As a videographer who has shot extensively on Sony mirrorless cameras since the first A7S, selecting the right codec for your production is one of the most pivotal (and sometimes bewildering) decisions you have to make…
I still remember years ago spending weeks shooting several short films on my original A7S, recording over 96 hours of XAVC-S footage to external SSD drives. When I finally loaded all those files into Premiere Pro to start editing, my excitement rapidly turned into a nightmare – even on a high-end PC, the playback was unbearably choppy with constant dropped frames. Caching helped temporarily, but exports were still interminably long. I had to downgrade the timeline to 720p proxies just to get through basic editing tasks without everything grinding to a halt.
So when Sony announced their new H.265 based XAVC H-S codec in 2020 promising smaller file sizes with better quality, I was understandably skeptical. But out of curiosity I tested it in the field on a documentary shoot, recording identical B-roll clips in both XAVC-S and XAVC H-S…and I was blown away. Not only were the H-S files sometimes nearly 1/3 smaller, but I was able to edit and export them far faster than the equivalent H.264 footage (about 2x in my benchmarks). I could punch in aggressively in post without any stuttering or lag even on 4K timelines. And subjectively I could barely discern any difference between the straight out of camera image quality.
Clearly there had been major advancements under the hood. This one test completely changed how I looked at codecs and compelled me to better understand what was actually happening inside all that video data to have such an impact on my real-world workflow.
Fast forward to having the A7S iii in my hands, and Sony now offers three different sophisticated codecs to choose between: XAVC-S, XAVC H-S and XAVC S-I. Based on hard lessons learned, I‘m no longer intimidated by endless acronyms. Instead I‘m focused on asking what each codec can (or can‘t) deliver for any given production based on how they handle compression. This guide breaks down exactly what you really need to know about how each of these codecs across dimensions like color fidelity, artifact susceptibility, editing/rendering performance and output quality.
Long GOP vs All-Intra Compression
While the specifics get technical (especially when arguing with other camera nerds in forums), at a high level Sony groups the A7S III codecs into two families:
Long GOP codecs: XAVC-S and XAVC H-S
All-Intra codec: XAVC S-I
The core difference comes down to how they handle compression between frames:
Long GOP Compression
"GOP" stands for Group of Pictures and refers to analyzing pixel data across a sequence of frames to identify areas that DON‘T change dramatically. And then for those more static areas, only partial data is stored. By not capturing redundant pixels that stay unchanged frame to frame, massive reductions in file size are possible.
Here‘s a graph showing an example GOP structure:
With Long GOP compression, instead of storing full data for every single frame, you have:
- I-frames: Complete images stored independently without referencing other frames
- P-frames: Only partial data stored based on changes from preceding I-frames or P-frames in the GOP sequence
- B-frames: Based on changes from the nearest I & P frames (both forward & backward) for maximum compression
In the example above, a full image is stored for the initial I-frame, but after that only selective pixel change data is saved for the P & B frames that follow. This results is massively smaller overall files sizes, but does introduce risks we‘ll cover shortly.
First though, within the long GOP family the XAVC-S and XAVC H-S codecs have some notable differences in their GOP structures worth calling out…
XAVC-S GOP Structure
The XAVC-S codec uses H.264 compression with typical Long GOP parameters of:
- I-frame frequency: ~1 second
- I to P ratio: 1:4 to 1:6
- GOP size: 8-12 frames
- 1 I-frame
- 4 to 6 P-frames
- 0 to 4 B-frames
So in general you‘ll have a complete I-frame followed by 4-6 P-frames pulling selective data from that initial I-frame (and each other). Finally 0-4 B-frames may reference the enclosing I & P frames backwards and forwards in time to maximize inter-frame data sharing.
This IPP structure repeats roughly each second as shown above, though the exact timing of I-frames, number of P & B frames in between, and the sequencing will vary clip by clip.
XAVC H-S GOP Structure
The XAVC H-S codec utilizes the more advanced H.265 standard, which is designed to provide better compression efficiency than H.264. It accomplishes this in part by supporting much larger GOP sizes vs XAVC-S, with more P & B-frame references between each I-frame.
