Shooting RAW Video — A Technical Guide for Professional Videographers
RAW video captures sensor data prior to camera-specific processing (debayering, color matrix, sharpening, compression and gamma transform). It’s a fundamentally different artifact than conventional camera codecs and is intended for maximum post-production latitude. This guide covers the real technical advantages and disadvantages, practical trade-offs and decision criteria for professionals.
What “RAW” actually is
Sensor-level data: per-pixel photodiode values (usually organized as a Bayer/CFAs pattern) plus metadata (exposure, ISO/gain, white balance, lens metadata, frame timing).
Not a single standard: formats include CinemaDNG, REDCODE RAW, Blackmagic RAW (BRAW), ProRes RAW, Fuji F-Log RAW variants, ARRIRAW, RAW from mirrorless/DSLR manufacturers. Each uses different quantization, entropy coding, and optional sensor-side pre-processing (e.g., linear vs. log encoding, noise reduction).
Raw pipeline tasks postponed to post: debayer, white balance, color transform, highlight reconstruction, denoising, chroma upsampling, and gamma/log conversion.
Advantages (technical)
Maximum tonal and color latitude
Full per-channel sensor readout: preserves extended dynamic range and highlight roll-off for reconstruction and highlight recovery.
12–16+ stops accessible depending on sensor headroom and RAW bit-depth (12–16+ bits typical). Floating- or high-bit encodings retain subtle midtone gradation and reduce banding after heavy grading.
Independent white balance and color temperature control
WB is a metadata parameter applied in debayering rather than baked, enabling accurate balance adjustments without chroma artifacts.
Better denoising and grading
Access to linear sensor data enables temporal/spatial denoising algorithms to perform more effectively (noise statistics better understood), and advanced color transforms (ACES, camera-id transforms) work on native linear data.
Superior chroma reconstruction and color fidelity
Debayering in post allows selecting advanced demosaicing algorithms tuned to the shot (edge-preserving, deep-learning debayers), improving detail and reducing color moiré.
Flexible ISO/Exposure manipulation
When ISO is stored as metadata with linear sensor data, pushing/pulling exposure in post is more robust; highlight reconstruction algorithms can succeed where baked codecs clip.
Future-proofing and archival
RAW retains maximum metadata and sensor information suitable for future-grade processes (HDR mastering, improved debayering algorithms).
Efficient grading workflows with camera raw transforms
Using standardized pipelines (e.g., ACES) preserves intent while leveraging device-specific input transforms.
Disadvantages (technical)
Massive data rates and storage footprint
RAW can be several× to an order of magnitude larger than compressed log codecs. Example: modern 6K/8K RAWs frequently require 1–6 GB/min depending on compression and bit-depth.
Requires high sustained write speeds (NVMe, CFast 2.0, CFexpress) and large-capacity long-term storage (RAID arrays, LTO).
Processing and hardware demands
Debayering and decoding (especially for high-bitrate or high-res RAW) are CPU/GPU intensive; realtime playback may require proxies or powerful workstations with GPU acceleration and ample RAM.
Hardware transcodes and accelerated decoders vary by vendor — e.g., BRAW/ProRes RAW have GPU/hardware acceleration in specific NLEs, while CinemaDNG relies heavily on CPU.
Complex, heavier workflows
Requires proxy workflows, color management, careful media management, and additional archiving steps; increases project complexity and turnaround time.
Need for consistent LUTs, IDTs (Input Device Transforms), ACES/OCIO configurations; misconfiguration can cause unpredictable color shifts.
Increased risk on-set (write failures, heat)
Sustained high throughput increases thermal load and potential for dropped frames; card/controller failures can be catastrophic. Monitoring and redundancy are essential.
Interoperability and software dependency
Proprietary RAW formats may lock you to certain NLEs or color tools; older formats may become unsupported without conversion. RAW-grade software support varies.
Loss of in-camera processing conveniences
No in-camera sharpening, noise reduction or color baked-in; often requires more post time to produce deliverables that match client expectations without proper workflows.
Technical trade-offs and nuances
Bit-depth vs. effective SNR: 16-bit encodings don’t magically increase sensor SNR; they reduce quantization error and preserve grading headroom. Real benefit depends on sensor noise floor and exposure technique.
Compression type matters: lossy vs. lossless RAW (e.g., REDCODE variable compression vs. lossless CinemaDNG) — heavier compression can introduce coder artifacts impacting fine detail and chroma.
Chroma subsampling: many RAW formats store full per-pixel CFA data (effectively 4:4:4 after debayer), whereas typical log codecs use 4:2:2/4:2:0 — RAW gives better chroma fidelity post-demosaic.
Latency vs. quality: on-set proxies or camera-generated low-res proxies are essential for monitoring/fast turnarounds.
