I am writing to address a severe workflow issue regarding the GUI’s handling of large 4K / 60 fps batch exports, specifically when using the Aion model and exporting to ProRes Standard.
Project: 4K video, 60 fps, ~25 minutes duration (resulting in a massive ~350 GB file).
Settings: Max Processes set to 1. “Clean-up preview headers” disabled.
The Problem:
When rendering these heavy files, the Aion model processes everything to 100% perfectly. However, at the absolute final stage—when multiplexing and writing the final ProRes header—the GUI encounters a timeout or internal memory-shift error. It falsely flags the job as a “Recoverable Error” or just “View Export”, but fails to finalize the video container.
Even worse: Despite having “Clean-up preview headers” explicitly disabled in the preferences, the GUI’s batch-logic overrides this safety measure for the Aion model. Instead of leaving the unfinalized .part / _temp file in the destination folder, the GUI automatically deletes the 350 GB temp-file when switching to the next item in the batch list! This completely destroys hours of rendering time.
The Workaround (proving the file is fully processed):
If I do a single job and catch the error before the GUI cleans it up, I can copy the unfinalized temp-file. Passing this file through a third-party tool like Bandicut (Smart Rendering) to reconstruct the missing header results in a 100% flawless, fully playable ProRes video. This proves that the AI engine works perfectly, but the Topaz GUI / finalizer fails catastrophically on large file sizes and punishes the user by deleting the data.
Feature Request / Bug Fix Needed:
Fix the finalization timeout error for ProRes files exceeding 200–300 GB.
Ensure that “Clean-up preview headers: Disabled” is strictly respected across all models (including Aion) during batch processing, so files are never automatically deleted on error.
Professional users rendering 4K, 6K, or 8K VR master files overnight cannot rely on a batch system that randomly deletes terabytes of successfully processed data just because the GUI fails to close the file container.
I just ran a crucial test to see if this issue is strictly related to giant file sizes, and the results are revealing.
I processed a tiny 14-second test clip (4K ProRes Std) using the exact same pipeline: Aion model → ProRes Standard.
The result is identical: The processing finishes perfectly, but the GUI fails to finalize the file and refuses to write the header, leaving it as an unplayable temp-file.
This completely rules out any hardware-related system instability, thermal throttling, or drive-space/timeout issues with 350 GB files. The bug occurs regardless of file size. There is a fundamental disconnect between the Aion model’s processing loop and the ProRes file-container closure routine in the current GUI version.
Thanks for the detailed report and the additional testing!
The temp file deletion overriding your preferences is something the team is aware of and looking into.
That said, looking at your logs, the finalization error appears to be linked to an outdated GPU driver. Could you try updating to the latest NVIDIA driver for your RTX 5080 and testing again? That may resolve the container closure issue independently.
I agree.
Yeah, I get it—if I worked for a company, this would be exactly the right thing to do, and the boss would just have to foot the bill for those ridiculously expensive NVMe drives…
But I’m a home user, and if I were to buy NVMe drives like that for personal use—just to render 45 minutes of 4K at 60 fps as TIFF files—it wouldn’t be my wife talking to me anymore, but her lawyer.
Note: I have two 4GB - SRY 4 TB - Samsung SSDs connected via PCIe for Topaz and DaVinci ! ! … which I use in ping-pong mode (one reads, the other writes)…
Thanks for the update. I have completed the requested testing, but unfortunately, the issue remains completely unresolved.
Here are the exact details of my test environment and results:
GPU Driver: Updated to the absolute latest:
NVIDIA Studio Driver 610.62 (Released June 16, 2026). I performed a clean installation.
GPU: NVIDIA GeForce RTX 5080.
ProRes Flavors: The issue occurs on both ProRes Standard (STD) and ProRes LT.
Video Length: It does not matter at all. The container finalization fails equally on a short 5-second clip and a long 45-minute 4K video.
Crucial Update / Cross-Test & Performance Comparison:
I just cross-tested this with Topaz Video AI v7 (my purchased standalone version) using the exact same 4K source file.
In Video AI v7: ProRes finalizes perfectly without any issues!
But - However, the performance of AION in the old model is more limited to 2.4 fps.
In the latest version: The engine utilizes my RTX 5080 much better, reaching a fantastic 3.8 to 4.0 fps with AION on 4K. But at the absolute end, the finalization fails.
