68 Prompts both models were run on. Each slide shows the two clips for one Prompt, with the Prompt and the human preference split underneath.
Summary and how to choose
Duration: 4s / 6s / 8s (8s required for extension, reference images, 1080p and 4K) for Google Veo 3.1, versus not confirmed for Wan 2.1. Resolution: 720p (default) · 1080p (8s only) · 4K (8s only, not on Lite) · 24fps for Google Veo 3.1, versus not confirmed for Wan 2.1. Controls: T2V · I2V · Ingredients (up to 3 reference images, not on Lite) · First–last frame · Scene extension (~148s max, not on Lite) for Google Veo 3.1, versus not confirmed for Wan 2.1. Audio: Included in video pricing for Google Veo 3.1, versus not confirmed for Wan 2.1.
Google Veo 3.1
Strengths: a documented output envelope (4s / 6s / 8s (8s required for extension, reference images, 1080p and 4K) · 720p (default) · 1080p (8s only) · 4K (8s only, not on Lite) · 24fps), documented generation controls (T2V · I2V · Ingredients (up to 3 reference images, not on Lite) · First–last frame · Scene extension (~148s max, not on Lite)), documented audio support (Included in video pricing), published exact per-second API rates, and a checked plan, credit or no-rate path.
Wan 2.1
Strengths: a checked plan, credit or no-rate path.
Trade-offs: the full output envelope is not confirmed, generation modes are not confirmed, audio support is not confirmed, and pricing is not normalized to a per-second rate.
Choose Google Veo 3.1 when transparent exact API pricing, documented generation controls, documented audio output, a checked plan, credit or no-rate path, and its documented duration and resolution envelope matter most. Choose Wan 2.1 when a checked plan, credit or no-rate path matter most.
For output quality, use the paired vote split below; it is specific to the shared Prompts shown here. This summary compares confirmed documentation and pricing only; it does not infer visual quality or rank models.
Overall · 68 shared Prompts
no ties
- Google Veo 3.1
- 58%
- preferred on 46/68
- Wan 2.1
- 42%
- preferred on 22/68
Average share of weighted human votes across every shared Prompt, published by Rapidata. This is the dataset’s vote split, not a rating by this site.
Prompt 1 of 68
Human preference on this pairing
Share of weighted human votes for these two clips, collected by Rapidata. It compares this exact pair of clips — it is not an absolute score and not our rating.
Prompt
A 'day in the life' of a cutting-edge urban dance crew rehearsing for a global competition. Showcase their intense training, creative choreography sessions, high-tech stage setups, team dynamics, and the adrenaline-filled moments of their electrifying performance under the city lights.
All 68 Prompts in this matchup
List-price signal
Checked 2026-09-19
Exact and estimate rows are USD per output second. Rows marked Exact are published provider rates; rows marked Estimate are official examples calculated from token consumption, not fixed per-second plan prices. Plan, credit and token paths are grouped separately. A dash means no comparable checked price.
| Basis | Google Veo 3.1 | Wan 2.1 |
|---|---|---|
| USD / output second | ||
| Lite · 720pexact | $0.05 | No comparable checked price |
| Lite · 1080pexact | $0.08 | No comparable checked price |
| Fast · 720pexact | $0.10 | No comparable checked price |
| Fast · 1080pexact | $0.12 | No comparable checked price |
| Fast · 4Kexact | $0.30 | No comparable checked price |
| Standard · 720p / 1080pexact | $0.40 | No comparable checked price |
| Standard · 4Kexact | $0.60 | No comparable checked price |
| Other checked billing paths | ||
| Lite · 4K | Not offered | No comparable checked price |
| Model fee | No comparable checked price | $0 |
| CPU Basic | No comparable checked price | Free |
| Nvidia T4 small | No comparable checked price | $0.40 / hour |
| 1x Nvidia L4 | No comparable checked price | $0.80 / hour |
| 1x Nvidia A10G small | No comparable checked price | $1.00 / hour |
| Google Veo 3.1: Audio included. Re-verified 2026-09-28 against Google's pricing page: rates unchanged; billed only when a video actually generates. | ||
| Wan 2.1: Compute examples, not a generation price; actual cost depends on VRAM, throughput and hosting path. | ||
This is a price signal, not a quality ranking. Resolution, capabilities, input materials, failed generations, taxes, and regional account terms are not normalized.
- Google Gemini API pricing · checked 2026-09-28
- Hugging Face pricing · checked 2026-09-19
API comparison
7/7 and 1/7 fields confirmed from official sources
Version-level API facts, taken from the provider documentation cited below. Unconfirmed means we have not verified it from an official source — it does not mean the feature is unsupported.
| Field | Google Veo 3.1 | Wan 2.1 |
|---|---|---|
| Official model ID | veo-3.1-generate-preview · veo-3.1-fast-generate-preview · veo-3.1-lite-generate-preview | Unconfirmed |
| Duration | 4s / 6s / 8s (8s required for extension, reference images, 1080p and 4K) | Unconfirmed |
| Resolution | 720p (default) · 1080p (8s only) · 4K (8s only, not on Lite) · 24fps | Unconfirmed |
| Modes | T2V · I2V · Ingredients (up to 3 reference images, not on Lite) · First–last frame · Scene extension (~148s max, not on Lite) | Unconfirmed |
| Aspect ratios | 16:9 · 9:16 | Unconfirmed |
| Native audio | Included in video pricing | Unconfirmed |
| Weights license | Closed (proprietary) | Apache-2.0 |
Sources · checked 2026-09-19
- Google Gemini API pricing · checked 2026-09-28
- Gemini API Veo docs · checked 2026-09-28
- Wan 2.1 repository (Apache-2.0) · checked 2026-09-11
Where these clips come from
Both clips and the preference split are published by Rapidata under APACHE-2.0, retrieved 2026-09-11. Media is served directly from their Hugging Face repositories; this site re-hosts nothing and adds no rating of its own.
- Rapidata/text-2-video-human-preferences
- Rapidata/text-2-video-human-preferences-veo3.1
- Rapidata/text-2-video-human-preferences-veo3
- Rapidata/text-2-video-human-preferences-veo2
- Rapidata/text-2-video-human-preferences-seedance-1-pro
- Rapidata/text-2-video-human-preferences-kling-v2.1-master
- Rapidata/text-2-video-human-preferences-wan2.1
- Rapidata/text-2-video-human-preferences-pika2.2
- Rapidata/text-2-video-human-preferences-moonvalley-marey
- Rapidata/text-2-video-human-preferences-genmo-mochi-1
Where to run these models
This site does not generate video and does not sell generation credits, so it has no reason to prefer either model. To run them, use the provider’s own API or app.
Frequently asked questions
Did these two models run the same Prompts?
Yes. Google Veo 3.1 and Wan 2.1 were both run on the same 68 Prompts, with identical Prompt text sent to each model. Each slide on this page shows the two clips generated for one Prompt.
Which model did people prefer overall?
Across 68 Prompts, Google Veo 3.1 averaged 58% of the weighted human preference for each pair, and Wan 2.1 averaged 42%. These are vote shares published by Rapidata; they are not a rating by this site.
How many Prompts was each model preferred on?
Google Veo 3.1 was preferred on 46 of 68 Prompts and Wan 2.1 on 22, with no ties.
Can I generate these videos on this site?
No. This site does not generate video and does not sell generation credits; it compares models and links out to each provider. Running these models means using the provider's own API or app.