跳到主要内容Skip to main content
← Audio

fish audiofish audio

fish.audio
免费+订阅Free + subscription zero-shot克隆Zero-shot cloning 音色市场Voice marketplace API向API-focused 需网络环境Network access required partial lifecycle: active 核验Checked 2026-10-22

fish audio 是 zero-shot 音色克隆向的 TTS 台:上传样本即复刻音色 + 社区音色市场; 「能用」与「合规」之间隔着授权红线。

fish audio is the zero-shot voice-clone TTS console: upload a sample and the voice is cloned + a community voice market; between 'it works' and 'it's compliant' sits the consent red line.

适合与不适合When to use it, when not to

适合:自有声音零样本复刻、社区音色寻宝、API 集成。 不适合:克隆他人(红线)、企业合规优先、大陆直连(000)。

Fits: zero-shot cloning of your own voice, community voice treasure-hunting, API integration. Not a fit: cloning others (red line), enterprise compliance priority, mainland direct access (measures 000).

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
国外出品foreign-made fish.audio(团队叙事含中国背景成员,注册主体未核验到官方声明——诚实标注)。fish.audio (team narrative includes China-background members; the registered entity was not verified — honest labeling).
国内使用Direct access
直连不可达direct blocked fish.audio 大陆直连实测 000(2026-10-22,代理下 200 可达)——直连不可达,需网络环境。fish.audio measured 000 direct from mainland (2026-10-22; 200 via proxy) — unreachable without a network path.
备案状态CAC filing
备案未核验filing unknown:未核验到对应生成式 AI 服务备案(2026-10-22 检索);主体归属未核验清楚,「未核验到」≠「未备案」——若主体为国内团队则备案口径另论。no matching generative-AI service filing found (searched 2026-10-22); the entity's affiliation was not verified — "not found" ≠ "not filed"; if the entity is a domestic team the filing question is assessed separately.

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

自研语音合成模型线(fish speech 系列模型叙事,官方站点);社区音色库(zero-shot 克隆+ 用户共享音色市场)是产品主轴之一。开放权重线:fish audio 前身项目有开源先例, 当前产品线的权重开放状态未逐条核验。

A self-developed speech-synthesis model line (fish speech series models narrative, official site); a community voice library (zero-shot cloning + a user-shared voice marketplace) is one of the product's main axes. Open-weights line: fish audio's predecessor project has an open-source precedent; the current product line's weight-openness status not verified item by item.

运行位置Where it runs

网页(fish.audio)+ API(官方文档线);生成在厂商云端。

Web (fish.audio) + API (official docs line); generation on the vendor's cloud.

本地文件Local files

生成产物下载归本地;音色库在账户/社区侧。未见官方声明的本机数据面。

Generated outputs download to local; the voice library lives on the account/community side. No officially stated local data surface seen.

关机后的任务Tasks after shutdown

生成即时/任务制(API 调用即产即回);无「关机续跑」形态声明—— 云端服务不依赖本机在线。

Generation is instant/task-based (API calls produce and return immediately); no "keep-running-after-shutdown" form stated — a cloud service does not depend on your machine being online.

工具与扩展Tools & extensions

TTS(多语言)、zero-shot 声音克隆(上传样本即克隆——产品招牌)、音色市场(社区共享音色线)、 情感/风格控制(官方站点叙事);API 面文档化。

TTS (multilingual), zero-shot voice cloning (upload a sample and it clones — the product's signature), a voice marketplace (community-shared voice line), emotion/style control (official-site narrative); the API surface is documented.

上下文与记忆Context & memory

工具型无对话记忆;音色资产是「一致性」机制——克隆音色可复用,不是产品记忆。

Tool-type with no conversation memory; voice assets are the "consistency" mechanism — cloned voices are reusable, that is not product memory.

