A 30-second animation is a much bigger render job than it sounds, and that catches almost everyone off guard at least once. At 30 frames per second,
A 30-second animation is a much bigger render job than it sounds, and that catches almost everyone off guard at least once. At 30 frames per second, a 30-second clip is 900 separate frames. If each frame takes 5 minutes, your machine is busy for roughly 75 hours with no breaks. The way out is not usually a faster computer. It is understanding that render time tracks frame count multiplied by per-frame complexity, then cutting per-frame time with sampling, denoising, and texture cleanup. After reworking my own 30-second shot, per-frame time dropped from about 5 minutes to roughly 90 seconds, and renting a few extra GPUs for the final pass took the wall-clock time from days down to a single afternoon.
Years ago I took on a 30-second animated explainer for a small product company. Stylized characters, a bit of motion graphics, nothing exotic. I quoted the job thinking the render would be a one-night thing. I set it going on a Friday evening feeling pretty relaxed about it. By Sunday afternoon it was still grinding, and it had stalled twice along the way because a couple of frames choked and I had to restart the queue. Three days for thirty seconds. I sat there genuinely confused, because the clip felt small.
It was not small. I just did not understand what I was actually asking my computer to do.
Why thirty seconds turns into three days
The thing nobody tells you starting out is that runtime and render time have almost nothing to do with each other. A clip is not one thing your computer renders. It is hundreds of still images rendered one after another, and then played back fast enough to look like motion. So the real question is never “how long is the animation,” it is “how many frames, and how heavy is each one.”
Here is the same clip length at different per-frame times, so you can see how fast it stacks up:
| Clip length (at 30 fps) | Total frames | 2 min / frame | 5 min / frame | 10 min / frame |
|---|---|---|---|---|
| 10 seconds | 300 | 10 hours | 25 hours | 50 hours |
| 30 seconds | 900 | 30 hours | 75 hours | 150 hours |
| 60 seconds | 1,800 | 60 hours | 150 hours | 300 hours |
Illustrative. Per-frame time depends entirely on your scene, engine, and settings. Run your own test render and multiply.
My explainer was sitting around 5 minutes a frame. Nine hundred frames at 5 minutes is 75 hours of pure render, and that is before the two stalls and restarts that pushed it across a full weekend. The clip being short did nothing for me. Each of those 900 frames was carrying the same bloat.
What I changed to get the per-frame time down
Before touching any hardware, I went looking for where those 5 minutes were going. The render log told the story fast. A big chunk was sampling set far higher than the shot needed, and a smaller chunk was motion blur on background elements that were soft and out of focus anyway.
The changes that moved the needle, roughly in order of how much they gave me back:
- Adaptive sampling with a denoiser. I had been brute-forcing samples out of habit. Dropping to a sane sample count and letting OptiX clean up the rest pulled per-frame time down by close to a third on its own, with no visible quality loss once the sequence played back.
- Texture sizes that matched the screen. I had 8K maps on objects that filled maybe a quarter of the frame. Knocking the ones far from camera down to 2K, and the hero objects to 4K, took another slice off and made the scene load faster too.
- Motion blur and bounces where they actually mattered. I cut blur steps on the background and trimmed the global illumination bounces from a number I had never questioned down to something reasonable. The image looked the same.
Per-frame time landed around 90 seconds. Nine hundred frames at 90 seconds is about 22 to 23 hours on my own machine. Better than 75, but still an overnight-plus job, and I had a client waiting.
Getting from a full day down to an afternoon
With the scene already lean, the only thing left between me and the deadline was raw compute. Buying a stronger machine for one job made no sense, so I rendered the final pass in the cloud and split it across several GPUs at once.
This is the part where people ask which render farm to use, and the answer changes depending on what you care about. Since this site does not belong to any one farm, here is how I actually see the main options for a job like this.
- GarageFarm is the one I send beginners to. You drag your file in, it renders, you download the result. If the weekend I lost taught you that you never want to babysit a render again, this is the gentlest landing. The catch is that you render inside their system, so if your project uses an unusual plugin or a specific software version, you can hit walls.
- RebusFarm has the best pre-render scene checker I have used. For a frame that keeps failing, it often catches the missing texture or broken path before you waste render time on it. It leans strong on CPU rendering. The interface feels a little old, and it is not the cheapest.
- Fox Renderfarm tends to come in lowest on price for big batches of frames, which is exactly what a 900-frame sequence is. Worth a quote if cost is your main worry. Support and queue times at peak can be hit or miss.
- iRender works differently from all three. Instead of handing your file to their system, you rent a full remote machine with up to 8x RTX 4090 and 256GB of RAM, and you set it up like it is your own workstation. You install your exact Redshift or Cycles version, your plugins, your scene the way you know works. Their current line is “your renders, your rules,” and for a job where my setup was a bit particular, that control is the whole reason I reached for it.
A couple of things about iRender you should know before you assume it is the easy button. It is a remote machine, so the first time you connect you install your software and move your project over, which ate maybe 35 minutes on my first run. And the clock runs the entire time the server is powered on, not only while it renders, so leaving it on after the job finishes quietly burns money. I shut mine down the second the last frame wrote out.
I rented a multi-GPU server and the final pass that would have taken most of a day on my own card finished in a few hours, including the setup and upload time. Thirty seconds, delivered the same afternoon I started the cloud render. The thing I want you to take from this is the order: I optimized first, so I was paying to render a clean scene instead of paying to render my own waste at high speed.
On cost: iRender gives new accounts a 100% bonus on the first deposit, and its Credit Back program returns roughly 10% to 20% of what you spend as credits depending on timing (10% standard, 12% on weekday Happy Hours, 20% on weekend Golden Hours, GMT+7). For a one-off batch like a 900-frame sequence, lining the render up for a weekend and stacking the first-deposit bonus can bring the effective rate down a long way. (Check current rates before relying on these numbers.)
FAQ
How long does it take to render a 30-second animation?
The clip length barely matters here; per-frame time decides it. A 30-second clip at 30 fps is 900 frames, so at 2 minutes per frame that is about 30 hours, at 5 minutes per frame about 75 hours, and at 10 minutes per frame around 150 hours. Lowering per-frame time through sampling and texture cleanup is what changes the total, not the fact that the clip is short.
Why is my short animation taking days to render?
Because runtime and render time are unrelated. A short clip still contains hundreds of individual frames, and if each frame carries oversized textures, too many samples, or heavy motion blur, that cost multiplies across every frame. Read your render log to find the slow stage, fix it once, and the saving applies to the whole sequence.
Is it cheaper to render a short animation on a render farm?
For a one-off heavy sequence, usually yes, because you avoid buying hardware you will rarely use. Optimize the scene first so you are not paying to render waste. Then a SaaS farm like Fox Renderfarm suits cheap batch work, while an IaaS service like iRender suits projects that need a specific software setup or multi-GPU speed.
Your renders, your rules
When a clean sequence still needs to be done by this afternoon, I render the final pass on a full GPU server I control. New accounts get a 100% first-deposit bonus, and weekend renders earn up to 20% back through Credit Back. See iRender GPU servers and pricing → iRender RTX 4090 servers. Shut the server down the moment the last frame writes.
See more: A Single Frame Takes 25 Minutes. Is That Normal for My Scene?
Image source: BlenderNation

COMMENTS