HyperFrames vs Remotion: the same edit, two workflows
This page builds one sixty second composition in HyperFrames and in Remotion, then reports render time, cost and how many prompts each needed to reach a clean output. Existing comparisons describe the tools without measuring them. Here the project files and prompts are shared so you can rerun it yourself.
This page is built but kept out of search until it carries the evidence it needs: No measured results yet; the protocol is written and the page is labelled in testing.
Status: in testing
In testing. Nothing on this page has been measured yet, and it shows no result until a test has been run and reviewed.
Nothing has been measured yet. This page describes a test we plan to run, and it will stay without results until the test has been run and reviewed.
The question this page answers
If you build the same video job in code, how do HyperFrames and Remotion compare on render time, cost and the effort it takes to get a clean output? Nobody has published a measured comparison, so we plan to run one.
The protocol
The job is one composition built twice from the same written brief: a 30 second, 1080p, 30 fps vertical video with a title card, a lower third and word-highlight captions driven by a supplied transcript file. No footage is needed for the first version of the test.
- Write the brief and the acceptance criteria before building. The criteria: text never clipped, captions follow the supplied timings, no frame has an error, the automated flash check passes.
- Build it in each tool, using the same coding agent and the same starting prompt. Record every prompt and every manual fix.
- Render on one machine with the tool versions pinned and written down, with no other heavy job running.
- Render each build three times. Keep all three times, not the best one.
- A second person checks the output against the criteria and signs off or lists defects.
Anyone can rerun it: we will publish the brief, the transcript file, the prompts, the project files and the versions.
What we will measure
- Render time on the same machine, per run, and the spread across runs.
- Prompts and manual fixes needed to reach an output that meets the criteria.
- Output file size and whether the output passes the criteria and the flash check.
- Cost of the licence for the stated team size, taken from each tool's own terms on the date we check.
Scoring is pass or fail against the criteria first. Only passing builds are compared on time and effort. We do not combine the measures into one score.
What we need to run it
We can run this on our own laptop. What is missing is the run itself, and a second machine if we want to show how much the result depends on hardware.
Remotion's licence terms apply to the build. Individuals and teams of 3 or fewer people can use the free licence. Larger organisations need a company licence. Check the current terms on Remotion's licence FAQ, which we read on Nov 4, 2026.
Method and limits
No measurements are shown until the benchmark has been run.
Results will cover one composition on one machine. Render time changes with hardware and tool versions.
Check each tool's licence for your own use before building a product on it. Our method is on How we curate and test AI video workflows.
Both tools build the same composition from the same brief, driven by the same coding agent. We record tool versions, machine, render settings and every prompt, and we publish the files so the numbers can be reproduced.
- No measurements are shown until the benchmark has been run, so this page describes the method and the deliverables only.
- Results are for one composition on one machine, and render time changes with hardware and tool versions.
- Check each tool's licence for your own use before building a product on it.
Page checked on Nov 4, 2026.