Wan vs Veo
Wan and Veo are both current Top 10 AI 视频生成器. Wan has the stronger current scoring signal in this dataset, while Veo may still win for specific access, workflow, pricing, or open-source needs.
快速建议
Most ordinary users should start with Wan when they want the safer overall pick from the current evidence. Pick Veo instead when its winning scenarios below match your workflow better.
选择 Wan 的情况
- price / cost原因: Wan has the clearer cost/access advantage from open weights or confirmed price facts.证据: github.com, wan.video, gemini.google.com
- access / availability原因: Wan has broader visible access evidence across web, API, aggregator, or open-weight routes.证据: wan.video, gemini.google.com, labs.google
- open / closed source原因: Wan publishes open weights or open-source artifacts, which is a clear advantage for self-hosting, research, and reproducible evaluation.证据: github.com, deepmind.google
选择 Veo 的情况
- camera control原因: Veo has stronger camera control evidence from product features, workflow tags, or access data.证据: github.com, deepmind.google
- native audio原因: Veo has stronger native audio evidence from product features, workflow tags, or access data.证据: github.com, ai.google.dev
- API / developer availability原因: Veo has stronger API availability evidence from product features, workflow tags, or access data.证据: github.com, ai.google.dev
- version maturity原因: Veo has the newer tracked version event in the local version dataset.证据: artificialanalysis.ai, krea.ai, docs.dev.runwayml.com
购买前仍需确认的证据
Wan 概览
Alibaba's open and commercial video model family, with strong community adoption and variants for text, image, speech, and animation-driven video.
公司: Alibaba Group
模型 page: Wan details
当前分数： 87
适合场景
Veo 概览
Google DeepMind's flagship video generation family for cinematic clips, image-referenced video, vertical output, native audio, and API workflows.
公司: Google DeepMind
模型 page: Veo details
当前分数： 86
适合场景
优势总结
Every dimension below names an advantage, a reason, evidence status, and the user type it matters for. 证据 pending means the page needs live pricing, leaderboard, benchmark, or product-term verification before naming a winner.
Wan 的优势
- price / cost原因: Wan has the clearer cost/access advantage from open weights or confirmed price facts.证据: github.com, wan.video, gemini.google.comBest for: Budget-sensitive teams and API buyers
- access / availability原因: Wan has broader visible access evidence across web, API, aggregator, or open-weight routes.证据: wan.video, gemini.google.com, labs.googleBest for: Ordinary users who need the product to work today
- open / closed source原因: Wan publishes open weights or open-source artifacts, which is a clear advantage for self-hosting, research, and reproducible evaluation.证据: github.com, deepmind.googleBest for: Researchers, self-hosters, and teams needing model control
Veo 的优势
- camera control原因: Veo has stronger camera control evidence from product features, workflow tags, or access data.证据: github.com, deepmind.googleBest for: Directors who need repeatable camera language
- native audio原因: Veo has stronger native audio evidence from product features, workflow tags, or access data.证据: github.com, ai.google.devBest for: Dialogue, sound effects, music, and audio-video sync
- API / developer availability原因: Veo has stronger API availability evidence from product features, workflow tags, or access data.证据: github.com, ai.google.devBest for: 开发者s, automation builders, and product teams
- version maturity原因: Veo has the newer tracked version event in the local version dataset.证据: artificialanalysis.ai, krea.ai, docs.dev.runwayml.comBest for: Teams choosing stable current products
持平或视情况而定
- feature depth原因: Feature coverage is close in the current profiles, so choose by the specific workflow rather than raw feature count.证据: github.com, arxiv.org, deepmind.googleBest for: Creators choosing by workflow breadth instead / one demo
- generation quality原因: Fallback scores are close, so current use case and benchmark evidence matter more than the raw score.证据 pending: needs live pricing, benchmark, or product documentation source.Best for: Teams optimizing final visual quality
No public source link is attached yet.
- speed / delivery原因: Speed evidence pending; measure current queue time and generation latency before naming a winner.证据 pending: needs live pricing, benchmark, or product documentation source.Best for: Users who need faster iteration or production throughput
No public source link is attached yet.
