test_id stringlengths 20 26 | prompt_id stringlengths 11 20 | prompt stringlengths 129 156 | workflow stringclasses 2
values | use_case stringclasses 8
values | test_goal stringlengths 55 73 | expected_behavior stringlengths 105 136 | status stringclasses 1
value | observed_result stringclasses 1
value | motion_score null | camera_score null | subject_consistency null | visual_quality null | prompt_adherence null | commercial_readiness null | failure_mode null | strengths stringclasses 1
value | limitations stringclasses 1
value | evidence_url stringclasses 1
value | tested_at stringclasses 1
value | tester_type stringclasses 1
value | source_type stringclasses 1
value | notes stringclasses 2
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
h3-benchmark-product-001 | h3-product-001 | Luxury perfume bottle in a controlled commercial scene, natural secondary motion, slow push-in, balanced studio lighting, polished product styling. | text-to-video | product | Evaluate premium product shape in a repeatable MiniMax H3 workflow. | The Luxury perfume bottle should remain coherent while the requested slow push-in and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-product-002 | h3-product-002 | mechanical wristwatch in a controlled commercial scene, natural secondary motion, dolly out, balanced studio lighting, polished product styling. | text-to-video | product | Evaluate fine mechanical motion in a repeatable MiniMax H3 workflow. | The mechanical wristwatch should remain coherent while the requested dolly out and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-product-003 | h3-product-003 | mirrorless camera in a controlled commercial scene, natural secondary motion, pan, balanced studio lighting, polished product styling. | text-to-video | product | Evaluate lens-ring reveal in a repeatable MiniMax H3 workflow. | The mirrorless camera should remain coherent while the requested pan and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-product-004 | h3-product-004 | wireless headphones in a controlled commercial scene, natural secondary motion, crane, balanced studio lighting, polished product styling. | text-to-video | product | Evaluate suspended silhouette in a repeatable MiniMax H3 workflow. | The wireless headphones should remain coherent while the requested crane and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-product-005 | h3-product-005 | electric car in a controlled commercial scene, natural secondary motion, low-angle tracking, balanced studio lighting, polished product styling. | text-to-video | product | Evaluate vehicle reflections in a repeatable MiniMax H3 workflow. | The electric car should remain coherent while the requested low-angle tracking and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-ecommerce-006 | h3-ecommerce-001 | white sneakers in a controlled commercial scene, natural secondary motion, orbit, balanced studio lighting, polished ecommerce styling. | text-to-video | ecommerce | Evaluate catalog-accurate geometry in a repeatable MiniMax H3 workflow. | The white sneakers should remain coherent while the requested orbit and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-ecommerce-007 | h3-ecommerce-002 | skincare jar in a controlled commercial scene, natural secondary motion, macro push-in, balanced studio lighting, polished ecommerce styling. | text-to-video | ecommerce | Evaluate packaging fidelity in a repeatable MiniMax H3 workflow. | The skincare jar should remain coherent while the requested macro push-in and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-ecommerce-008 | h3-ecommerce-003 | smartphone in a controlled commercial scene, natural secondary motion, tilt-up reveal, balanced studio lighting, polished ecommerce styling. | text-to-video | ecommerce | Evaluate screen continuity in a repeatable MiniMax H3 workflow. | The smartphone should remain coherent while the requested tilt-up reveal and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-social-009 | h3-social-001 | creator holding a drink in a controlled commercial scene, natural secondary motion, handheld push-in, balanced studio lighting, polished social styling. | text-to-video | social | Evaluate hand-object interaction in a repeatable MiniMax H3 workflow. | The creator holding a drink should remain coherent while the requested handheld push-in and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-social-010 | h3-social-002 | compact appliance unboxing in a controlled commercial scene, natural secondary motion, overhead tilt, balanced studio lighting, polished social styling. | text-to-video | social | Evaluate multi-step demo in a repeatable MiniMax H3 workflow. | The compact appliance unboxing should remain coherent while the requested overhead tilt and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-character-011 | h3-character-001 | young athlete in a controlled commercial scene, natural secondary motion, forward tracking, balanced studio lighting, polished character styling. | text-to-video | character | Evaluate walking continuity in a repeatable MiniMax H3 workflow. | The young athlete should remain coherent while the requested forward tracking and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-character-012 | h3-character-002 | character in a coat in a controlled commercial scene, natural secondary motion, slow orbit, balanced studio lighting, polished character styling. | text-to-video | character | Evaluate identity preservation in a repeatable MiniMax H3 workflow. | The character in a coat should remain coherent while the requested slow orbit and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-camera-013 | h3-camera-001 | cyclist in a city street in a controlled commercial scene, natural secondary motion, side