What in fact is the "log" and how should we work with?

[Lucas Piont]

  I just started recording in RAW now with the ML and, coming from a photography background, I find myself completely lost on how to edit my scenes, since every damn tutorial tells me to apply the log profile to my raw scenes and then apply a correction LUT and only then do the color grading.

  With some research I discover that the "log or linear" feature is already assigned on the raw file itself in a certain way, like I found in this commentary:
_______________________________________

"LOG is not to preserve highlight it is to evenly spread data in non linear way to have an even amount of data for each stop of dynamic range.

Lets say we have 12 stop of dynamic range each stop causes the light to double which takes up a bigger number. So if we have 10 bit non log linear video we would have 512 values assigned for the last stop of light and about 32 values for your skin tones which is complete waste since it is a half of your values consumed by the brightest part of your image and shadows/midtones are poorly represented.

In order to combat that engineers decided to write data with LOG function which effectively gives each stop more space to be reprensted with binary data. And that gives us more or less the same range of values to wirte each stop of dynamic range.

LOG and RAW are different things, sensor can see only in linear way which is doubling value for each stop of light, so it in order to put 16 stops in 12 bit file you would encode 16 bit linear data with log function and it will be writen as 12 bit "log" file."
________________________________________


  Assuming that the statement described above is correct, it makes everything even more confusing. If Log is an away to organize the raw data (and not a gamma curve, profile or whatever) what is the "log" that we applie on the editing programs? Or the Log "file" that some cameras rec in?

I don't even know if my camera (6d) is recording in a logarithmic or linear raw way, so is even correct apply a log profile on it? Why should I do this when I can simply take a LUT that was designed to be aplied directly to the raw file? Like we do with presets in photography!


I'm trying to understand all this concepts, but everyone says something different... So here's what I've understood so far: Is it just a convention due to the popularity of compressed log filming over the past few decades? So most of the LUTs are just already pre-designed to be applied above the log profiles?

OR my other theory and line of reasoning: apparently ARRI (and maybe other brands) created their own RAW file (ARRIRAW) to be used with their own log curve (ARRI Log C) to be used with their own correction LUTs, so they can have a lightweight yet flexible final file to optmize the worflow. Was this the motivation behind of still using the log curve to this days? And this is the only correct way to use it? (Within a system already calculated and pre-designed by the brand to prioritize the workflow)
If that is the case, I still don't understand how this could be more practical than applying a LUT on top of the already RAW file... Would that be because the files would be too heavy to send to the colorists if they were still in RAW? (in a big company scenario)



To summarize and get to the heart of the matter:
How and when exactly should we use the "log" in our RAW files? And what are the real advantages it can bring to our real world RAW editing workflow? (homemade scenario)

Concepts to be related:
RAW-Log/RAW-linear  VS  Compressed log filming  VS  Log curve (or profile, etc)

p.s.: Sorry for my English

[names_are_hard]

MLV raw video is not recorded in log.  You have no choice in this.  You cannot use log in MLV files.

I will leave other people to describe what might be sensible editing choices.

[ML700D]

In the context of digital cameras, both RAW and Log formats capture a wider dynamic range than standard video formats, offering more flexibility in post-production. RAW captures the unprocessed data directly from the sensor, while Log is a video format that applies a specific gamma curve to the data to preserve dynamic range.
Here's a more detailed breakdown:

RAW:
Unprocessed: RAW files contain the raw, unprocessed data from the camera sensor, including all the color and light information captured.

Maximum Flexibility: This offers the most flexibility in post-production for adjusting white balance, exposure, and color grading.
   
Large File Sizes: RAW files are typically large due to the vast amount of data they contain.
   
Not Directly Viewable: RAW footage cannot be directly viewed or played back without processing.
   
Requires Conversion: RAW footage needs to be converted to a viewable format before editing.

Log:

Video Format:
Log is a specific video format, meaning it's designed to be directly viewable and playable.

Dynamic Range Preservation:
Log profiles are designed to capture the widest possible dynamic range, meaning more detail in shadows and highlights.

Flat Image:
Log footage often appears flat and washed out when viewed directly, requiring color grading to bring out the colors and contrast.

