Log vs. Raw: What to Actually Shoot and When

A director of photography says "let's shoot raw, just to be safe." The cards fill fast, the DIT cart groans through the day, and three weeks later the edit is still waiting on proxies because nobody budgeted the transcode time. The final grade is a clean, natural corporate look. Log would have delivered that exact frame. Raw did not make the video better. It made it slower, heavier, and more expensive, and the client paid for latitude that never left the drive.

That is the whole argument in one paragraph. Raw is not the premium option you upgrade to when the budget allows. It is a specific tool with a specific cost, and the job of a working production company is to know when the cost buys something real and when it just buys reassurance. This post breaks down what log and raw each give you in the grade, what they take from you in storage and turnaround, and how to decide per job instead of defaulting to the heaviest format on the menu.

The question is not which is better

Both formats exist to solve the same problem: a camera sensor captures far more tonal range than an 8-bit delivery file can hold, and you want to keep that range alive until the grade. They just solve it at different points in the pipeline.

Log is a way of encoding a normal video file. The camera still debayers the sensor, bakes in a white balance and ISO, and writes a compressed codec like ProRes or a camera-native intra-frame format. What log changes is the transfer curve. Instead of spending most of the file's data on the mid-tones the way standard video does, a log curve like ARRI LogC, Sony S-Log3, RED Log3G10, Canon C-Log3, or Blackmagic Film redistributes the tones so highlights and shadows survive. The result looks flat and desaturated on set, which is the point. It is a container built to be graded, not watched.

Raw skips the bake. Instead of processing the sensor readout into a finished image, the camera records the sensor data itself, or a lightly compressed version of it, and defers the decisions to post. Debayer, white balance, and often ISO become adjustable later, non-destructively. ARRIRAW, REDCODE RAW, Blackmagic RAW, ProRes RAW, and Sony X-OCN are all versions of this idea at different compression ratios.

Here is the part the "raw is safer" reflex skips: a good log codec already holds the dynamic range you shot. ProRes 422 HQ off a modern sensor carries 12 to 14 stops without breaking a sweat. On a normal delivery, graded to a normal look, log and raw resolve to the same image. You do not see the difference. You see it only when you push the frame past where the baked decisions were set, and most jobs never push that far.

What each one actually buys you in the grade

The gap between log and raw is not quality at the finish line. It is how much abuse the frame survives on the way there.

Log gives you real latitude. You can lift shadows, pull back a slightly hot sky, warm or cool the image, and shape contrast hard before it falls apart. For the overwhelming majority of commercial work, a branded video, a testimonial, a product spot, a corporate piece, a social cut, that latitude is more than enough. The white balance you set on the day was correct, and you are grading toward a look, not rescuing a mistake.

Raw gives you latitude at the extremes, plus a few things log cannot do at all. You can shift white balance by thousands of Kelvin after the fact with no penalty, because it was never baked. You can recover a blown window or a crushed shadow with more room. You can rate the image differently in post because ISO is metadata, not a committed value. And raw keys and composites better: a green screen, a sky replacement, a heavy VFX plate, or a shot that will be relit in the grade all benefit from the extra bit depth and the pre-debayer data.

So the real question is not "do I want more latitude." Everyone wants more latitude. The question is whether this specific job will ever spend it. A stylized music video that gets relit shot by shot will. A two-camera talking-head day graded to look like a nice office will not. If the grade is going to be a natural, balanced look inside the range you exposed for, raw and log deliver the same frame and only one of them cost you a second hard drive.

The cost nobody prices at the bid stage

Raw's price tag does not show up in the camera rental line. It shows up in storage, in the DIT cart, in post hours, and in archive, and it compounds across every one of them.

Start with raw data rates. A log delivery codec like ProRes 422 HQ at UHD runs roughly 900 megabits per second, about 113 megabytes per second, which is close to 0.4 TB per recorded hour. Raw ranges wildly by format and compression. Blackmagic RAW at a heavy 12 to 1 ratio can sit near 0.14 TB per hour, genuinely light. ProRes RAW HQ runs closer to 0.74 TB per hour. ARRIRAW at 4.6K lands around 1.24 TB per hour, roughly three times the ProRes 422 HQ footprint for the same running time.