tracking, balanced studio lighting, polished camera styling. | text-to-video | camera | Evaluate camera-direction compliance in a repeatable MiniMax H3 workflow. | The cyclist in a city street should remain coherent while the requested side tracking and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-camera-014 | h3-camera-002 | ceramic cup in a controlled commercial scene, natural secondary motion, orbit, balanced studio lighting, polished camera styling. | text-to-video | camera | Evaluate static-subject framing in a repeatable MiniMax H3 workflow. | The ceramic cup should remain coherent while the requested orbit and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-cinematic-015 | h3-cinematic-001 | rainy neon street in a controlled commercial scene, natural secondary motion, dolly forward, balanced studio lighting, polished cinematic styling. | text-to-video | cinematic | Evaluate atmosphere continuity in a repeatable MiniMax H3 workflow. | The rainy neon street should remain coherent while the requested dolly forward and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-cinematic-016 | h3-cinematic-002 | fantasy mountain pass in a controlled commercial scene, natural secondary motion, crane reveal, balanced studio lighting, polished cinematic styling. | text-to-video | cinematic | Evaluate large-scale composition in a repeatable MiniMax H3 workflow. | The fantasy mountain pass should remain coherent while the requested crane reveal and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-food-017 | h3-food-001 | ramen bowl in a controlled commercial scene, natural secondary motion, macro push-in, balanced studio lighting, polished food styling. | text-to-video | food | Evaluate steam and noodle continuity in a repeatable MiniMax H3 workflow. | The ramen bowl should remain coherent while the requested macro push-in and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-food-018 | h3-food-002 | chocolate cake in a controlled commercial scene, natural secondary motion, gentle orbit, balanced studio lighting, polished food styling. | text-to-video | food | Evaluate viscous food motion in a repeatable MiniMax H3 workflow. | The chocolate cake should remain coherent while the requested gentle orbit and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-fashion-019 | h3-fashion-001 | model in a structured jacket in a controlled commercial scene, natural secondary motion, tracking shot, balanced studio lighting, polished fashion styling. | text-to-video | fashion | Evaluate garment silhouette in a repeatable MiniMax H3 workflow. | The model in a structured jacket should remain coherent while the requested tracking shot and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-product-020 | h3-product-006 | wireless speaker in a controlled commercial scene, natural secondary motion, reveal shot, balanced studio lighting, polished product styling. | text-to-video | product | Evaluate surface and logo stability in a repeatable MiniMax H3 workflow. | The wireless speaker should remain coherent while the requested reveal shot and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-ecommerce-021 | h3-ecommerce-004 | running shoes in a controlled commercial scene, natural secondary motion, turntable orbit, balanced studio lighting, polished ecommerce styling. | text-to-video | ecommerce | Evaluate sole consistency in a repeatable MiniMax H3 workflow. | The running shoes should remain coherent while the requested turntable orbit and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-social-022 | h3-social-003 | lifestyle product in hand in a controlled commercial scene, natural secondary motion, quick push-in, balanced studio lighting, polished social styling. | text-to-video | social | Evaluate creator framing in a repeatable MiniMax H3 workflow. | The lifestyle product in hand should remain coherent while the requested quick push-in and secondary motion stay smooth and readable. | pending | null | null | null | null | null | null | null | community-curated | Linked to the existing Prompt Dataset; no run has been performed yet. | ||||||
h3-benchmark-i2v-001 | h3-i2v-portrait-001 | Animate the provided portrait with subtle blink and slow push-in. Preserve the source composition, identity, proportions, and key details. | image-to-video | social | Evaluate subtle blink and slow push-in from a reference image. | The source portrait should remain stable while subtle blink and slow push-in stays natural and controlled. | pending | null | null | null | null | null | null | null | community-curated | Image-to-video case; source image is required before testing. | ||||||
h3-benchmark-i2v-002 | h3-i2v-fashion-002 | Animate the provided fashion look with runway step and side tracking. Preserve the source composition, identity, proportions, and key details. | image-to-video | fashion | Evaluate runway step and side tracking from a reference image. | The source fashion look should remain stable while runway step and side tracking stays natural and controlled. | pending | null | null | null | null | null | null | null | community-curated | Image-to-video case; source image is required before testing. | ||||||
h3-benchmark-i2v-003 | h3-i2v-product-003 | Animate the provided product image with rotation and highlight sweep. Preserve the source composition, identity, proportions, and key details. | image-to-video | product | Evaluate rotation and highlight sweep from a reference image. | The source product image should remain stable while rotation and highlight sweep stays natural and controlled. | pending | null | null | null | null | null | null | null | community-curated | Image-to-video case; source image is required before testing. | ||||||