Post-Production Flexibility:
Similar to RAW, Log provides significant flexibility in post-production, particularly for color grading.

Smaller File Sizes:
While still larger than standard video formats, Log files are generally smaller than RAW files.

Key Differences:
RAW is not video: RAW is sensor data, not a video file. Log is a video format.
RAW is uncompressed: RAW data is typically uncompressed, while Log is a compressed format.
RAW offers more control: RAW provides more control over the final look in post-production than Log.
Log is easier to preview: Log footage can be previewed on an external monitor, while RAW needs to be processed first.

In short: If you need the absolute maximum control and flexibility in post-production, and don't mind the large file sizes and processing time, RAW is the way to go. If you want a good balance of dynamic range, file size, and post-production flexibility, Log is a good option.

source: paste from google search 

[a.sintes]

Maybe this link will help a bit ("Picture profiles" chapter).
Initially, log profiles were specifically designed to retain more useful data when dealing with lossy compression algorithms (e.g.: h.264), nowadays it's also used over RAW data in case of decimation process (e.g.: internal 14-bits vs 12/10-bits RAW output).
It's a very smart way to save data storage with a good quality compromise.

[Walter Schulz]

In the context of digital cameras, both RAW and Log formats capture a wider dynamic range than standard video formats, offering more flexibility in post-production.

You begin this part with a mix of true and false info.
The part

In the context of digital cameras, both RAW and Log formats capture a wider dynamic range than standard video formats

is exactly what log formats are not doing!

The goal is indeed

offering more flexibility in post-production.

[iaburn]

You begin this part with a mix of true and false info.
The part  is exactly what log formats are not doing!

Not sure if I understood you correctly, but Log is indeed designed to capture a wider dynamic range. While standard gamma curves, designed for direct viewing like rec709 can capture 5 or 6 stops of DR, Log gamma curves are designed usually to capture 12+ stops of DR.

Log is a mathematical function that plays along very well with binary representation, as the amount of data that can be stored increase exponentially with every bit added (one stop per bit), and the Log function "counteract" this by mapping every stop of information to roughly the same amount of values.

To answer the original question on the thread, RAW video is just a series of RAW photos, so you just edit RAW video like you are used to edit your RAW photos.

[koljanych]

BM RAW, ProResRAW etc, these formats with applied LOG gamma curve. MLV is linear.

if we talk about developing in an intermediate LOG format (CLOG, SLOG, FLOG) - this is done for subsequent color correction during editing.
personally, I don't see the point in this for myself - it's more convenient to immediately develop MLV as I need - in MLVApp, AdobeAE, Fast CinemaDNG. if at the time of development there is not even a rough idea of ��what the final color correction will be - then the intermediate LOG format may make sense.

if we talk about shooting in LOG formats, with h265 codecs, let's be honest, LOG is RAW for the poor. it turns out YUV with beautiful DD values, but the color in them is not even close to stretching as in RAW. with experience with RAW photography, this is obvious.

in 2025, there are no problems with processing ProResRAW, BMRAW - you can immediately edit these formats in Premiere. mlv cannot be imported directly into Premiere, so in my opinion the easy way is to learn MLVApp, export from it to ProRes 422, and happily edit in Premiere.

[Skinny]

How and when exactly should we use the "log" in our RAW files? And what are the real advantages it can bring to our real world RAW editing workflow? (homemade scenario)

I think, in the homemade scenario if we are going to use the MLV App, we have at least four choices:

1. Simple - import MLVs, color grade in MLV App, export in your preferred format, done.

2. Just export as cdng (it is still raw) and do all processing and grading in other apps (resolve, etc). Some people prefer this option.

3. Select a LOG preset in MLV App (Alexa log, etc), export in ProRes 4444, DNxHR or something appropriate. This way you can do all color grading in your NLE. But personally I think you can achieve better results working with RAW instead.

4. If you like a particular LUT but it is designed for the LOG footage, you can select a Log preset in MLV App, then apply your favorite LUT in the LUT section. And here is the thing - if you want to change how it looks now, select "allow creative adjustments". The picture will go very dark, no problem - set "dark strength" to 0, it should return to normal. Now you can do all your color grading in the MLV App with all raw benefits, but you still have your favorite LUT on top of it. Maybe this is not 100% technically correct thing to do, but I really like what you can get with this option.

