Claude Fable 5.2 for Motion Graphics: Is It Worth It?
Claude Fable 5.2 came out on November 9 with a price to match: $8 per million input tokens and $40 per million output, twice Opus 5.5 and four times Sonnet 5.5 (OpenRouter). So is it worth it for animation and video? I ran the same seven motion graphics prompts through all three models and put every film below, along with the time and the bill.
The short answer: Fable 5.2 made clearly better films on four of seven prompts: kinetic typography, a model launch video, a product launch video and an open brief. On a logo reveal and a data chart the gap was small, and Sonnet at a fifth of the price or less was the better deal. Opus 5.5 was the real competition: it won the pixel-art scene and matched Fable on the product video, but fell behind wherever the job was pure visual design. Fable's edge comes in two parts. It picks the right idea for the brief, and it carries the idea out with better detail and timing: camera moves, how one shot hands off to the next, and when each element lands. None of the prompts was a detailed brief: the shortest is 9 words and the longest 28, and Fable worked out on its own what to show and how to move it.
How I tested
- Same agent, system prompt, tools and renderer as iArt's video agent. Only the model changed.
- Thinking set to medium for all three, which is what Auto runs Fable at in iArt. One run per model per prompt, no retries, no edits.
- Time is the agent's wall-clock time until the film was finished. Cost is what the run costs on iArt.
- A turn is one round of the agent: think, call tools, read the result. Fewer turns means fewer fix-ups.
- Prompts are written the way people actually type them, from one line to a short paragraph.
Results at a glance
| Prompt | Sonnet 5.5 | Opus 5.5 | Fable 5.2 |
|---|---|---|---|
| Kinetic typography | $0.22 0.8 min · 4 turns | $1.24 4.2 min · 11 turns | $2.39 4.3 min · 8 turns |
| Launch video for a new AI model | $0.42 2.2 min · 10 turns | $1.02 3.0 min · 8 turns | $2.07 4.3 min · 6 turns |
| Product launch video from screenshots | $0.29 2.0 min · 14 turns | $1.92 7.0 min · 20 turns | $2.49 5.7 min · 16 turns |
| 3D logo reveal | $0.34 2.2 min · 7 turns | $1.30 6.0 min · 16 turns | $1.63 3.2 min · 9 turns |
| Pixel-art scene | $0.36 1.4 min · 5 turns | $1.87 10.0 min · 9 turns | $4.14 8.3 min · 7 turns |
| Animated data chart | $0.32 1.7 min · 9 turns | $0.95 11.2 min · 11 turns | $1.40 2.2 min · 7 turns |
| An open brief | $0.46 2.8 min · 10 turns | $3.49 12.6 min · 23 turns | $4.25 7.8 min · 11 turns |
| All seven | $2.41 13.3 min · 59 turns | $11.79 54.1 min · 98 turns | $18.37 35.8 min · 64 turns |
Cost on iArt, then agent time and turns. One run each, so treat single rows as examples rather than averages.
Across all seven, Fable cost 7.6× what Sonnet did but only 1.6× what Opus did, even though its tokens cost twice as much. It needed fewer turns than Opus on every prompt (64 against 98) and finished sooner on five of the seven.
The seven films
Each player starts on Fable 5.2. Click Sonnet or Opus below it to switch films at the same moment, with sound.
1. Kinetic typography
Make a kinetic typography video for this line: "The best way to predict the future is to invent it."
The biggest gap of the seven, and Fable's film is the one to beat. Sonnet crosses out predict the future, drops in invent it., then holds a still frame for the last four seconds, with sound effects and no music. Fable treats the line as a decision: dotted branches fan out from the best way, ghost versions of future flicker, and one red stroke draws through to the answer, while the camera pans and pulls back and each word lands on a piano note. Opus builds a similar idea, with drafting lines under invent and a scrambling predict, but packs more type detail into each frame and moves with less drive.
Try this prompt with Fable 5.2 →
2. Launch video for a new AI model
Make a launch video for Claude Fable 5.2, Anthropic's new model.
Another big gap, and it's about motion. Sonnet tells three separate metaphors (a knotted line, an edited text card, a node graph) joined by identical full-screen wipes, with long silences in between. Fable follows one line for the whole film: a typed prompt becomes a dot, the dot becomes code, a chart and handwriting under Through code / data / words, then spirals into the logo. It's one tracking shot with no cuts, in 19 seconds. Opus uses the same one-line structure with a nice concrete demo (a prompt asks to explain the tides as a story, and the line draws the Moon pulling the sea), but it slows down to type out a three-line story in the middle.
Try this prompt with Fable 5.2 →
3. Product launch video from screenshots
Make a launch video for iArt (iart.ai): you type an idea, it makes a motion graphics video. Screenshots of the app attached.
Sonnet loses the thread. It types a prompt, shows three still cards, pushes into the app, moves to a library screen, and never shows a video the product made; the soundtrack is mostly silence. Opus shows the most product: one coffee-brand film in three styles, including a section cut to the beat. For a product video, that much information is earned. Fable is shorter and tighter: a hook, the prompt box morphs into a player, the result plays on drum hits, then the style library and the call to action, though it shows less of the output than Opus. Both had a split-second glitch: a tiled frame in Opus near 0:02, a blank flash in Fable at 0:15.
Try this prompt with Fable 5.2 →
4. 3D logo reveal
make my logo reveal in 3D with chromatic metal and grids
The gap is small. Sonnet gets the chrome look too, but the grid lines make the frame busy, and the action is over in three seconds. Opus traces the logo in cyan and magenta wireframe, fills it with a scanline and sends a shockwave across the floor. Fable goes synthwave: neon strokes sketch the mark, which extrudes, fills and turns to chrome. All three hold the finished logo for the second half. For a logo sting, Sonnet at a fifth of the price is good enough.
