro and Pro Max and offers a very sophisticated system, including filters, macro lens and anamorphic lenses. Another provider is Freewell with the “Sherpa” – the name of course says it all, especially for the iPhone. It works with clever magnetic holders for filters or lenses.

Freewell not only offers filters, but also anamorphic lens attachments. These are available in both 1.33 and 1.55 and at Freewell optionally with streaks in classic blue as well as in “gold”, i.e. warm yellow. However, the streaks are very pronounced, including the typical side effects such as edge blurring and distortion. However, the 1.55s in particular offer much more horizontal image area, even if they are not easy to align. There is no gain in resolution associated with the anamorphic lenses in ProRes Log. When recording film, the full area of the 4:3 sensor can only be used with H.265 in 8-bit HDR or Rec 709 and then shows all the typical “blurring” effects.

Be careful if you want to use existing filters with a larger diameter via the step-up ring: the 24 mm main camera can then vignette. In addition, the stabilisation of the iPhone does not work with anamorphic lenses and sometimes the iPhone “sees” the adapter as an object and goes into macro mode. So is this more for wannabe Spielbergs? The only filters you really need are NDs and possibly a circular polarising filter, as neither of these can be replaced in post. There are excellent software solutions for everything else, such as “Scatter”.

Advantages and disadvantages

So where do we stand with such a powerful camera phone? Well, first of all, an “always with you” camera is definitely better than no camera at all. In low light, a good hybrid camera like the Sony A7 IV is still clearly superior and you have an almost infinite choice of lenses, but all of this is still much more expensive, even compared to Apple’s prices. If you know your camera really well, you can
blindly with the viewfinder and at the same time see the subject in isolation without distraction. Nowadays, the monitor can usually be swivelled and rotated, but the iPhone’s monitor is still visible even in sunlight.
But you don’t always have all those nice lenses with you. Dust on the sensor when changing lenses is also not an issue with mobile phones. The iPhone is even waterproof, where larger cameras can only withstand splashing water. And you hardly stand out with it, whereas a “real” camera with a longer lens might be eyed suspiciously or even rejected at the entrance. The iPhone 15 can achieve cinema quality in the documentary sector and would certainly have been a dream come true for authors such as Michael Winterbottom with “In This World” or Wim Wenders with “Buena Vista Social Club” – these films were released to the cinema in large parts by DV.

Commentary

Apple has managed an amazing balancing act with the iPhone 15 Pro: 99 per cent or more of users get nice pictures without any great expertise and professionals get a camera that, with the right app, can do far more than a conventional smartphone. The pig called “game changer” is driven through the digital village almost every day – mostly ephemeroptera (mayflies). But the Pro models of the iPhone 15 could actually signal a paradigm shift, similar to the Canon 5D Mark II 15 years ago. Just as that marked the beginning of hybrid photo/film cameras, the iPhone is now a working tool for professionals. And unlike a standard camera, you can even use it to make phone calls..

The splash-proof iXM microphone from Yellowtec is a combination of handheld microphone and audio recorder. The device was developed at the suggestion of public broadcasters who wanted a reliable and easy-to-use recording device for interviews.

A standard SD card, as used in DSLR cameras, is inserted at the bottom as a recording medium. It is operated with large buttons, which are also completely silent. Operation is almost self-explanatory. However, the device must first be switched on by pressing the REC and STOP buttons simultaneously. Depending on the size of the memory card, the boot process takes a few seconds – in my case around 24 seconds – but as the internal battery lasts for 18 hours, you can do this in advance. And if the battery is empty, the device automatically switches to the additional battery compartment with three standard AA batteries without interruption or cracking.

When the boot process is complete, this is indicated by the flashing red LED. Then press the REC button and the recording starts. This is indicated by the permanently lit red LED. Press the REC button again to set a marker. In addition, a pre-roll time of up to 30 seconds can be set via the software so that the beginning of an interview cannot be missed.

The device has a 3.5 millimetre yellow line-in socket on the underside, which can also be used to record line signals – but only in mono. You should make sure that the level is set correctly via the software. On delivery, it was far too low for me, so that the sound from the headphone output of my ThinkPad was barely audible. But to check this, there is a 3.5 millimetre headphone socket marked in green right next to it, which can be used to listen to the sound during both recording and playback. Playback and volume are controlled using the buttons on the back of the device. A voice announcement tells you which file is being played. This is very practical and replaces the missing display. The markers can also be started using the control buttons.

The solid metal housing has an edge to prevent the microphone from rolling away so easily. The microphone capsule from Beyerdynamic is screwed on and is available in three characteristics – omnidirectional, cardioid and supercardioid, each with premium electret condenser capsules, which we received with our test set
received with our test set. There is also a slightly cheaper dynamic version of the two cardioids from Yellowtec.

And how does the microphone sound?

The integrated LEA levelling technology ensures distortion-free levelling and the suppression of interference signals. Even if you shout into the microphone and it already sounds distorted on the headphones, the recording itself is distortion-free because the LEA technology is bypassed in the monitoring path to avoid latency problems. The level control is rather slow. Technically, this is realised by two differently levelled preamplifiers, between whose signals the best levelled signal is always written to the memory card, depending on the level. This means that there is no pumping as with automatic levelling or compression from other devices and the volume can be subsequently adjusted in audio software. This creates a homogeneous sound image. Adjustments are only made in the event of extreme level fluctuations anyway. Even with the omnidirectional capsule and strong background noise from a trade fair, speech intelligibility is still very good.