- motion realism原因: Benchmark evidence pending; use current leaderboard or controlled tests.证据: artificialanalysis.aiBest for: Ads, action shots, camera moves, and physical scenes
- prompt adherence原因: Wan and Veo both have usable evidence here; choose based on workflow fit and current access.证据: github.com, deepmind.googleBest for: Prompt-heavy storyboards and precise scene requests
- subject consistency原因: Both products have comparable subject consistency evidence or the difference depends on workflow depth.证据: github.com, deepmind.googleBest for: Character, product, and object continuity work
- reference control原因: Both products have comparable reference control evidence or the difference depends on workflow depth.证据: github.com, deepmind.google, ai.google.devBest for: Brand/product references and character consistency
- editing / extension原因: Both products have comparable editing or extension workflow evidence or the difference depends on workflow depth.证据: github.com, deepmind.googleBest for: Iterative editors and production teams
Advantage By 维度
This is the decision layer: not just yes/no support, but who leads, why they lead, what evidence backs it, and who should care.
| 维度 | 优势 / 胜出方 | 原因 | 证据 | 适合用途 |
|---|---|---|---|---|
| price / cost | Wan advantage | 原因: Wan has the clearer cost/access advantage from open weights or confirmed price facts. | 证据: github.com, wan.video, gemini.google.com | Budget-sensitive teams and API buyers |
| access / availability | Wan advantage | 原因: Wan has broader visible access evidence across web, API, aggregator, or open-weight routes. | 证据: wan.video, gemini.google.com, labs.google | Ordinary users who need the product to work today |
| feature depth | 持平或视情况而定 | 原因: Feature coverage is close in the current profiles, so choose by the specific workflow rather than raw feature count. | 证据: github.com, arxiv.org, deepmind.google | Creators choosing by workflow breadth instead / one demo |
| generation quality | 持平或视情况而定 | 原因: Fallback scores are close, so current use case and benchmark evidence matter more than the raw score. | 证据 pending: needs live pricing, benchmark, or product documentation source. No public source link is attached yet. |
Teams optimizing final visual quality |
| speed / delivery | 证据 pending | 原因: Speed evidence pending; measure current queue time and generation latency before naming a winner. | 证据 pending: needs live pricing, benchmark, or product documentation source. No public source link is attached yet. |
Users who need faster iteration or production throughput |
| motion realism | 证据 pending | 原因: Benchmark evidence pending; use current leaderboard or controlled tests. | 证据: artificialanalysis.ai | Ads, action shots, camera moves, and physical scenes |
| prompt adherence | 持平或视情况而定 | 原因: Wan and Veo both have usable evidence here; choose based on workflow fit and current access. | 证据: github.com, deepmind.google | Prompt-heavy storyboards and precise scene requests |
| subject consistency | 持平或视情况而定 | 原因: Both products have comparable subject consistency evidence or the difference depends on workflow depth. | 证据: github.com, deepmind.google | Character, product, and object continuity work |
| camera control | Veo advantage | 原因: Veo has stronger camera control evidence from product features, workflow tags, or access data. | 证据: github.com, deepmind.google | Directors who need repeatable camera language |
| reference control | 持平或视情况而定 | 原因: Both products have comparable reference control evidence or the difference depends on workflow depth. | 证据: github.com, deepmind.google, ai.google.dev | Brand/product references and character consistency |
| native audio | Veo advantage | 原因: Veo has stronger native audio evidence from product features, workflow tags, or access data. | 证据: github.com, ai.google.dev | Dialogue, sound effects, music, and audio-video sync |
| editing / extension | 持平或视情况而定 | 原因: Both products have comparable editing or extension workflow evidence or the difference depends on workflow depth. | 证据: github.com, deepmind.google | Iterative editors and production teams |
| API / developer availability | Veo advantage | 原因: Veo has stronger API availability evidence from product features, workflow tags, or access data. | 证据: github.com, ai.google.dev | 开发者s, automation builders, and product teams |
| commercial use | 证据 pending | 原因: Commercial use depends on current product terms, regional plans, and license language; verify live terms before recommending a winner. | 证据: github.com, deepmind.google | Brands, agencies, and revenue-generating projects |
| open / closed source | Wan advantage | 原因: Wan publishes open weights or open-source artifacts, which is a clear advantage for self-hosting, research, and reproducible evaluation. | 证据: github.com, deepmind.google | Researchers, self-hosters, and teams needing model control |
| version maturity | Veo advantage | 原因: Veo has the newer tracked version event in the local version dataset. | 证据: artificialanalysis.ai, krea.ai, docs.dev.runwayml.com | Teams choosing stable current products |
| external ratings / leaderboard | 持平或视情况而定 | 原因: Fallback scores are close, so current use case and benchmark evidence matter more than the raw score. | 证据 pending: needs live pricing, benchmark, or product documentation source. No public source link is attached yet. |
Buyers who want market validation |
| SEO / popularity / web authority | 证据 pending | 原因: SEO, popularity, and authority metrics need richer imports before naming a winner. | 证据 pending: needs live pricing, benchmark, or product documentation source. No public source link is attached yet. |
Teams using adoption and discoverability as risk signals |
| open-source GitHub metrics | 证据 pending | 原因: GitHub metrics only apply when a model has open-source or open-weight repositories; neither side has enough repo evidence here. | 证据 pending: needs live pricing, benchmark, or product documentation source. No public source link is attached yet. |
Open-model evaluators and infra teams |
| company background | 持平或视情况而定 | 原因: Both companies have usable profile evidence; brand strength alone should not decide the product choice. | 证据: home.alibabagroup.com, deepmind.google, github.com | Users who care about product longevity, distribution, and support risk |
Feature 维度 Comparison
Supported values are source-backed where product detail records include source URLs; inferred values are marked.
