<report>

Pruna AI

P-Video 2

v-benchmark v2 by Megaton benchmark
Generated via Segmind · September 10, 2026Evaluated · September 14, 2026Leaderboard rank #13 of 16Last updated · September 14, 2026

65.90

/ 100

<Our Take>

P2 is a fast and cheap model that can deliver great performances. A few generations genuinely surprised us with its good taste given its pricing. It's animation, text rendering and scenes with complex physics are areas that could be improved with subsequent runs.

Categories

C01

Prompt AdherenceWhether every explicit, observable instruction in the prompt is fulfilled.

This Video

100.0

This Video · Assessment

This video fulfills its explicit, observable instructions with near-complete consistency.

This Category

73.0

37 videos were rated for this category

This Category · Assessment

Most prompt requirements are fulfilled, though complex combinations or secondary details recur as limitations.

Models that scored similarly on this category

Violations

This Video
Essential2
Major1
Minor0
This Category
Essential30
Major36
Minor20

37 videos

C02

Scene ConsistencyWhether established subjects, objects, spatial relationships, environment, and lighting remain coherent over time.

This Video

100.0

This Video · Assessment

This video maintains subjects, objects, spatial relationships, environment, and lighting over time with near-complete consistency.

This Category

80.2

36 videos were rated for this category

This Category · Assessment

Temporal and spatial continuity is highly reliable, with only isolated drift or regeneration.

Models that scored similarly on this category

Violations

This Video
Essential0
Major0
Minor0
This Category
Essential39
Major23
Minor19

36 videos

C03

PhysicsWhether the central motion, force, material, and reaction behave coherently in the requested world and style.

This Video

100.0

This Video · Assessment

Fully convincing physical behavior, including under close inspection.

This Category

58.2

17 videos were rated for this category

This Category · Assessment

The intended physical events are recognizable, but repeated defects leave overall performance uneven.

C04

Human FidelityWhether faces, eyes, skin, bodies, hands, motion, and human presence remain anatomically and perceptually convincing.

This Video

100.0

This Video · Assessment

Fully convincing within the requested style, with no meaningful uncanny or anatomical defects.

This Category

81.1

27 videos were rated for this category

This Category · Assessment

Human rendering is highly convincing across the set, with only small localized defects.

C05

2D Animation Craft & ArtistryThe quality and artistic authorship of intentionally 2D animation within its requested conventions.

This Video

80.0

This Video · Assessment

Technically excellent and fully professional, but more conventional or less expressive than the highest tier.

This Category

38.3

12 videos were rated for this category

This Category · Assessment

The intended 2D style is present, but animation execution is commonly weak.

Models that scored similarly on this category

C06

3D Animation Craft & ArtistryThe quality, coherence, and artistic authorship of intentionally 3D animation.

This Video

80.0

This Video · Assessment

Technically excellent and fully professional, but more conventional or less expressive than the highest tier.

This Category

62.5

5 videos were rated for this category

This Category · Assessment

3D animation is generally coherent, but synthetic, mechanical, stiff, or underdeveloped execution recurs.

C07

Cinematography Craft & ArtistryShot design, framing, camera movement, lighting, focus, visual hierarchy, and intentional filmmaking authorship.

This Video

90.0

This Video · Assessment

Exceptional professional work with highly intentional framing, light, movement, and visual storytelling.

This Category

78.6

7 videos were rated for this category

This Category · Assessment

Shot design is technically strong overall, though recurring choices are more conventional or less expressive.

Models that scored similarly on this category

C08

Taste & Art DirectionThe strength, specificity, coherence, and artistic authorship of the result as a finished creative choice.

This Video

90.0

This Video · Assessment

Excellent and memorable art direction with highly intentional, sophisticated choices.

This Category

62.2

60 videos were rated for this category

This Category · Assessment

Art direction is generally capable, but generic, unresolved, or less-confident choices recur.

Models that scored similarly on this category

C09

Animal & Creature FidelityWhether real animals and designed creatures preserve convincing anatomy, identity, motion, behavior, and interaction.

This Video

100.0

This Video · Assessment

Fully convincing anatomy, surfaces, movement, and behavior within the requested species or creature design.

This Category

67.1

7 videos were rated for this category

This Category · Assessment

Animals and creatures remain believable in most cases, though synthetic anatomy, motion, or behavior recurs.

Models that scored similarly on this category

C10

Object & Product FidelityWhether requested objects have correct identity, structure, materials, details, motion, and interactions.

This Video

100.0

This Video · Assessment

This video preserves the requested objects’ identity, structure, materials, details, motion, and interactions with near-complete consistency.

This Category

59.8

8 videos were rated for this category

This Category · Assessment

Object fidelity is mixed: requested products remain recognizable, but important properties are uneven.

Violations

This Video
Essential0
Major0
Minor0
This Category
Essential18
Major10
Minor4

8 videos

C11

Causal & Semantic CoherenceWhether events, relationships, transformations, and cause-and-effect remain semantically coherent, including surreal premises.

This Video

100.0

This Video · Assessment

Actions, roles, relationships, and causal progression are completely coherent within the premise.

This Category

62.7

11 videos were rated for this category

This Category · Assessment

Central events are generally understandable, but strained or unexplained actions and relationships recur.

Models that scored similarly on this category

C12

Text FidelityWhether requested text content, typography, placement, attachment, legibility, and temporal stability are correct.