Try this prompt with Fable 5.2 →
5. Pixel-art scene
Make a cozy pixel-art animated scene: a forest village seen from above, with paths, a cabin, a pond and a cave, and little bits of life moving around.
Opus made the best film here. It's the only one with a camera: it starts close on the cabin, follows a boy to the pond, pans up to a miner leaving the cave, then pulls out to reveal the whole village as day turns to night. Fable keeps the whole map in view and tells a small story (two characters meet, one fishes, one lights a campfire, the lanterns come on); it's lively, but the night scene gets busy. Sonnet makes a pleasant loop of villagers walking, with no arc. A tour is exactly what this scene asked for, so Opus's habit of showing more pays off.
Try this prompt with Fable 5.2 →
6. Animated data chart
Make an animated chart video of world population: 1950 2.5 billion, 1975 4.1 billion, 2000 6.1 billion, 2025 8.2 billion.
The gap is small. Four numbers make a short brief, and all three get the chart right. Fable's is the cleanest: one line, counters in the corners, a pulse on each milestone and a ×3.3 card at the end. Sonnet draws almost the same chart, but its last caption overlaps the year as it fades in. Opus adds a second chart, a grid of dots where each dot is 100 million people. It helps you feel the scale and makes the frame busier.
Try this prompt with Fable 5.2 →
7. An open brief
Make the most impressive motion design video you can.
With no brief at all, the model has to decide what impressive means, and Fable got it right: it made a showreel of motion design itself. A single point drops, gets plucked into a line, becomes a circle, triangle and square, splits into a grid, rises into 3D columns, orbits, bursts into a swarm and assembles the word MOTION. Every change is a transformation, not a cut. Opus turned the prompt into a science story about a photon's journey from the Sun to your eye, with counters and labels. It's well made, but it's an explainer, and for a brief that asks to show off design, the story is extra weight. Sonnet tries a shorter version of Fable's idea and leaves almost two seconds of empty orange screen in the middle.
Try this prompt with Fable 5.2 →
Where Fable 5.2 is worth it
Fable's lead was widest where a video has to feel designed rather than assembled: typography, launches and open briefs. Those depend on two things cheaper models do less well. The first is the idea. Given "the most impressive video you can", Fable picked a concept and built the whole film around it. The second is execution: camera moves with weight, transitions that turn one shape into the next instead of cutting, and words and objects that land on the beat. Sonnet usually gets the content right and the timing flat. Most of its films also had no music, only isolated sound effects with silence in between, and several ended on a frozen frame for three to four seconds.
Where the brief already contains the idea, as with four population numbers or one logo, the gap shrinks to polish, and polish is rarely worth 5–10× the price.
Fable 5.2 vs Opus 5.5
Opus 5.5 is in the same league for ideas and story, and it won one prompt outright. The difference is how much each puts on screen. Opus adds information: labels, counters, a second chart, a story. When the video's job is to explain or show around, that's the right call. It's why Opus made the best pixel-art tour, the most informative population chart and the fullest product demo. When the job is to look good, the extra information is extra weight: a science story in a brief that asked for impressive design, type detail that slows down a kinetic line. Fable does one thing per frame and keeps it moving.
Fable is also the more efficient worker. It took fewer turns on every prompt (on the open brief, 11 against Opus's 23) and spent 35.8 minutes in total against 54.1 for Opus. That's why a model with twice the token price ended up only 1.6× the bill.
Which model should you use?
- Sonnet 5.5 for routine edits and simple pieces: a chart from a few numbers, a basic logo sting, text changes.
- Opus 5.5 for explainers, product walk-throughs and scenes that benefit from a guided tour, where more information on screen helps.
- Fable 5.2 for the video people will actually watch: a launch, a title sequence, kinetic type, or an open brief where the idea has to come from the model. Make the first version with Fable, then switch to a cheaper model for small edits.
Using Fable 5.2 in iArt
Pick Fable 5.2 in the model menu next to the prompt box. It's marked with five cost dots, the most of any model, and is available on paid plans. Leave Thinking on Auto, which runs it at medium, the same setting as this test. While Fable works, iArt reminds you that it's a costly choice for routine edits and offers a one-click switch to DeepSeek 4.1 for the next one.
FAQ
Is Claude Fable 5.2 good at motion graphics?
Yes. Across seven prompts it made the strongest film on four of them: kinetic typography, a model launch video, a product launch video and an open brief. Its edge is picking the right idea for the brief, plus the timing and camera work. On a logo reveal and a simple data chart, cheaper models came close.
How much does a Fable 5.2 video cost?
In this test, one Fable 5.2 film cost between $1.40 and $4.25 at iArt prices, against $0.22 to $0.46 for Sonnet 5.5. Across all seven prompts Fable cost 7.6× Sonnet and 1.6× Opus 5.5.
Fable 5.2 or Opus 5.5 for video?
Opus 5.5 is close and costs less. It won our pixel-art scene and showed the most product in the launch video, because it tends to put more information on screen, which suits explainers and tours. Fable keeps frames cleaner and the motion going, and it worked faster: fewer turns on all seven prompts and less time on five of the seven.
Can I use Fable 5.2 on a free iArt plan?
No. Fable 5.2 is available on paid plans. Free accounts can use the other models, including Sonnet 5.5 and DeepSeek 4.1.
Should I use Fable 5.2 for every edit?
No. Make the first version with Fable, then switch to a cheaper model for small changes like text, colours or timing. Every edit re-reads the project, so a top-tier model on routine edits costs far more than it adds.