| 维度 | Wan | Veo |
|---|---|---|
| text-to-video | Supported | Supported |
| image-to-video | Supported | Supported |
| video-to-video | Partial Animate/replacement workflows use video inputs; general V2V coverage depends on wrapper. | Supported Veo 3.1 supports video input for extension in Gemini API docs. |
| reference control | Supported | Supported |
| native audio | Partial Speech-to-video supports audio-driven generation; general native generated audio for T2V needs separate confirmation. | Supported |
| lip sync | Supported S2V audio-driven generation covers speech/video synchronization workflows. | Partial Native audio is confirmed; standalone lip-sync controls need feature-level confirmation. |
| camera control | Partial | Supported |
| editing/extension | Partial | Partial |
| API | Partial Self-hosting and hosted wrappers exist; Alibaba Cloud commercial endpoints should be normalized separately. | Supported |
| commercial license | Provisional Check Apache-2.0 license before commercial use. No public source link is attached yet. | Provisional Commercial use depends on current hosted product terms. No public source link is attached yet. |
| open/closed source | [Object Object] Wan2.2 publishes code and model weights with Apache-2.0 repository licensing shown by GitHub repository metadata. | [Object Object] Veo weights are not open. |
Wan 价格 / 入口
Open weights can be downloaded, but practical cost depends on GPU hardware or hosted inference. Commercial Alibaba/partner endpoints should be priced separately.
If exact public pricing is not listed here, treat it as provisional and check the linked source before publishing a cost claim.
访问渠道
价格事实
- Self-hosting costHardware-dependent; S2V examples note high VRAM requirements.
Veo 价格 / 入口
Gemini API pricing lists Veo 3 at USD 0.40 per second, Veo 3 Fast at USD 0.15 per second, and Veo 2 at USD 0.35 per second; check page for Veo 3.1 pricing before display.
If exact public pricing is not listed here, treat it as provisional and check the linked source before publishing a cost claim.
访问渠道
价格事实
- Veo 3 APIUSD 0.40 per second on paid tier
- Veo 3 Fast APIUSD 0.15 per second on paid tier
Performance / Quality 证据
Rows say source-backed when comparison JSON provides a cited claim; otherwise they are provisional.
| Quality 维度 | Wan | Veo |
|---|---|---|
| motion realism | source-backed high Wan has a current evidence-weighted score / 87.4. Treat this as source-aware provisional unless refreshed against live Artificial Analysis, Arena, and product tests. (high) | source-backed high Veo has a current evidence-weighted score / 86.4. Treat this as source-aware provisional unless refreshed against live Artificial Analysis, Arena, and product tests. (high) |
| prompt adherence | source-backed high Wan has a current evidence-weighted score / 87.4. Treat this as source-aware provisional unless refreshed against live Artificial Analysis, Arena, and product tests. (high) | source-backed high Veo has a current evidence-weighted score / 86.4. Treat this as source-aware provisional unless refreshed against live Artificial Analysis, Arena, and product tests. (high) |
| temporal consistency | source-backed high Wan has a current evidence-weighted score / 87.4. Treat this as source-aware provisional unless refreshed against live Artificial Analysis, Arena, and product tests. (high) | source-backed high Veo has a current evidence-weighted score / 86.4. Treat this as source-aware provisional unless refreshed against live Artificial Analysis, Arena, and product tests. (high) |
| subject consistency | source-backed high Wan has a current evidence-weighted score / 87.4. Treat this as source-aware provisional unless refreshed against live Artificial Analysis, Arena, and product tests. (high) | source-backed high Veo has a current evidence-weighted score / 86.4. Treat this as source-aware provisional unless refreshed against live Artificial Analysis, Arena, and product tests. (high) |
| speed | source-backed high Wan has a current evidence-weighted score / 87.4. Treat this as source-aware provisional unless refreshed against live Artificial Analysis, Arena, and product tests. (high) | source-backed high Veo has a current evidence-weighted score / 86.4. Treat this as source-aware provisional unless refreshed against live Artificial Analysis, Arena, and product tests. (high) |
| resolution/duration/public benchmark evidence | source-backed high Wan has a current evidence-weighted score / 87.4. Treat this as source-aware provisional unless refreshed against live Artificial Analysis, Arena, and product tests. (high) | source-backed high Veo has a current evidence-weighted score / 86.4. Treat this as source-aware provisional unless refreshed against live Artificial Analysis, Arena, and product tests. (high) |
版本进展
Latest and historical versions are pulled from product profiles plus version-events JSON.
