Architecture and specs, no unmeasured claims

FlashFX vs CapCut vs DaVinci Resolve

Three tools that are genuinely different products, not three versions of the same one. This page compares how they are built, what hardware each demands, and which features exist in which, not which is fastest.

We have not published performance figures here, and the reason is on this page.


How each one renders

Most of the differences below follow from these three pipelines. Where a tool is strong or weak is usually predictable from its architecture rather than from how hard anyone worked on it.

FlashFX

Browser, client-side

  1. File.slice byte source
  2. Streaming demuxer
  3. WebCodecs decode, one worker per asset
  4. WebGPU texture upload
  5. WebGPU render graph
  6. WebCodecs encode and mux
  • Compute runs at native speed once frames are on the GPU: WebGPU compute shaders are not emulated.
  • Codec work goes through the browser media stack, which does reach hardware encoders and decoders.
  • Format support is whatever the browser supports, not what we support. That is the hard ceiling.
  • Memory is bounded by tab and GPU process limits rather than by installed RAM.

CapCut

Native desktop, plus mobile and web

  1. Direct OS media framework access
  2. Media Foundation, VideoToolbox, NVENC, QuickSync
  3. Fixed-function effect stack
  4. Template-oriented composition
  • No sandbox between the app and the hardware encoder.
  • Effects are a fixed list rather than a general compositing graph.
  • Capability differs across desktop, mobile and web. They are not the same product.

DaVinci Resolve

Native desktop, professional

  1. Full native GPU pipeline
  2. CUDA, Metal or OpenCL
  3. 32-bit float internal processing
  4. Fusion node compositor, Fairlight DAW
  • Hardware decode for professional and camera RAW formats.
  • Node-based compositing rather than a layer-and-timeline model.
  • A genuine colour pipeline built by a company that also makes cameras.

What each one demands of your machine

These are published requirements, not our measurements. It is the largest and least arguable difference between the three.

FlashFX

A current browser

Runs in a tab. No installer, no driver requirement, no dedicated GPU. Works on integrated graphics, Chromebooks and locked-down machines where the other two cannot be installed at all.

Architecture

CapCut

8 GB RAM

8 GB is adequate for typical short-form editing; 16 GB is recommended for 4K source or effects-heavy multi-layer projects. Modest by design. It targets fast accessible editing rather than deep grading or compositing.

Published system requirements

DaVinci Resolve 20

16 GB RAM, 4 GB VRAM

Windows minimum is 16 GB system memory, rising to 32 GB when Fusion is in use, with a GPU of at least 4 GB VRAM and OpenCL 1.2 or CUDA 12.8. Practical 4K work sits far above the minimum: roughly 24 GB VRAM, 64 GB RAM and a dedicated NVMe cache.

Blackmagic Design published requirements


What exists in which

Presence or absence of a feature, not how fast it runs. The empty cells are the point, and all three columns have them.

Feature availability across FlashFX, CapCut and DaVinci Resolve
CapabilityFlashFXCapCutResolve
Design & vector
Pen tool with bezier pathsSmoothing, per-point corner radius, dash arrays, trim start and endYesNoYes
Animatable corner radiusPixel-level, animatable for morphing transitionsYesNoYes
Auto-layout containersHorizontal and vertical, with padding, margin and nestingYesNoNo
Reusable UI component shapesButton, input, toggle, modal, chat bubble, progress barYesNoNo
SVG import with inline editingFill and stroke override, ViewBox preservationYesNoYes
Built-in searchable icon libraryYesYesNo
Smart guides and snappingSnap to grid, elements, markers and keyframesYesYesYes
Blend modes16 element-level modesYesNoYes
Animation
Keyframe graph with bezier velocity curves16 easing functions with per-keyframe bezier handlesYesNoYes
ExpressionsOwner-confirmed 2026-08-06YesNoYes
Repeaters and clonersOwner-confirmed 2026-08-06YesNoYes
Per-character text animationBy character, word, line or objectYesNoYes
Compositing
Masking with feather and invertYesNoYes
Groups, nesting and shared parent transformsEnter and exit group editing, lock and hide per layerYesNoYes
Nested compositionsSequence compositorYesNoYes
Chroma keyYesYesYes
3D with model importGLB, OBJ, FBX, STLYesNoYes
Collaboration
Real-time collaborationFlashFX: Teams planYesNoYes
Version history30 days free, 90 days on paid plansYesNoYes
Media & output
Transparent exportPNG sequenceYesNoYes
Auto-captions and stock libraryNoYesNo
Camera RAW and professional codecsBRAW, R3D, ProRes, DNxHRNoNoYes
ACES and HDR colour grading32-bit float, DaVinci Wide GamutNoNoYes
Full digital audio workstationFairlightNoNoYes
Platform
Runs with no installationYesNoNo
Mobile appNoYesYes
Works offlineNoYesYes

Which one to reach for

These are three different jobs. Picking by benchmark score would be the wrong way to choose between them.

FlashFX

  • Motion graphics: kinetic type, logo animation, social motion design
  • You need it to run on a machine that cannot install the alternatives
  • You want to open a URL rather than manage media, relink files and warm a cache
  • The work is synthetic: shapes, text, gradients and particles rather than hours of footage

CapCut

  • Cut, caption, publish: short-form social video on a deadline
  • You are editing on a phone
  • You want templates, auto-captions and one-click publishing
  • You need to work offline, or on a long 4K timeline

DaVinci Resolve

  • Colour grading is the point: ACES, HDR, scopes, control panels
  • You are cutting camera RAW, ProRes or DNxHR
  • Long-form: features, documentaries, multicam, hours-long timelines
  • You need a real DAW in the same application

Why there are no speed numbers here

We have an internal benchmark framework for exactly this comparison. It predicts how FlashFX should perform against both tools across export throughput, playback frame rate, seek latency and time-to-result.

Those are predictions derived from architecture. They have not been measured. So they are not on this page, and they will not be until they have been run properly : three passes on one machine, median reported, cold GPU, thermals settled, with the method published next to the result.

Benchmarks written by the vendor of one of the products are worth exactly as much as their method, and FlashFX being ours makes that standard higher rather than lower. Everything above is either how the software is built or a specification its publisher has stated, both of which you can check without taking our word for it.

When the numbers exist, they will appear here, including the ones we lose. On raw codec throughput: long timelines of straight cuts, especially at 4K: a native application with direct hardware encoder access should beat a browser tab, and we expect it to.


Frequently Asked Questions

Also worth reading: FlashFX vs After Effects, the comparison closest to what FlashFX is actually for.


The only benchmark that matters is your own

FlashFX opens in a tab, so trying it costs a click rather than a download and an install. Build something and see whether it fits how you work.