Feature matrices are the least useful format in tooling advice. Every tool claims the same rows, the ticks are self-reported, and the answer depends on a context the table cannot see.
So here is a different format: six questions. Answer them for the piece of work in front of you and the tool usually falls out. I will do the same for the six tools, and name the cases where each one is the wrong choice — including the tool I work on.
Question 1: Is this a picture or a behaviour?
If the artefact is a picture that needs to be reviewed, you need a design canvas. If it is a behaviour that needs to be run, you need a runtime.
Most arguments happen because a team asks one tool to be both. Figma is a picture tool with some behaviour features. GSAP is a behaviour library with no picture surface. Neither is broken; the expectation is.
Question 2: Does the animation need to know something at runtime?
The pointer position, the item count, the error, the user's plan, the viewport width, prefers-reduced-motion.
If the answer is no, a recorded format is cheaper and smaller. This is where Lottie wins and where a lot of teams over-engineer.
If the answer is yes, you need something that stores rules — a state machine, physics parameters, layout constraints — and you should stop evaluating playback formats.
Question 3: Is the artefact a character or an interface?
Characters want skeletal animation and mesh deformation: Spine, Rive.
Interfaces want layout, typography, accessibility and data: Exode, or an environment like it.
A character inside an interface is the genuinely hard case, and it usually means two tools plus a shared state machine. Pretending one tool covers both is how you end up with a mascot that breaks at 768px.
Question 4: Who tunes the motion?
If an engineer tunes it, a library is fine and often better. No new format, no new tool, full control.
If a designer tunes it, they need a visual surface, and the output has to be consumable without a translation step. That is the handoff question, and it is organisational more than technical.
If both tune it, you need the tuning values to live in one artefact both roles can read. That is the strongest argument for an open JSON format over a binary.
Question 5: What has to ship?
This one is brutal and clarifying.
- A prototype for a review: Figma. Nothing else is worth the setup.
- A marketing loop: Lottie or a video. Tiny files, no state.
- A component in a React app: whatever compiles to a component, or GSAP inside it.
- The same component on iOS and Android: a format that compiles to multiple platforms, or you accept drift.
- A game: Spine, Rive, or your engine's animation system.
- A whole interactive screen: an environment that owns layout + logic + export.
Question 6: What is the cost of the file being closed?
A binary animation file cannot be reviewed in a pull request, merged, linted in CI, or edited by an agent. Sometimes that does not matter — a decorative asset is a decorative asset.
It matters when motion is part of your product surface and changes every sprint. Then version control stops working at exactly the moment you need it most, and someone starts attaching screenshots to PRs.
The six tools, placed
Figma — picture tool, best-in-class collaboration, weak runtime story. Use it for exploration, systems, review. Do not use it to specify behaviour and expect behaviour to arrive.
Rive — character and vector state machines with a small, proven runtime. Excellent when the artefact is art and the target is a canvas or a game. Weak when the artefact is a responsive interface with text and focus states.
Lottie — recorded playback, tiny files, universal support. Perfect for illustration. Fundamentally unable to react, so do not build interaction on segment scrubbing.
GSAP — imperative animation library, unmatched sequencing, no design surface. Ideal when motion is authored in code by the people who ship it.
Spine — game skeletal animation, deepest rigging, engine pipelines. Wrong shape for UI.
Exode — an environment that treats layout, physics, state and export as one artefact: flexbox canvas, hierarchical state machines, simulated motion, .exode JSON, compiled components. Right when the whole screen is the deliverable. Wrong when you need one tiny asset, or when your team has no appetite for a new tool.
The combinations that actually work
The best stacks I have seen are not monogamous.
- Figma for exploration → Exode for the shipped screen. Draw freely, then build the real thing once behaviour matters.
- Spine for characters → Exode for the interface around them. Shared state machine, two rigs.
- GSAP for page choreography → component-level motion in the design artefact. Each tool does what it is best at.
- Lottie for decorative assets → everything interactive elsewhere. Resist the urge to make the illustration interactive.
The failure mode is not picking the "wrong" tool. It is picking four and making the integration everyone's side project.
A note on switching costs
Every migration costs more than the demo suggests: retraining, rebuilding libraries, re-earning trust. A tool has to remove a recurring, expensive problem to justify that. "It has more features" does not.
The recurring problem worth paying to remove, in my experience, is the gap between a screen that was designed and a screen that behaves. If that gap is not expensive for you, you do not need a new tool. You need a better design review.
Priya works on design systems at Exode. If question 5 landed on "a whole interactive screen", that is the case Exode is built for — exodeui.com. If it landed anywhere else, use the tool that fits.
Top comments (0)