Now put that on a real day. Commercial shoots record a lot of footage relative to what they use: a lot of takes, safety, and coverage. Call it about 1.5 hours of recorded material per camera, two cameras. In ProRes 422 HQ that is around 1.2 TB of camera originals. In ARRIRAW it is closer to 3.7 TB. And camera originals are the smallest number in the chain, because you never keep one copy of anything.

The industry standard is 3-2-1: three copies, on two types of media, one off-site. That triples the footprint before post even opens the files. The ARRIRAW day is now roughly 11 TB provisioned, the ProRes day closer to 4. Add proxies so the edit can move, add the color pipeline's cache and renders, add long-term archive, often LTO tape for anything that has to survive past the project, and the raw job's storage bill is not double the log job. It is several times over, at every stage, for years.

This is the part clients and first-time producers underprice, because the camera looked the same on set. The cost is downstream and it is real.

Turnaround is the tax you feel first

Storage you can throw money at. Time you cannot, and raw spends more of it before anyone touches the grade.

Log footage is close to edit-ready. Drop it on a reasonably specced machine and cut. Raw usually cannot be cut natively without a serious workstation, so the workflow adds a transcode-to-proxies step at ingest, which on a multi-terabyte day is measured in hours, not minutes. Somebody has to run it, verify checksums on the copies, and manage the media. On a raw job that somebody is a DIT, a data-management specialist whose day rate runs roughly 500 to 850 dollars depending on the market, plus the cart and drives. On a log corporate shoot you can often skip that role entirely and let the shooter or an AC handle offloads.

Then the grade itself is heavier. Raw means debayering every frame, a more demanding color pipeline, and a colorist working with more data, which is exactly what you want when the job needs it and pure overhead when it does not. A same-week social deliverable does not have room for a raw ingest and archive cycle. A long-form brand film with a month of post and a real color session does.

Turnaround is where the "just to be safe" decision gets expensive quietly. The client feels a slow edit long before they would ever have felt the missing two stops of highlight recovery they were never going to use.

A framework for deciding per job

Stop asking which format is better and start asking what this specific job needs. Five questions settle almost every case.

• How far does the grade get pushed? Natural, balanced look inside your exposure: log. Heavy stylization, aggressive relighting, or an HDR deliverable that needs every stop: raw earns its keep.
• Is there VFX, keying, or compositing? A clean edit and a color pass: log is fine. Green screen, sky replacements, or heavy plates: raw's extra depth pays off in the composite.
• What is the turnaround? Same-week or same-day delivery: log, because you do not have hours to burn on raw ingest. A month of post with a real color session: raw fits.
• What is the storage and infrastructure reality? A laptop and a couple of shuttle drives: log, and do not pretend otherwise. A DIT cart, RAID, and an LTO archive plan: raw is supportable.
• Who is on the crew? No dedicated data wrangler: lean log, because raw with nobody managing it is how footage gets lost. A DIT on set: raw is handled properly.

The pattern underneath these is simple. Raw is the right call when the job will actually use the latitude and the infrastructure exists to carry it. When either of those is missing, log is not the compromise. It is the correct answer, and choosing it is a sign you priced the job honestly.

There is also a middle ground worth naming: light-compression raw. Blackmagic RAW at a high ratio, or REDCODE at a friendly setting, gets you most of raw's post flexibility at a data rate that sits between log and full-fat raw. If you want raw's white-balance and ISO freedom but cannot carry an ARRIRAW storage bill, that is often the honest split.

What we actually shoot, and when

Most of the work a production company ships lives comfortably in log. Branded content, testimonials, product and corporate pieces, event coverage, and social cuts are graded to look correct, not reinvented, and a strong log codec holds everything that grade will ever ask for. Shooting those jobs raw does not raise the ceiling on quality. It raises the floor on cost.

Raw comes out for the jobs that spend what it costs: high-end commercials with a real color session, VFX-driven work, HDR deliverables, and anything where the grade is a creative act rather than a correction. On those, the storage, the DIT, and the post time are line items you planned for, and the latitude is the reason the job looks the way it does.

The move is not to pick a favorite and shoot it every time. It is to make the call per job, out loud, at the bid stage, so the storage and the turnaround are budgeted before the cards start filling. Raw is not the safe default. The safe default is knowing which one this job needs, and being able to say why.