Benchmarks · measured 2026-10-24

Up to 7.5× faster. No browser.

We rendered the same motion graphics with Volito and HyperFrames on one Mac: same encoder settings, both tools on their defaults, medians of three runs. Here is every number, how we got it, and where HyperFrames is still ahead.

1.9–7.5×: faster renders, across 11 render tests
9–14×: faster agent loop: 0.15–0.19 s a call, not 1.5–2.2 s
51 MB: installed, against 304 MB with Chrome

Render time, in seconds.

Shorter is faster. Short 1080p graphics, 3 to 5 seconds long, rendered to video or frames. The gap is widest on short renders, where HyperFrames spends about 3 s starting Chrome and loading the page.

Volito 0.1.0
HyperFrames 0.8.188

Lower third 1080p · 4.8 s

H.264 MP4

1.02 s7.60 s

7.5×

ProRes 4444 with alpha

2.80 s8.84 s

3.2×

ProRes 4444, every frame drawn

3.56 s8.84 s

2.5×

PNG frames

1.69 s6.63 s

3.9×

Kinetic title 1080p · 5 s

H.264 MP4

2.01 s5.80 s

2.9×

PNG frames

3.61 s11.49 s

3.2×

Moving squares 1080p · 3 s

10 squares

0.70 s3.92 s

5.6×

100 squares

1.53 s4.61 s

3.0×

1,000 squares

2.97 s6.27 s

2.1×

Show the numbers as a table

Test	Volito	HyperFrames	Faster
Lower third (1080p · 4.8 s): H.264 MP4	1.02 s	7.60 s	7.5×
Lower third (1080p · 4.8 s): ProRes 4444 with alpha	2.80 s	8.84 s	3.2×
Lower third (1080p · 4.8 s): ProRes 4444, every frame drawn	3.56 s	8.84 s	2.5×
Lower third (1080p · 4.8 s): PNG frames	1.69 s	6.63 s	3.9×
Kinetic title (1080p · 5 s): H.264 MP4	2.01 s	5.80 s	2.9×
Kinetic title (1080p · 5 s): PNG frames	3.61 s	11.49 s	3.2×
Moving squares (1080p · 3 s): 10 squares	0.70 s	3.92 s	5.6×
Moving squares (1080p · 3 s): 100 squares	1.53 s	4.61 s	3.0×
Moving squares (1080p · 3 s): 1,000 squares	2.97 s	6.27 s	2.1×

Long renders

Thirty seconds of moving squares, 900 frames. Chrome’s start-up matters less over a long render, so the gap settles near 2×.

Volito 0.1.0
HyperFrames 0.8.188

Moving squares, long 1080p · 30 s

100 squares

10.02 s23.42 s

2.3×

1,000 squares

20.79 s38.90 s

1.9×

Show the numbers as a table

Test	Volito	HyperFrames	Faster
Moving squares, long (1080p · 30 s): 100 squares	10.02 s	23.42 s	2.3×
Moving squares, long (1080p · 30 s): 1,000 squares	20.79 s	38.90 s	1.9×

The agent loop, 9–14× faster.

An agent writing a graphic looks at it, checks it, fixes it and looks again, many times. Each of those calls comes back from Volito in under a fifth of a second.

Volito 0.1.0
HyperFrames 0.8.188

Agent loop on the lower third

One still frame

0.19 s2.23 s

11.7×

Four-frame preview

0.15 s2.02 s

13.5×

Check

0.16 s1.52 s

9.5×

Show the numbers as a table

Test	Volito	HyperFrames	Faster
Agent loop (on the lower third): One still frame	0.19 s	2.23 s	11.7×
Agent loop (on the lower third): Four-frame preview	0.15 s	2.02 s	13.5×
Agent loop (on the lower third): Check	0.16 s	1.52 s	9.5×

HyperFrames does more per call: its snapshot writes full-size frames plus a contact sheet, and its check also audits layout and contrast. Re-checked from a fresh install with all 96 items, Volito’s still, preview and check took 0.14, 0.20 and 0.16 s.

Why it’s faster: no browser.