This Video

52.8

This Video · Assessment

This video preserves requested text content, typography, placement, attachment, legibility, and temporal stability unevenly, mixing clear successes with repeated failures.

This Category

20.6

5 videos were rated for this category

This Category · Assessment

Severe spelling, legibility, placement, attachment, or temporal failures dominate much of the set.

Violations

This Video
Essential4
Major2
Minor0
This Category
Essential32
Major9
Minor0

5 videos

C13

Character Performance & ActingWhether character intention, emotion, timing, reactions, eyelines, gesture, and performance read convincingly.

This Video

90.0

This Video · Assessment

Highly compelling and memorable performance with strong timing, attention, and expressive choices.

This Category

71.0

21 videos were rated for this category

This Category · Assessment

Performances are professional and convincing overall, though recurring choices are more conventional or less nuanced.

Models that scored similarly on this category

Differences between other models

Megatonmegaton.ai

C01Prompt Adherence

P-Video 2Kling 3 StandardSeedance 2.0 MiniVeo 3.1Gen-4.5
Megatonmegaton.ai

C03Physics

P-Video 2Kling 3 StandardSeedance 2.0 MiniVeo 3.1Gen-4.5
Megatonmegaton.ai

C02Scene Consistency

P-Video 2Kling 3 StandardSeedance 2.0 MiniVeo 3.1Gen-4.5
Megatonmegaton.ai

C04Human Fidelity

P-Video 2Kling 3 StandardSeedance 2.0 MiniVeo 3.1Gen-4.5
Megatonmegaton.ai

C052D Animation Craft & Artistry

P-Video 2Kling 3 StandardSeedance 2.0 MiniVeo 3.1Gen-4.5
Megatonmegaton.ai

C063D Animation Craft & Artistry

P-Video 2Kling 3 StandardSeedance 2.0 MiniVeo 3.1Gen-4.5
Megatonmegaton.ai

C11Causal & Semantic Coherence

P-Video 2Kling 3 StandardSeedance 2.0 MiniVeo 3.1Gen-4.5
Megatonmegaton.ai

C08Taste & Art Direction

P-Video 2Kling 3 StandardSeedance 2.0 MiniVeo 3.1Gen-4.5
Megatonmegaton.ai

C07Cinematography Craft & Artistry

P-Video 2Kling 3 StandardSeedance 2.0 MiniVeo 3.1Gen-4.5
Megatonmegaton.ai

C09Animal & Creature Fidelity

P-Video 2Kling 3 StandardSeedance 2.0 MiniVeo 3.1Gen-4.5
Megatonmegaton.ai

C10Object & Product Fidelity

P-Video 2Kling 3 StandardSeedance 2.0 MiniVeo 3.1Gen-4.5
Megatonmegaton.ai

C12Text Fidelity

P-Video 2Kling 3 StandardSeedance 2.0 MiniVeo 3.1Gen-4.5
Megatonmegaton.ai

C13Character Performance & Acting

P-Video 2Kling 3 StandardSeedance 2.0 MiniVeo 3.1Gen-4.5
CategoryP-Video 2Kling 3 StandardSeedance 2.0 MiniVeo 3.1Gen-4.5
C01Prompt Adherence73.083.087.979.776.2
C03Physics58.257.650.645.338.8
C02Scene Consistency80.282.277.980.078.1
C04Human Fidelity81.179.664.865.967.4
C052D Animation Craft & Artistry38.346.461.745.047.5
C063D Animation Craft & Artistry62.566.062.049.044.0
C11Causal & Semantic Coherence62.766.470.965.572.7
C08Taste & Art Direction62.263.160.751.949.7
C07Cinematography Craft & Artistry78.675.767.162.970.0
C09Animal & Creature Fidelity67.171.482.967.161.4
C10Object & Product Fidelity59.890.088.495.474.5
C12Text Fidelity20.630.363.841.747.2
C13Character Performance & Acting71.059.063.062.063.0

Leaderboard

Megatonmegaton.ai
Megaton

Video Index

Measuring video model quality and performance

Video Index · 0–100

Scores are from the active Megaton v-benchmark v2.0 publication. See how our scores are calculated in our Methodology.

<Cite this report>

Citation snippet

Copy a plain-text citation for articles and research.

Megaton. “P-Video 2 V-Benchmark v2 report.” V-Benchmark results publication 1.5.0, September 14, 2026. https://megaton.ai/evaluations/pruna/p-video-2/report

Report badge

The badge links to this report so readers can verify the ranking.

P-Video 2 is ranked #13 in Megaton V-Benchmark v2

Methodology

Our score is based on our v-benchmark v2 methodology. This is a Core 36 evaluation with 83.98% weighted coverage, not a Full evaluation.

Read the full methodology →

Changelog

Published
September 14, 2026
V-Benchmark version
2.2.0
Megaton Index
65.90
Published rank
13 of 16
View model changelog →

<What you get>

  • The prompts and outputs used in our tests.
  • Scores and notes explaining each result.
  • Comparisons with other models and versions.

<Who can get access>

  • Are you part of the Pruna AI team? Please reach out for complimentary access.
  • University and nonprofit researchers can ask about research access.
  • Paid access starts at $2,500 per report.
Access option

We’ll use these details to reply to your request.