跳到主要内容Skip to main content
← Image

SeaArtSeaArt

海艺(成都海艺互娱)Haiyi (Chengdu Haiyi Huyu)
免费+订阅Free + subscription 社区模型库Community model library LoRA训练LoRA training 二次元向Anime-focused 模型聚合Model aggregation partial lifecycle: active 核验Checked 2026-10-22

SeaArt 是二次元/插画向的社区型聚合创作台:云端 LoRA 训练+导出、社区模型库(许可逐模型标注)、 多家第三方模型聚合;中国团队(成都海艺互娱)的出海产品。

SeaArt is the community-type anime/illustration aggregation console: cloud LoRA training with ComfyUI export, a community model library whose licenses are labeled per model, multi-vendor model aggregation; a going-overseas product by Chengdu-based Haiyi.

适合与不适合When to use it, when not to

适合:二次元插画生态、云端训练 LoRA 再导 ComfyUI、社区模型寻宝、一平台多模型。 不适合:要官方自研前沿模型、企业合规叙事优先、对社区内容治理敏感。

Fits: anime/illustration community ecosystem, cloud LoRA training then export to ComfyUI, community model treasure-hunting. Not a fit: needing official frontier self-developed models (it aggregates), enterprise compliance narratives.

中文生态China availability 国产 / 国内可直接用 / 备案状态 · 快照domestic / direct access / CAC filing · snapshot 2026-10-22

出品属地Origin
国产domestic 成都海艺互娱科技有限公司(中国)——主站面向海外的出海产品(官网口径:全球创作者社区)。Chengdu Haiyi Huyu Technology Co., Ltd. (China) — a going-overseas product whose main site targets global creators (per official positioning).
国内使用Direct access
大陆可直连direct OK seaart.ai 大陆直连实测 200(2026-10-22)。seaart.ai measured 200 direct from mainland (2026-10-22).
备案状态CAC filing
备案未核验filing unknown:未核验到 seaart.ai 主站对应的生成式 AI 服务备案(2026-10-22 检索);主站面向海外,未核验到≠未备案——公司另有国内线产品,其备案情况不在本档案口径内。no generative-AI service filing found for the seaart.ai main site (searched 2026-10-22); the main site targets overseas users — "not found" ≠ "not filed"; the company's domestic-line products are outside this profile's scope.

备案信息以网信办《生成式人工智能服务已备案信息》公示名单为准(按批次滚动更新,本站只抄编号不解读);「大陆可直连」是本站从大陆网络直连实测的结果,不是厂商承诺。GitHub 类站点当日实测有波动,口径写在依据里。Filing status is quoted from the CAC public registry (updated batch by batch; this site copies the number and does not interpret it). "Direct access" is this site's own reachability measurement from a mainland network, not a vendor promise. GitHub-hosted sites fluctuated on the measurement day — the note says so.

固定坐标系Fixed coordinate system

8 个维度dimensions, 与同分区其他对象逐项可比each comparable item by item against others in the same section.

模型与开放条件Model & openness

多模型聚合平台:官网列第三方模型清单(Seedance/Wan/GPT Image/Seedream/Flux/Qwen Image 等, 官网导航与隐私页所见)+ 社区模型库(checkpoint/LoRA 托管,模型页标 Base Model 与许可范围); 自研模型线未核验到官方声明。

Multi-model aggregation platform: the official site lists third-party models (Seedance/Wan/GPT Image/Seedream/Flux/Qwen Image etc., as seen in site navigation and the privacy page) + a community model library (checkpoint/LoRA hosting; model pages mark Base Model and license scope); an in-house model line — no official statement verified.

运行位置Where it runs

网页(seaart.ai / seaart.me)+ App(官网有 APP 入口);生成在厂商云端。 社区模型可下载(模型页有 Download 面)——下载后本地跑走本地生态,不是 SeaArt 提供本地运行时。

Web (seaart.ai / seaart.me) + app (an APP entry on the official site); generation on the vendor's cloud. Community models can be downloaded (model pages have a Download surface) — running them locally after download goes through the local ecosystem; SeaArt does not provide a local runtime.

本地文件Local files

作品与模型存账户侧云端;社区模型可下载至本地(模型页 Download 面)。未见官方声明的 本机数据面——下载后的模型文件归本地文件系统,与平台无关。

Works and models stored account-side in the cloud; community models downloadable to local (Download surface on model pages). No officially stated local data surface — downloaded model files belong to the local filesystem, unrelated to the platform.

关机后的任务Tasks after shutdown

生成即时/排队(高峰期排队是社区常见反馈,档位差异未逐条核验);无「关机续跑」任务面 (未核验到声明);LoRA 训练提交后在云端跑。

Generation is instant/queued (queuing at peak hours is a common community feedback; tier differences not verified item by item); no "keep-running-after-shutdown" task surface (no claim verified); LoRA training runs on the cloud after submission.

