Quality, measured
Sharper text. Calmer zoom starts. Measured.
We exported the same three recordings from Dolly and from Screen Studio, at the same scale and the same moments, and measured the picture and the motion. Here is what we found, including where Screen Studio is just as good.
Measured 15 October 2026Dolly 0.4.12 · Screen Studio 4.0
Text sharpness
0.79–0.95Dolly
0.58–0.76Screen Studio
1.00 is a perfect resample
VMAF, after alignment
88–96Dolly
49–82Screen Studio
An independent check, 0 to 100
Zoom speed in frame 1
1–3%Dolly
27–30%Screen Studio
Lower is a softer start
In our test: three recordings, 15 October 2026. Ranges run from the lowest to the highest valid moment.
01
Sharper text, in all three recordings.
How much of the original detail each export keeps, at the same scale. Dolly kept 0.79–0.95, Screen Studio 0.58–0.76. The gap is largest on dense, light interfaces and small on the website.
Dolly · High 1080p60Screen Studio · its own export1.00 = a perfect resample
DashboardA web app: tables and settings, small text
DialogA web app: a modal over a list
WebsiteA white page with a dark chat panel
Sobel gradient energy of a text region, relative to a perfect Lanczos resample of the recording. Zoomed moments (1.1× to 1.6×): Dolly 0.92–0.93, Screen Studio 0.65–0.71. Read each recording on its own: a noisy source lowers both apps.
02
What that looks like, up close.
The same words, magnified, softened in proportion to the sharpness we measured on the dialog recording at rest.
An illustration drawn in code to show the measured difference, not a capture of either export. In the real crops, Dolly’s letters could not be told apart from the original.
03
Checked a second way, with VMAF.
VMAF is Netflix’s open measure of how close a video looks to the original. After the same alignment for both apps, Dolly scored higher on every valid moment. PSNR agrees: 41–48 dB against 25–32 dB.
DashboardA web app: tables and settings, small text
DialogA web app: a modal over a list
WebsiteA white page with a dark chat panel
Not the bitrate: a perfect resample encoded at Screen Studio’s own bitrate keeps 95–99% of its sharpness. The softening happens before the encoder.
04
Zooms that ease in, not jump.
The same 1.38× zoom at the same spot, tracked frame by frame. Screen Studio’s camera starts at speed; Dolly’s starts from rest and builds up. The pattern held on all six zoom-ins we compared.
Screen Studio: 30% of top speed in the first frameDolly: 3%, then a steady rise
Speed in the first frame
Dolly1–3%
Screen Studio27–30%
Share of top speed. A step at the start reads as a jolt.
Peak acceleration
Dolly0.4–3.1
Screen Studio4.9–31
Pixels per frame², zoom-ins of 1.38× and more. Lower is calmer.
Frames to reach speed
Dolly8–13
Screen Studio4–6
10% to 90% of top speed. Screen Studio gets going sooner.
Speed of a tracked block of the picture, as a share of that zoom’s top speed, so a faster and a slower zoom can be compared. The tiles cover six clean zoom-ins per app, 1.1× to 2.38×.
05
Link copies stay sharp too.
The 1080p web copy a Dolly link plays keeps almost all of the full export’s sharpness, at a fraction of the size.
DashboardLink copy 3.7 Mbit/s · Screen Studio’s export 1.9 Mbit/s
DialogLink copy 2.8 Mbit/s · Screen Studio’s export 1.5 Mbit/s
WebsiteLink copy 1.0 Mbit/s · Screen Studio’s export 1.3 Mbit/s
Same measure as above, 1080p30 web copy against High 1080p60. Bitrates from this test, which cuts every 3 seconds and so runs higher than a normal video.
06
Where Screen Studio holds its own.
Screen Studio is a good app. These are the places where our numbers don’t favour Dolly, or don’t say anything.
- Zoom-outs: no clear winner.Peak acceleration when zooming out was 2.4–4.3 px/frame² for Dolly and 0.9–5.8 for Screen Studio. Neither is calmer across the board.
- Screen Studio gets going sooner.Its zooms reach top speed in 4–6 frames, Dolly’s in 8–13. If you like a snappier zoom, that is a fair preference.
- Zero padding, off-centre zoom.With no padding and a zoom toward a point near the edge, Dolly’s camera picks up speed in a step half-way through (8.4 px/frame² tracked). With 8% padding the same zoom peaks at 3.1.
- Little fine detail, little difference.On the website recording, a low-bitrate source with large type, the gap was 5% (0.79–0.80 against 0.76), and by eye both look like the original.
- Not tested.A code editor and a dark-mode app. We had no Screen Studio export of either, so we don’t claim anything about them.
How we measured.
- Three recordings, three Screen Studio exports. We used the three Screen Studio recordings we had that already had an export, made in August and September 2026. No Screen Studio project was changed and nothing was re-exported for this test.
- The same pixels in. Each recording was opened in Dolly and exported as High 1080p60 and as the 1080p30 web copy a Dolly link plays, around the same moments, with a zoom of the same factor at the same spot where Screen Studio zoomed.
- The same scale. We measured where Screen Studio puts the recording in its frame (0.5089, 0.5100 and 0.6976 output pixels per recording pixel) and set Dolly's padding so it matches to 0.01%.
- The same moment. Screen Studio's cuts move its timeline, so each moment was found from the pixels. A moment only counts when the picture is still and the content matches: 11 of 23 Screen Studio moments qualified, and every excluded one is kept with its reason.
- Sharpness is the gradient energy (Sobel) of a text region, divided by that of the recording resampled perfectly (ffmpeg Lanczos) to the same grid. VMAF and PSNR are measured after the same alignment search for both apps (offset, scale and aspect), so a sub-pixel misalignment cannot penalise either one.
- Bitrate check. The perfect resample, encoded at each Screen Studio export's own bitrate, keeps 95 to 99% of its sharpness. Bitrate explains at most 5% of the difference.
- Motion. A block of the picture is tracked frame by frame through six zoom-ins per app; Dolly's camera was also computed from its code. Speed in pixels per frame, acceleration in pixels per frame².
- Not covered. A code editor and a dark-mode app: we had no Screen Studio export of either. Measured on a MacBook Air (Retina). All test exports were deleted afterwards.
Measured 15 October 2026. Dolly 0.4.12 development build (the resampler and camera code are unchanged in 0.4.14). Screen Studio exports made in August and September 2026 (Screen Studio 4.0.4 installed when we measured). These are the results of our test on our recordings; yours depend on your screen, your content and your settings.