[ML700D]

You begin this part with a mix of true and false info.
The part  is exactly what log formats are not doing!

The goal is indeed

ah ic.. thanks for your correction walter 

I was curious and asked google gemini ai "can you explain how log format capture wide dynamic range?"

here the answer:
Of course. This is the heart of why Log is so powerful. Explaining how it works requires understanding a little bit about how digital video handles light.

Let's use a great analogy: A Bucket and a Water Hose.

The Bucket is your video file. It has a fixed capacity (e.g., a 10-bit file can hold 1,024 distinct levels of brightness).

The Water is the light from your scene, from the darkest shadows to the brightest highlights.

The Hose is the "gamma curve"—the rule that decides how to put the water into the bucket.

1. The Problem with Standard Video (Rec.709)

Imagine your hose is a standard, rigid metal pipe.

It pours water into the bucket at a constant, predictable rate.

When it gets to the dark parts of the scene (a trickle of water), it fills the bottom of the bucket.

When it gets to the bright parts (a strong flow), it fills the top of the bucket.

Here's the problem: The real world has an enormous range of light. What happens when you point the camera at the bright sky? The flow of water is a fire hose blast! Your rigid pipe can't handle it. The water violently overflows the top of the bucket.

This "overflow" is clipping. All the detail in the bright sky—the subtle clouds, the color gradients—is lost. It's just recorded as "completely full" (pure white). Similarly, the darkest shadows might not even be enough of a trickle to register, so they become "completely empty" (pure black).

This is what a standard video profile (like Rec.709) does. It's designed to look good on a standard TV, but it throws away tons of data at the extremes to create that pleasing contrast.

2. The Logarithmic Solution

Now, imagine your hose is a highly advanced, flexible, and smart rubber hose. This is your Log gamma curve.

Instead of pouring water in at a constant rate, this smart hose does something clever: it redistributes the flow.

For the Shadows (the trickle): The hose squeezes itself very narrow. It uses only a tiny portion of the bucket's capacity to store all the dark information.

For the Mid-tones (normal flow): The hose is at a normal width.

For the Highlights (the fire hose blast): The hose magically expands, becoming incredibly wide. It's able to capture that entire blast of water and carefully guide it into the top portion of the bucket without spilling a drop.

This is the essence of how Log works. It's a non-linear encoding process.

The Technical Explanation: The Gamma Curve

A gamma curve is a graph that maps the brightness from the real world (Input) to the signal level recorded in your video file (Output).

(This is a simplified visual representation)

The Horizontal Axis (X-axis): The amount of light captured by the sensor, from pure black to blindingly bright.

The Vertical Axis (Y-axis): The brightness level recorded in your digital file (from 0 to 1023 in a 10-bit file).

Standard Rec.709 Curve (The Rigid Pipe):

It has a steep curve in the middle and then hits a hard "ceiling."

Any light brighter than that ceiling is simply recorded at the maximum value (1023). All the information above that point is clipped and lost forever.

Logarithmic Curve (The Smart Hose):

It has a very gentle, gradual slope.

Most importantly, it has a long, soft "shoulder" at the top. It doesn't have a hard ceiling.

It dedicates a huge amount of the recording space (the Y-axis values) to the brightest parts of the scene (the far right of the X-axis).

It squeezes the vast range of real-world highlights into the top end of the digital file's capacity.

Why Does it Look "Flat"?

Because of this extreme compression of highlights and shadows into the top and bottom of the file's data range, the resulting image looks like it has no contrast. The mid-tones are effectively stretched out.

This flat, desaturated look is a necessary side effect of the primary goal: maximum data retention.

When you take this Log footage into your editing software and apply a "conversion LUT" or do a color grade, you are essentially applying an inverse curve. You are telling the software: "Expand the highlights and shadows that were squeezed into this file, and display them in a way a normal screen can understand."

Because no data was clipped (the bucket didn't overflow), you can now see all the beautiful detail in the clouds that a standard profile would have thrown away.

In summary, Log captures a wide dynamic range by using a clever encoding curve that prioritizes information in the highlights and shadows, squeezing a vast range of light into a standard video file without clipping.

maybe this is what it meant..