Typical XAVC H-S GOP structuring is along the lines of:
- I-frame frequency: 1 every 2-3 seconds
- GOP size: 16-48 frames
- 1 I-frame
- Often 14+ P-frames
- 1+ B-frames
So in general you should expect to see about half (or fewer) I-frames, longer chains of inter-connected P-frames, and strategic B-frames peppered throughout.
This more complex GOP pattern is what allows H.265 to achieve its claimed 2x compression ability over H.264 with comparable image quality. But as we‘ll discuss shortly, this added compression power comes with some notable workflow tradeoffs.
All-Intra Compression
The XAVC S-I codec forgoes Long GOP compression entirely. Instead every single frame is encoded as an I-frame and stored independently. This completely avoids any inter-frame pixel analysis and predictions over time. The benefit is that it provides the cleanest, artifact-free preservation of image data possible.
But without any data sharing from one frame to the next, the storage price tag is obviously much higher. Just know that if quality is your primary consideration, all-intra is the way to go.
Now that we‘ve covered the core differences in how Long GOP and All-Intra handle compressing video data across frames, how does this actually impact you in real-world shooting and post-production scenarios?
The Pros and Cons of Long GOP Compression
By only storing selective pixel change data over time, Long GOP compression used in the XAVC-S and XAVC H-S codecs can enable dramatically smaller file sizes compared to all-intra. But this comes with some notable drawbacks:
Potential Quality Loss from Compression Artifacts
By relying on pixel analysis between frames, degraded image quality or visual artifacts can creep into your footage with Long GOP under certain conditions:
- Complex motion and textures
- Significant lighting changes
- Exposure shifts
- Random movement like foliage
When adjacent frames have lots of variation, the encoding process has to work much harder to evaluate motion vectors, occlusion data, deformations etc. By tossing data away aggressively, video codecs can miscalculate the ideal inter-frame pixel mappings, resulting in:
- Blockiness
- Color banding
- Aliasing
- Noise
These compression artifacts will be most noticeable in high motion shots or transitions from dark to bright environments. Scenes with fine repetitive textures like brickwork or foliage can also confuse encoders.
In general XAVC H-S renders better footage than XAVC-S in challenging conditions due to its more robust compression algorithms. But neither can fully protect your clips from encoding deficiencies like All-Intra can.
Restricted Editing & Playback Performance
During editing in programs like Premiere Pro or Final Cut Pro, playing back Long GOP media requires analysing pixel data across multiple frames. This makes real-time playback and scrubbing more demanding compared to independent I-frame only video like all-Intra provides.
And during export, Long GOP codecs must reconstruct the inter-frame processing that is normally handled in camera during recording. This after-the-fact encoding boosts export times, consuming extra computing horsepower.
With XAVC H-S it tends to be 2-3x slower vs XAVC-S for editing because of the expanded GOP sizes and more rigorous compression logic applied. So expect even choppier playback and longer export times if your editing workstation isn‘t a powerhouse.
Benefits of All-Intra Video Quality
By storing pristine data for each frame individually, All-Intra XAVC S-I codec used in the A7S iii brings some compelling benefits:
No Compression Artifacts
Without trying to selectively analyze pixels across frames, All-Intra side steps the risk of encoding errors producing visible artifacts in your footage during shooting or post-production. This makes it a favorite codec for green screen work where even minor banding or blocking can hamper compositing and keying results.
And because full image data is preserved losslessly scene to scene, creative latitude in post-production is maximized for color correction, sharpening adjustments and any graphics overlays or effects you want to composite later.
Faster Editing & Rendering
When editing all-intra footage, playback is much smoother with no dropped frames on moderately capable systems. Scrubbing through clips is far more responsive as well. This lets you rapidly iterate without performance bottlenecks getting in the way of your creative flow state.
And with no inter-frame data to rebuild on export, final rendering times are 2-3x faster than Long GOP footage, especially for longer timelines. So you spend less time staring at progress bars and more time crafting exceptional productions.
Of course this high-performance creative flexibility comes at the cost of radically larger storage requirements…
Comparing File Sizes Across the Sony Codecs
To appreciate just how much extra data all-intra recording demands, here‘s a breakdown of typical bitrates used for 4K shooting on the A7S iii across the different codecs (officially specified by Sony here):
| Codec | Bitrate |
|---|---|
| XAVC-S | 100 Mbps |
| XAVC H-S | 140 Mbps |
| XAVC S-I | 240 Mbps |
You can immediately see that XAVC S-I more than doubles the data rate vs XAVC-S. And 60% higher demands versus XAVC H-S as well.