工具与扩展Tools & extensions

在线 LoRA 训练(官方页:上传数据集选基础模型 FLUX/SDXL/Pony/Wan 等,可导出对接 ComfyUI)、 ControlNet/inpainting、图像+视频双线(官网导航所见)、社区画廊(作品带提示词分享)、 模型市场变现面(官方 FAQ 口径)。

Online LoRA training (official page: upload a dataset, pick a base model FLUX/SDXL/Pony/Wan etc., exportable to interoperate with ComfyUI), ControlNet/inpainting, image+video dual lines (as seen in site navigation), a community gallery (works shared with prompts), a model-marketplace monetization surface (official FAQ framing).

上下文与记忆Context & memory

以作品/模型页组织;跨会话记忆未核验到官方声明。社区模型库是它的「一致性」机制——风格模型可复用不是记忆。

Organized by works/model pages; cross-session memory — no official statement verified. The community model library is its "consistency" mechanism — reusable style models are not memory.

权限与限制Permissions & limits

账号+订阅体系(D9:价格不采);商用权逐模型标注(模型页 License Scope 面所见); 训练用途开关未逐条核验。主体是中国团队出海产品(海外用户占比高,官网主站面向海外)。

Account + subscription system (D9: prices not collected); commercial rights labeled per model (License Scope surface on model pages); the training-use switch not verified line by line. The entity is a Chinese team's going-overseas product (high share of overseas users; the main site targets overseas).

适合什么任务Best for

适合:二次元/插画向社区生态、云端 LoRA 训练+导出 ComfyUI、一个平台聚合多家模型、下载社区模型。 不适合:要官方自研前沿模型(它是聚合方)、要企业级合规叙事、对社区内容治理敏感的场景。

Fits: anime/illustration community ecosystem, cloud LoRA training + export to ComfyUI, aggregating multiple vendors' models on one platform, downloading community models. Does not fit: wanting official in-house frontier models (it is an aggregator), enterprise-grade compliance narratives, scenarios sensitive to community content moderation.

头号误解Top misconception

  • 社区模型的商用许可逐模型不同——模型页的 License Scope 才是依据,别按「平台会员=全部可商用」理解Community-model commercial licenses differ per model — the model page's License Scope is the basis; 'platform membership = everything commercial' is wrong
  • 「可下载模型」≠「SeaArt 在本地跑」——下载后走的是本地生态(ComfyUI 等),平台本身是云端服务'Downloadable models' ≠ 'SeaArt runs locally' — downloads feed the local ecosystem (ComfyUI etc.); the platform itself is a cloud service
  • 中国团队出海产品:主站 seaart.ai 面向海外,大陆可达性与备案口径要分开看(见 cn 块)Chinese team going overseas: the main site targets global users — mainland reachability and filing status must be read separately (see the cn block)

未知项清单Known unknowns

  • 自研模型线(未核验到官方声明)
  • 训练用途开关与默认态(未逐条核验)
  • 主站与国内线(海艺 AI)账号/数据是否互通(未核验)
  • 跨会话记忆机制(未核验到官方声明)
  • The in-house model line (no official statement verified)
  • The training-use switch and its default state (not verified line by line)
  • Whether the main site and the domestic line (Haiyi AI) share accounts/data (unverified)
  • Cross-session memory mechanism (no official statement verified)

证据来源Evidence sources

判断可回到以下一手源复核。本站核验日Every judgement can be rechecked against the primary sources below. Verified on 2026-10-22,内容更新日content updated on 2026-10-22。

类型Type名称Name链接Link
officialSeaArt · 官网(产品与能力叙事、模型清单)SeaArt · official site (product and capability narrative, model list)https://www.seaart.ai
officialSeaArt · Terms(官方条款,2026-10-22 直抓核验)SeaArt · Terms (official terms, direct-fetched and verified 2026-10-22)https://seaart.ai/en/terms
officialSeaArt · Privacy(官方隐私页,2026-10-22 直抓核验)SeaArt · Privacy (official privacy page, direct-fetched and verified 2026-10-22)https://seaart.ai/en/privacy
changelogSeaArt · Blog(产品更新线,changelog)SeaArt · Blog (product update line, changelog)https://seaart.ai/blog

实测记录Measured runs

本档未跑生成质量与速度测试;结论只覆盖能力面与许可面。

实测建议(三条,都指向官方可读页没给的东西):

  • 在线 LoRA 训练产物导出 ComfyUI 的兼容性(格式/触发词是否无损)
  • 社区模型库的下载许可面与生成许可面是否一致
  • 聚合各家模型的排队与质量波动

This site has not completed measured runs. The coding-oriented test protocol is at protocols/coding/_protocol.md (the coding task suite does not apply to creative products).

Suggested measurements (three, all pointing at things the official readable pages do not give):

  • Compatibility when exporting online-LoRA-training artifacts into ComfyUI (whether format/trigger words are lossless)
  • Whether the community model library's download-license surface matches its generation-license surface
  • Queuing and quality variance across the aggregated vendors' models

相关条目Related entries

本页由This page is generated from the ai-handbook 数据层生成(CC BY 4.0)。data layer (CC BY 4.0). 方法论与坐标系定义见仓库内For the methodology and coordinate definitions, see METHODOLOGY.md.