Software

To make further settings, you must first install the software. There is user and admin software. After starting the user software, you should first click on the Help button. The help window will then open, in which you can find out what is possible and how you can connect the iXM Micro in the first place. To do this, you first have to find out which com port the microphone is connected to in the Windows Device Manager. Unfortunately, this is not particularly intuitive and should at least be mentioned in the briefing book supplied. After that, you can also fine-tune the LEA Engine via the software and save the settings in the nine different presets, which you can then call up with the user software.

To get a good comparison, my friend Jan Kalt set up two classics next to the iXM in the recording room of his Schraubfabrik recording studio in Musikpark Mannheim and recorded them. A Neumann U87 and a Shure SMB7. I also plugged in a DJI and a Hollyland Lark MAX wireless microphone. While the two wireless microphones are easy to understand but sound quite mid-range, just like the SMB7, which is at least a little more brilliant, the iXM sounds full and transparent. Not quite as brilliant as the U87, of course, but a little fuller in the bass – but that’s my subjective listening impression. If you want to see for yourself, you can watch the video on my postproduction-tutorials.net website.

Conclusion

The iXM from Yellowtec really is a very good microphone, especially for voice recordings, even in difficult environments. It is optimised to work intuitively in every situation, which is invaluable, especially for non-repeatable events and interviews. However, this also comes at a price. It costs 899 euros with a dynamic capsule and 1099 euros with a premium capsule. You should definitely order a microphone holder, as the microphone is really very thick and does not fit into a standard holder. You shouldn’t skimp on pop protection either, as the supercardioid is particularly sensitive to pop sounds. It is a pity that the microphone cannot be used as an audio interface via USB. Perhaps there will be a firmware update in the future. After all, you can already see the level change in the iXM User software. If you want something a little cheaper, the iXM Podcaster with dynamic cardioid capsule is an alternative for 639 euros. However, you will then have to do without the internal battery and the interchangeable microphone capsules. Otherwise, the functions are the same.

In a nutshell, the Yellowtec Mika system is a set of columns, clamps, arms and brackets that mount microphones and other heavy equipment to a desk, wall or wherever and position everything as it should be in use.

What is it actually? It’s called “m!ka”, but internal punctuation… we call it Mika. You can find the overview at is.gd/yellowtec_mika – but that’s not all, because the accessories are where it gets really exciting – mounts, various arms and all kinds of accessories. Brackets are rods, clamps and rails that can be used to attach everything in the system, from the ceiling to the wall to the table (with and without holes) and so on. The arms are mounts for microphones, screens, cameras and tablets. And then there are on-air systems, monitor mounts, tablet clamps and much more.

The arm

Available in a wide range of sizes and designs, the microphone arms can support loads of up to 1.5 kg (there are probably few microphones out there that can go that far). There are also monitor arms in the range that can support up to 8 kg with VESA-compatible mounts – that’s easily enough for normal screens, and no-one hangs broadcast monitors from FSI, Eizo, Sony etc. on an arm anyway. The arm itself is made of sturdy aluminium and finished with a 5/8-inch steel thread – that’s “audio” standard. If you want to attach something else, you can find thread converters in all sizes – or probably already have some lying around somewhere.

The joints are made of solid plastic and can be locked securely, or, if the system is balanced
the system is balanced (counterweight screws are in the base of the respective arm), they can also be left open in order to have a movable arm – for example, if you want to reposition the microphone frequently for different people. The arm is not all “fully” rotatable, but the radius is sufficient and is described precisely in the documentation.

The built-in LEDs on the microphone arms are a delicacy – the indicators are directly above the microphone and switch with the “On Air” lights – each microphone shows whether it is live, communicates with the various mixing consoles (of which Yellowtec also has some in its range, but that would go too far for us VFXers) and thus provides built-in feedback.

In the test, we set up a column, the cam stud and three arms. The entire system is very stable and contains a complete set, from the cable holder to the shim disc. Everything is designed for “hard” use and is made of high-quality materials such as aluminium, solid plastic and stainless steel screws. Nothing wobbles, squeaks or rattles.

In our test version, we had several extension variants available. The system can be attached to the ceiling, wall, table or floor as required. The large columns in particular can be used to create flexible work areas that offer monitors and storage space without everything standing on the table and transmitting vibrations to the microphone or camera.

Another advantage is that you can order and extend everything individually. If you need another monitor, that’s no problem. You can simply add a bracket and an arm and attach it
to the column. Markers, screws, brackets and other elements can be exchanged, extended or personalised.

Enough praise, now for the grumbles! The first point is that it is relatively expensive compared to gaming/streaming accessories. Although it is more stable and flexible, it is not a bargain. If you can’t afford it as essential equipment, you might want to recalculate.

The second point of criticism concerns the selection. There is currently only one USB-C arm, while the rest of the range is supplied with a 5-pin XLR or XLR-compatible cable. However, this is probably only a matter of time. It would be desirable if the range were expanded.

And there is another point to criticise: Three different Allen key sizes are required for the table column mount, the mounting screws on the column and the settings to achieve a “normal” setup. You should therefore be careful not to throw any of the spanners away during assembly, as all of them are required.

If you browse the Yellowtec website, you will also see the “iXm” series – microphone-shaped recorders with built-in autolevelling. If you’re thinking: That can’t be right, you always have to level audio recordings! Nils Calles is already testing the device and will let you know how it works on page 128 of this issue.

Conclusion

If you want to set up a “professional studio”, there’s no getting around them – it’s not for nothing that the arms are used by all the well-known broadcasting companies. The interaction between the arms, the mixing consoles and the mounting options in all directions, including stands, brackets and the like, is perfectly coordinated – and stable enough that the broadcasting station is more likely to collapse than the brackets.