To put that in more practical shooting terms, here is how much 4K recording time you can expect on a 256GB CFe Type A memory card using each codec:
| Codec | Record Duration |
|---|---|
| XAVC-S 100Mbps | Roughly 5 hours |
| XAVC H-S 140Mbps | Over 3 hours |
| XAVC S-I 240Mbps | 1 hour 30 minutes |
Clearly if you‘ll be shooting longer pieces, all-intra can fast consume storage space. So scaling a documentary means needing beefier MP4 recorders, more memory cards in rotation, and possibly backup drives…definitely more dongles!
Real-Worldcodec Comparisons from an Active Videographer
Okay, by this point your head is probably spinning trying to parse technical compression algorithms. Let‘s boil it down to straightforward recommendations tailored to different production needs from my years of experience relying on these cameras to pay the bills:
When XAVC-S is "Good Enough" Even for Commercial Work
Despite testing literally every codec possible on every camera I work with, I still default to XAVC-S for certain types of budget commercial shoots. This includes:
- Corporate interviews (usually single cam)
- "Broll" at internal conferences or events
- Run-and-gun weddings
Basically scenarios where client expectations are focused more on technically sound footage as opposed to cinematic perfection.
Even though H.S may yield marginally better quality, or S-I would provide pristine footage…the storage costs, recording durations and editing flexibility of XAVC-S still make it a great workhorse codec for prosumer level DSLR/Mirrorless work.
And honestly most viewers probably can‘t discern the difference between XAVC-S quality vs the data hungry S-I version! As long as you nail focus and exposure, frame intelligently and light effectively, XAVC-S holds up remarkably well for times when maximum production value isn‘t the top priority.
When XAVC H-S Saves the Day on Documentary Shoots
After that initial documentary test shoot converted me to the power of H.265, XAVC H-S has become my default codec anytime I need to maximize recording time or require higher image quality for budget-conscious projects:
- Multi-episode docuseries
- Internal video magazines for tech companies
- Event live streaming
The ability to fit 2-4 hours of footage onto 256GB cards compared to under 90 minutes for XAVC S-I is game changing, especially when backing up onto SSDs.
And knowing I can push colors more aggressively in post or handle VFX overlays if needed provides welcome creative headroom over XAVC-S. All while retaining snappy playback on my (aging) editing laptop.
So don‘t underestimate how versatile XAVC H-S can be despite seeming like the middle child of the Sony codecs!
When Only XAVC S-I Will Cut It for Commercial Finishing
Now if bleeding edge image quality is an absolute requirement with no compromises, XAVC S-I remains the gold standard for professional production deliverables:
- TV commercials
- Branded content
- Music videos
- Short films
Recording all-intra guaranteeing pristine footage should be considered mandatory for these types of projects where quality is the expectation.
In an age of 4K (or even 8K!) consumer displays with HDR, clients are scrutinizing video closer than ever on bigger, more vivid screens. This means less forgiveness for any subtle banding, noise or blocking that creeps in during shooting or post.
Yes storage budgets inflate drastically, but this Clause in my production contracts explicitly permits charging for all-intra media costs. Because honestly the reliability and creative insurance for high stakes finishing is absolutely worth the surcharge!
Bottom Line Recommendations by Production
Hopefully reviewing real-world examples gives better context on when leaning towards smaller file sizes makes sense vs times when uncompromised quality should be the priority.
Here is a quick summary of how I would recommend each Sony A7S iii codec based on production:
- Corporate & Event Videos
- File Size Priority: XAVC-S
- Quality Priority: XAVC H-S
- Documentaries & Web Series
- Storage Flexibility: XAVC H-S
- Maximum Quality: XAVC S-I
- Commercial Finishing & Music Videos
- Only Option: XAVC S-I
Of course exact choice always depends on the client goals, viewing formats, post-production plans and your editing hardware capability. But these guidelines offer a starting point for factoring the key trade-offs around efficiency, flexibility and no-compromise image quality.
I hope breaking down both the technical inner workings and real-world performance of the Sony codecs removes any trepidation in selecting the optimal one each production. Feel free to reach out with any other questions!