Rolling shutter and sensor behavior: RAW doesn’t change rolling shutter readout or sensor smear; it simply preserves the native readout characteristics.
When RAW is worth it — decision criteria
Project requirements
High-end grade projects (feature films, high-end commercials, HDR deliverables, heavy VFX integration, archival master): default to RAW when budget/turnaround allow.
Quick-turn work, web content, news, run-and-gun, or social-first deliverables: consider log/proxy workflows; RAW costs likely outweigh benefits.
Exposure latitude and lighting complexity
Scenes with extreme dynamic range, mixed lighting, or critical highlight retention benefit strongly from RAW.
Controlled, well-lit scenes with adequate exposure margin may not need RAW.
Post pipeline capacity
Do you have workstation/GPU horsepower, storage bandwidth, and colorist expertise? If not, RAW will bottleneck deliverables.
If using ACES or heavy VFX, RAW simplifies accurate color transforms and matte extraction.
VFX and image-processing needs
Keying, camera-tracking, plate matching, and heavy compositing are simpler when working from RAW due to canonical linear data and richer metadata.
Client and deliverable specs
If deliverables must be in high dynamic range (Dolby Vision, HDR10+) or archived as masters, RAW is preferable.
If clients request small turnaround, low-res deliverables, or streaming specs, a high-quality log codec may be adequate.
Practical workflow recommendations
On-set capture strategy
Record RAW if the shot’s intent or lighting stresses require it; otherwise capture log with a high-quality codec as an insurance policy.
Produce camera-side proxies (ProRes/MP4) for editorial and dailies; ensure timecode, slate, and metadata integrity for relinking.
Use redundant recording where critical: simultaneous RAW and proxy, or dual-card mirrored recording.
Storage and ingest
Plan storage: calculate data rates per camera and per day. Provision redundant RAID for active projects and LTO for long-term archive.
Use checksum-verified offload (e.g., LTO verification, verified copy tools). Embed generation metadata and consistent folder naming.
Post-production pipeline
Create offline proxies at ingest for editorial. Select proxy codec with accurate color and frame-accurate timecode (e.g., ProRes Proxy or 422 LT).
Maintain a color-managed pipeline: determine whether to use ACES, OCIO or camera-specific transforms. Establish IDT/ODT rules and LUTs.
Debayer and perform initial color transforms at conform/grade stage; apply dedicated denoising on linear data before heavy grading.
Performance optimizations
Use GPU-accelerated RAW decoders (BRAW SDK, RED SDK, ProRes RAW hardware) where supported.
Pre-render intermediate mezzanine files (e.g., high-quality ProRes HQ or DNxHR) if real-time playback is required and workstation resources are limited.
For multiple cameras with disparate RAW formats, normalize to a single working space early (e.g., linear ACEScg) to simplify cross-camera matches.
Metadata and QC
Capture ISO/gain, shutter, lens metadata, and color metadata; use it for IDT pre-sets to speed grading.
Implement a robust QC step post-ingest to catch dropped frames, color shifts, or codec decode errors early.
Examples of where RAW changed the outcome
Highlight recovery: recovering blown highlights on a backlit subject where log footage retained chroma clipping but RAW allowed rolloff reconstruction.
Heavy VFX: plate extraction where RAW linear data improved keying and matched grain/noise for compositing.
HDR grading: preserving specular highlights in HDR10+/Dolby Vision mastering where filmic rolloff from RAW prevented clipping.
Cost/benefit checklist for a shoot
Does the shot require more than ~4–5 stops of post-exposure manipulation? (Yes → RAW recommended)
Will VFX or keying require clean linear channel data? (Yes → RAW)
Can your on-set and post infra handle the data (cards, RAID, workstations, time)? (No → consider log codec + good proxies)
Does budget/time permit additional storage and grade time? (No → avoid RAW)
Quick comparisons (practical shorthand)
RAW = Maximum flexibility, maximum resource cost, best for HDR/VFX/archival.
High-bitrate Log (ProRes/DNxHR/ARRIRAW alternatives) = Good compromise for faster workflows; less headroom but easier editing.
Camera-log baked into efficient codecs = Fast delivery, limited grade latitude, lowest storage/time cost.
Final notes and best practices
Expose for SNR: RAW helps with highlight reconstruction, but preserve midtone SNR by exposing as brightly as practical (ETTR principles) while avoiding clipping.
Use per-scene decision-making: don’t default to RAW for everything; weigh mission-critical shots for RAW capture and use proxies or log for the rest.
Keep provenance and metadata: accurate IDTs, LUTs, and camera reports save grading time and mitigate color surprises.
Test your entire pipeline before production: capture camera tests, ingest, debayer, grade, and deliver a finished clip in the intended deliverable specs to validate performance and look.
Good night.