All H2H comparisons

Odyssey anime / Seedance 2

Kimi K3 vs Claude Opus 5

Claude Opus 5 wins this prompt-writing challenge by 1 point. Watch both unedited Seedance 2 outputs and inspect the prompts below.

Comparison videos

Moonshot AI

Kimi K3

72%

Anthropic

Claude Opus 5

73%

Overall human score

Kimi K3

72%

Overall human score

Claude Opus 5

73%

Shared brief

Create an eight-second, 16:9, hand-drawn 2D cel-anime retelling of Odysseus and the Sirens in exactly four connected shots. Use historically grounded Late Bronze Age Aegean materials, precise character action, cinematic camera direction, and native sound.

Prompt record

What each model sent to Seedance 2

The model responses below are presented without editorial rewriting.

Moonshot AI

Kimi K3

Hand-drawn 2D cel anime; Late Bronze Age Aegean; 16:9; four connected shots, eight seconds; ship travels left-to-right throughout. Shot one (2s): low 35mm deck-level dolly along the weathered penteconter—pine planking, upswept sternpost, flax rigging, oxhide shields on the rail—barefoot sailors in wool chitons knead beeswax pellets between thumb and forefinger, pressing plugs into their ears, brows furrowed; bronze knife, rope coils; watercolor sky, harsh noon light, glassy calm. Shot two (2s): static 50mm at the mast; two crewmen lash Odysseus—short black beard, leather headband, linen chiton, bronze-studded belt—upright, rope crossing his chest, hands hauling knots taut behind him; he nods once, jaw set. Shot three (2s): slow push-in as Odysseus strains, cords creaking, tendons taut, mouth pleading toward a limestone islet at frame right, where three bird-bodied Sirens with women's faces perch among pale bones, ink-drawn wings half-furled; deaf crew rows past, blurred foreground multiplane. Shot four (2s): wide 28mm; wind returns, square sail snaps full, hand-inked spray and wake foam white; crew pulls wax from their ears and cuts him down as the islet recedes astern at frame left. Palette: sun-bleached ochre, Aegean teal, bone white, bronze. Visible cel grain, varied line weight, watercolor multiplane backgrounds, limited animation punctuated by fluid rowing smears. Hard cuts. Sound: creaking timbers, oar-thole thuds, dying wind, layered female song swelling then fading, sail snap, gulls.

Anthropic

Claude Opus 5

Hand-drawn 2D cel anime, painted gouache backgrounds, visible cel grain, expressive brush-ink linework, multiplane depth. 16:9, four shots, eight seconds, glassy noon Aegean calm. SHOT 1 (0.0–2.0s): 35mm-equivalent low deck-level view, slow dolly right along oar benches. Two bare-chested rowers in linen loincloths knead honey-amber beeswax between calloused thumbs, warmed over a small pottery brazier; a bearded sailor with bronze earring presses plugs into his ears, jaw tightening. Wax gleams with drawn highlight dots. Sound: creaking pine, wax squelch, lapping water. SHOT 2 (2.0–3.8s): Cut to matching screen direction, 50mm, camera cranes up the fir mast. Four crewmen wind twisted flax rope around Odysseus's chest, hands passing rope hand-over-hand, one bracing a knee on the mast step. He wears a saffron-bordered wool tunic, leather baldric, tin-inlaid dagger sheath, greying curls windless. Sound: rope hiss, grunts, hemp tension. SHOT 3 (3.8–6.0s): Close on Odysseus's face, subtle three-frame limited animation, then fluid snap: pupils dilate, brows crush, tendons flare, mouth opening in soundless command, fingers clawing rope. Rack focus past him to a distant limestone island—white cliffs, bone-pale scree, three slender winged silhouettes perched, mouths open. Sound: layered wordless female harmonics, thinning wind. SHOT 4 (6.0–8.0s): Wide starboard beam, telephoto compression, slow pan left as fifty painted oars lift, dip, and pull in unison, drawn foam swirls trailing. Island slides behind the stern; sunlight warms ochre, deep indigo sea. Sound: oar-thump cadence, receding song, single gull.

Scoring

A simple rubric

One overall human score, guided by four equally weighted questions. The percentages are not model-generated.

  1. 01

    Prompt adherence

    Does the video deliver the requested actions, sequence, and historical details?

  2. 02

    Visual craft

    Are the composition, cel-anime aesthetic, lighting, and material details convincing?

  3. 03

    Motion and continuity

    Do characters, props, geography, and transitions remain readable and stable?

  4. 04

    Story and sound

    Does the eight-second arc communicate clearly, with audio that supports the scene?

Method

Same setup, different prompt writer

01

Each frontier model received the same creative brief through Vercel AI Gateway when available, with a direct provider fallback.

02

Each response was passed to Seedance 2.0 without prompt editing.

03

Every output used the same 720p, 16:9, eight-second settings with native audio.

04

The percentages are human editorial scores guided by the rubric below.