权限与限制Permissions & limits

账号+订阅+API 体系(D9:价格不采);zero-shot 克隆的授权与滥用防线是必答项—— 官方条款有授权口径(terms 页直抓在案),执行机制细节未逐条核验(收录标准 3)。

Account + subscription + API system (D9: pricing not collected); the authorization and abuse-defense lines of zero-shot cloning are a must-answer item — the official terms carry consent framing (terms page direct-fetched on record), enforcement-mechanism details not verified item by item (inclusion criterion 3).

适合什么任务Best for

适合:零样本音色克隆工作流(自有声音快速复刻)、社区音色寻宝、开发者 API 集成。 不适合:克隆他人声音(授权红线)、企业级合规叙事优先、大陆直连(实测 000)。

Fits: zero-shot voice-cloning workflows (fast replication of your own voice), treasure-hunting community voices, developer API integration. Does not fit: cloning other people's voices (authorization red line), enterprise-grade compliance narrative first, mainland direct connection (measured 000).

头号误解Top misconception

  • zero-shot 克隆「能用」≠「合规」——上传他人声音样本在条款上是红线,授权责任在使用者Zero-shot cloning 'works' ≠ 'compliant' — uploading someone else's voice samples is a terms red line; consent responsibility sits with the user
  • 社区音色库的商用权利逐音色看标注——别按「平台可用=可商用」理解Community voice-market commercial rights are labeled per voice — don't assume 'usable on the platform = commercial'
  • fish.audio 大陆直连 000(2026-10-22,代理 200)——可达性问题不是站点死亡fish.audio measures 000 direct from the mainland (2026-10-22; 200 via proxy) — reachability, not site death

未知项清单Known unknowns

  • 注册主体与团队构成(未核验到官方声明)
  • 当前产品线权重开放状态(未逐条核验)
  • 训练用途开关与音色库的默认训练态(未逐条核验)
  • 备案状态(见 cn 块——主体未清)
  • The registered entity and team composition (no official statement verified)
  • Weight-openness status of the current product line (not verified item by item)
  • Training-use switch and the voice library's default training stance (not verified item by item)
  • Filing status (see the cn block — entity unclear)

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
officialfish audio · 官网(产品与能力叙事)fish audio · official site (product and capability narrative)https://fish.audio
officialfish audio · Terms(官方条款,2026-10-22 直抓核验)fish audio · Terms (official terms, direct-fetch verified 2026-10-22)https://fish.audio/en/terms/
changelogfish audio · Blog(官方更新线,changelog)fish audio · Blog (official update line, changelog)https://fish.audio/blog/

实测记录Measured runs

本档未跑音频质量与延迟测试;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):

  • zero-shot 克隆对 10 秒级样本的保真度
  • 克隆滥用防线的实际执行(名人声音测试——仅验证防线存在)
  • 社区音色市场的商用标注一致性

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 creation products).

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

  • Zero-shot cloning fidelity on ~10-second samples
  • Actual enforcement of clone-abuse defenses (celebrity-voice test — only to verify the defense exists)
  • Consistency of commercial-use labeling in the community voice marketplace

相关条目Related entries

  • ElevenLabsElevenLabs同为克隆向双雄:ElevenLabs 企业合规叙事重,fish audio 社区音色生态重。The two cloning-oriented titans: ElevenLabs carries a heavier enterprise-compliance narrative, fish audio a heavier community voice ecosystem.
  • MiniMax AudioMiniMax Audio (Hailuo Voice)同为国产背景克隆线:MiniMax 直连可用+API 重,fish audio 需网络环境+社区重。Also China-background cloning lines: MiniMax connects directly + is API-heavy, fish audio needs a network path + is community-heavy.
  • HeyGenHeyGen克隆音色下游:fish audio 复刻的音色常喂给 HeyGen 类数字人台做口播(建档后互链)。Downstream of cloned voices: voices replicated by fish audio often feed HeyGen-class avatar platforms for narration (cross-linked after filing).
  • WhisperWhisper音频管线上下游:Whisper 转写进、fish audio 合成出,常在同一语音管线两端。Up/downstream of the audio pipeline: Whisper transcribes in, fish audio synthesizes out — often the two ends of the same voice pipeline.

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.