Wan versions
- Wan 2.5 / 2.6 / 2.7Names appear in third-party catalogues and should not be treated as confirmed official releases yet.
- Wan2.2Released inference code and model weights. Supports T2V, I2V, and TI2V. Integrated with Diffusers and ComfyUI.
- Wan technical reportPublished the Wan open large-scale video generative model paper line.
Veo versions
- Veo negative prompt supportRunway API adds optional negativePrompt for veo3, veo3.1, and veo3.1_fast text-to-video and image-to-video requests.
- Veo model cardsOfficial DeepMind model-card index for Veo safety/version updates.
- Veo 3.1 Lite model cardGoogle DeepMind published the Veo 3.1 Lite model card for the text/image-to-video system with audio.
- Veo 3.1 in Google Vids free monthly accessGoogle announced that personal Google accounts get monthly no-cost Veo 3.1 video generations in Google Vids, with higher limits for Google AI Pro/Ultra and Workspace AI Ultra accounts.
- Veo 3.1Improves Ingredients to Video. Adds native vertical outputs. Adds 1080p and 4K upscaling options.
- Veo 3.1 Preview / Fast PreviewGemini API model codes include veo-3.1-generate-preview and veo-3.1-fast-generate-preview. Supports text, image, last-frame, reference-image, and video-extension inputs depending on model.
- Veo 3 PreviewGemini API changelog lists video with audio generation for Veo 3 preview.
- VeoVeo announced at Google I/O 2024 as Google DeepMind's generative video model.
外部评分
Fallback scores, SEO/web authority signals, GitHub status, and preset editorial winners are shown with caveats.
| Signal | Wan | Veo |
|---|---|---|
| Fallback score | 87 Wan appears in Artificial Analysis and Arena snapshots, and the open repository/paper provide reproducible public evidence.; professional score is provisional/evidence pending unless backed by scoring JSON. | 86 Google Gemini Omni Flash/Veo-family entries lead the Artificial Analysis audio-video snapshot and rank highly in Arena/Vercel snapshots.; professional score is provisional/evidence pending unless backed by scoring JSON. |
| SEO / web authority | Reachability collected; third-party authority metrics pending. | Reachability collected; third-party authority metrics pending. |
| GitHub / open source | Wan2.2 publishes code and model weights with Apache-2.0 repository licensing shown by GitHub repository metadata. | Veo weights are not open. |
| Comparison preset | Open 权重 Benchmarking; 开发者 工作流; Image To Video; Open 模型 Ecosystem; MoE Video Diffusion Newer Wan 2.5/2.6/2.7 names need official release-page confirmation.; Open checkpoints can require high-memory GPUs. | Cinematic Generation; Image Referenced Clips; Vertical Short Form Video; Ingredients To Video; Native Vertical Output 模型 availability and feature access vary by country, product surface, and plan.; Some Google products may expose only selected Veo versions. |
到底该选哪个？
- Wan best-fit workflows: Wanopen-weight benchmarking, developer workflows, image-to-video
- Veo best-fit workflows: Veocinematic generation, image-referenced clips, vertical short-form video
- Score-backed shortlist pick: WanScores are close (87.4 vs 86.4), so run a prompt-level test for the exact use case.
来源
本页面使用的官方来源、新闻/版本来源、外部排名或权威来源，以及开源仓库链接。