HyperFrames: HTML in Chrome

Start Chrome about 3 s
Load the page
Screenshot every frame
FFmpeg

Volito: a program per graphic

frame(t, $) returns SVG
resvg, ten frames at a time
FFmpeg

Nothing to boot or load. Node runs the program and resvg rasterises the SVG in the same process. HyperFrames’ own log puts its start at 0.8 s to compile and 2.5 s to set up, before the first frame.

Frames that don’t change are drawn once. A frame is a pure function of time, so a frame that comes out identical to the last is reused. The chart shows the every-frame render too.

Less CPU and memory. Volito used less CPU time in every test, from about the same to 9× less, and less peak memory in all but one (10 squares).

Safer agent code. With no browser, code an agent writes runs in a sandbox with no network, require or eval.

Install size

Volito 0.1.0
HyperFrames packages
Chrome headless shell

Volito

51 MB

HyperFrames

111 MB193 MB Chrome304 MB

Where each one wins.

HyperFrames is a whole video platform; Volito is a graphics engine. Speed is not the only thing that matters, and HyperFrames is ahead on breadth.

Pick Volito for

1.9–7.5× faster renders and a 9–14× faster agent loop in these tests
36 brand styles or your own DESIGN.md, and every aspect from 16:9 to 9:16 from one item
ProRes 4444 and HEVC with real alpha, straight onto footage
Text that lands on the words of a transcript
Agent-written code runs in a sandbox with no network
51 MB installed, no browser, no GPU

Fast, branded overlays, and an agent that iterates in a fifth of a second.

Pick HyperFrames for

448 ready-made items (210 blocks and 238 components) to Volito’s 96
Whole videos: footage, an audio mix, captions and colour grading
WebGL 3D with models and textures, shader effects, Lottie, and real sub-frame motion blur
Studio, a visual editor in the browser and on the desktop
Cloud rendering on HeyGen, AWS Lambda or Google Cloud Run
70k GitHub stars, 540 releases and production users

Whole videos, and anything that needs a browser’s full toolbox.

The two can combine: HyperFrames can drive any pure function of time through registerFrameSource, so a Volito graphic could play inside a HyperFrames project. We have not tried it yet.

How we measured

Setup

One Apple M4 Mac (10 cores, 16 GB, macOS 26.6.2), Node 22.19, FFmpeg 8.0.1.
Volito 0.1.0, measured under its earlier name MONARC Motion, against HyperFrames 0.8.188 with Chrome headless shell 152.
Both tools on their defaults: HyperFrames chose its own worker count (4 for the lower third); Volito draws 10 frames at a time.
The same encoder settings: x264 preset medium at CRF 16 for MP4, prores_ks 4444 for MOV.
29 jobs run in turn, alternating the two tools: one warm-up round, then three measured rounds (two for the 30-second jobs). Medians reported; wall time is the whole command.
Other apps were open, so absolute times are not lab-clean; both tools ran under the same load.

Scenes

Lower third: Each tool’s own ready-made lower third (Volito’s lower-third against the HyperFrames catalog block lt-dark-card), 1920×1080, 30 fps, 4.8 s.
Kinetic title: One title built the same way in both: same Inter files, sizes, timings and eases, with a drifting glow that changes every frame. 5 s.
Moving squares: 10, 100 or 1,000 squares on identical paths for 3 s, and 100 or 1,000 for 30 s (900 frames).
Agent loop: What an AI agent runs on the lower third: render one still, a four-frame preview, and a check.

Fairness notes

Volito draws a frame that comes out identical to the one before only once, which helps it on the lower third. The every-frame row turns that off: 3.56 s against 8.84 s.
The two lower thirds differ: the HyperFrames block is simpler (no full-frame scrim). The kinetic title and the squares are matched.
HyperFrames’ snapshot writes full-size frames plus a contact sheet, and its check also audits layout and contrast, so its agent calls do more work per call.
On macOS HyperFrames captures frames by screenshot; its docs say the BeginFrame capture on Linux is faster. These numbers are from macOS only.
Both tools gave identical frames on a second render, and byte-identical ProRes files.

HyperFrames ships often and has not been re-run since HyperFrames 0.8.188. These numbers are from one Mac on 2026-10-24; if you measure something different, especially on Linux, tell us and we will update this page.

$ npm install volito

Read the docs Use it from an agent