h3-benchmark-i2v-004 | h3-i2v-landscape-004 | Animate the provided landscape image with cloud and grass motion. Preserve the source composition, identity, proportions, and key details. | image-to-video | cinematic | Evaluate cloud and grass motion from a reference image. | The source landscape image should remain stable while cloud and grass motion stays natural and controlled. | pending | null | null | null | null | null | null | null | community-curated | Image-to-video case; source image is required before testing. | ||||||
h3-benchmark-i2v-005 | h3-i2v-food-005 | Animate the provided food image with steam and top-down push-in. Preserve the source composition, identity, proportions, and key details. | image-to-video | food | Evaluate steam and top-down push-in from a reference image. | The source food image should remain stable while steam and top-down push-in stays natural and controlled. | pending | null | null | null | null | null | null | null | community-curated | Image-to-video case; source image is required before testing. | ||||||
h3-benchmark-i2v-006 | h3-i2v-character-006 | Animate the provided character image with one step and three-quarter orbit. Preserve the source composition, identity, proportions, and key details. | image-to-video | character | Evaluate one step and three-quarter orbit from a reference image. | The source character image should remain stable while one step and three-quarter orbit stays natural and controlled. | pending | null | null | null | null | null | null | null | community-curated | Image-to-video case; source image is required before testing. | ||||||
h3-benchmark-i2v-007 | h3-i2v-ecommerce-007 | Animate the provided ecommerce product with turntable rotation and push-in. Preserve the source composition, identity, proportions, and key details. | image-to-video | ecommerce | Evaluate turntable rotation and push-in from a reference image. | The source ecommerce product should remain stable while turntable rotation and push-in stays natural and controlled. | pending | null | null | null | null | null | null | null | community-curated | Image-to-video case; source image is required before testing. | ||||||
h3-benchmark-i2v-008 | h3-i2v-fashion-008 | Animate the provided portrait-fashion image with shoulder turn and close-up reveal. Preserve the source composition, identity, proportions, and key details. | image-to-video | fashion | Evaluate shoulder turn and close-up reveal from a reference image. | The source portrait-fashion image should remain stable while shoulder turn and close-up reveal stays natural and controlled. | pending | null | null | null | null | null | null | null | community-curated | Image-to-video case; source image is required before testing. |
MiniMax H3 Commercial Prompt Benchmark
A community benchmark evaluating MiniMax H3 video-generation prompts across commercial, cinematic and motion-focused workflows.
The benchmark tracks expected behavior, observed results, strengths, limitations and common failure modes instead of publishing prompt examples alone.
Try MiniMax H3 Text to Video →
What this dataset is
This first release contains 30 linked benchmark cases covering product advertising, ecommerce, social advertising, character motion, camera motion, cinematic scenes, food, fashion, and image-to-video. It is a test set and protocol: cases that have not been run are explicitly marked pending.
The source prompt library is MiniMax H3 Structured Video Prompts. The Prompt Dataset is the discovery asset; this Benchmark is the evidence asset.
Schema
Each JSONL row includes test_id, prompt_id, prompt, workflow, use_case, test_goal, expected_behavior, status, observed_result, motion_score, camera_score, subject_consistency, visual_quality, prompt_adherence, commercial_readiness, failure_mode, strengths, limitations, evidence_url, tested_at, tester_type, source_type, and notes.
Methodology
Each benchmark case begins with a defined prompt and expected behavior. Tested cases are evaluated across motion, camera adherence, subject consistency, visual quality, prompt adherence, and commercial readiness.
Cases that have not been run are marked pending. No performance score should be assigned until an actual result has been reviewed. A case may be marked tested only when it was run through an authorized MiniMax H3 workflow, has a clearly linked real output, or has a verifiable historical output.
This is an independent community benchmark and not an official MiniMax evaluation. No credits are purchased automatically and no result is inferred from the prompt alone.
Community evaluation rubric
Scores are ordinal community judgments, not scientifically precise measurements:
| Score | Meaning |
|---|---|
| 1 | Poor |
| 2 | Limited |
| 3 | Usable for concept exploration |
| 4 | Strong |
| 5 | Excellent for the tested criterion |
commercial_readiness: 1 = not commercially usable; 2 = concept only with clear issues; 3 = useful draft; 4 = close to mature content after light post-production; 5 = close to directly usable commercial creative.
Benchmark by use case
- Product Advertising — Try this workflow
- E-commerce — Try this workflow
- Character Motion — Try this workflow
- Camera Motion — Try this workflow
- Cinematic — Try this workflow
- Image to Video — Try this workflow
Try the prompts yourself
Findings policy
There are no Early Findings in this release because the cases are pending and no real test results are being claimed. Once enough cases are genuinely tested, aggregate scores and failure-mode frequencies may be calculated directly from the records.
Disclaimer
Independent community resource. Not official, affiliated, or endorsed by MiniMax. Do not infer capability from a pending case. Review generated media, permissions, and third-party rights before commercial use.
- Downloads last month
- 65