This clearly proves two things:
1. My GPU driver, system configuration, and hardware are fully capable of finishing the ProRes container (as proven by v7). 2. The bug is strictly isolated to the newer version’s ffmpeg/container pipeline. 3. I really need the current version to work, as the 60%+ performance boost (4.0 fps vs 2.4 fps) makes a massive difference for my workflow.
Right now, the app renders the entire video beautifully at top speed, but then forcefully deletes the temp files due to the known overriding bug, completely destroying the entire render job.
Please let me know what logs or diagnostic files you need next to escalate this to the development team so this pipeline bug can be fixed.
I don’t quite understand what the SSD has to do with this, and what ‘4GB’ means. Judging by the log, you have several terabytes of free space on your hard drives. TIFF and PNG formats export individual frames instead of a finished video file, which can then be converted to ProRes or anything else without any loss of quality.
Subject: Re: ProRes Finalization Failure & The TIFF Workaround Reality
Hi,
P.S.: 4 GB SSD: Sorry, I meant 4 TB NVMe, of course
To clear up the confusion regarding SSDs and the suggestion of using TIFF/PNG image sequences:
While exporting to an image sequence prevents the finalization bug from losing data, it is a completely unviable and unprofessional workaround for long-form 4K video editing. Let’s look at the math for my actual project, which is a 45-minute 4K video at 60 FPS:
1 Single Frame (4K, uncompressed 16-bit TIFF) = ~50 MB
At 60 FPS (45 minutes = 162,000 frames) = ~7.72 Terabytes (TB) of storage.
Suggesting that users manage a chaotic avalanche of 162,000 individual files and waste nearly 8 Terabytes of high-speed SSD space just to bypass a container closure bug is not realistic. It degrades a modern, high-end production workflow into an absolute storage nightmare.
As a professional software, Topaz Video AI must be expected to reliably handle industry-standard video containers like ProRes .mov natively without crashing at the last second—especially since Topaz Video AI v7 handled this exact same file flawlessly.
The issue here is not my hard drive space; it is a clear regression bug in the current version’s ffmpeg/container pipeline that needs to be escalated and fixed by the dev team.
I will second the comment related to manage hundreds of thousands of TIFF files. This requires a lot of storage space, the O/S sometimes struggles to display all files - I even avoid using the file manager as it generates thumbnails for each file, ouch!
My video sources are usually 360p-480p, so I am “only” dealing with hundreds of GB of data, but I feel for you with 4K16bits! Wondering if PNG would be an option vs. TIFF, would you lose in color space?
The bug must be fairly new, because for the past few months (not every day, of course—since I’m a personal user), I’ve been working exclusively with ProRes.
Sometimes with LT, sometimes with Standard.
(Depending on the file size)
And of course, it doesn’t just apply to AION—as I mistakenly assumed—but presumably to all models.
… I just tested a few of them …
… as I mentioned above, the bug isn’t present in Topaz Video AI 7
Seems like this error has been around since 1.2 as far as I can tell. I just experienced it on a M3 Max fully loaded, on Tahoe, exporting to internal. 5 second clip SLP 2.5 4k Prores > 8K.
Hi,
Okay, at least now I know I’m not the only one with this problem.
If, like me, you use Topaz to sharpen and spruce up old footage—usually with Iris and Aion—and then do the editing and final touches in DaVinci, ProRes is simply ideal.
Of course, you need massive NVMe drives to handle the resulting temporary monster files.
4K—after upscaling with Iris, and 60 fps with AION—exported in ProRes Standard: a nightmare.
( Exporting in a 16-bit image format would be the worst-case scenario!)
But it’s worth it. It’s worth it twice over:
Less loss due to permanent GOP and SLICES re-encoding, and
Faster workflow—DaVinci also loves intra-frame-based footage, especially when you’re building complex timelines…
I agree with everything you said about Prores. My workflow using it is very close to yours. One thing, Margaux helped me out effectively with my specific situation pointing to a memory issue she spotted in the logs. I was running GAIA to 8K from 4K and Hyperion at the same time, because of the suggestion to run an enhancement with Hyperion. She suggested to run them separately and that worked in terms of the completion without error of the export. Not quite satisfied with the Hyperion results but it was a first pass. I will try 1.7.0b today or tomorrow as I believe you suggested. So did Margaux btw. Thanks for your response! David
I’ve given up trying to run multiple models simultaneously on my mid-range PC (i9 11900; 4200 MT/s and RTX 5080). In 4K or more. That’s probably only possible with NASA machines
… I’d like to run IRIS and AION at the same time, forget it..
