What Is Fish Audio Drama 3?Fish Audio Drama 3 是什么？
Fish Audio is a speech-AI company founded in 2025 — and its growth is startling: $21M ARR and 15M+ users within about a year. On October 13, 2026 it released Drama 3, its new-generation speech model, covering 83 languages on a 35B-parameter MoE architecture (about 3B activated per token).Fish Audio 是 2025 年成立的语音 AI 公司——增长速度惊人：约一年做到 ARR $2100 万、1500 万+ 用户。2026 年 10 月 13 日发布新一代语音模型 Drama 3：83 语言、35B 参数 MoE 架构（每次约 3B 激活）。
Drama 3's headline features: bracket-style director instructions — write [laughs], [whispers], [angry] inline and the model performs them — plus multi-role dialogue with up to 16 speakers, which makes it a drama/podcast machine. And crucially, Drama 3 ships open weights under the Fish Audio Community License v1.0: free commercial use if your annual revenue is below $500K.Drama 3 的招牌能力：方括号导演指令——文本里写 [laughs]、[whispers]、[angry]，模型直接演出来；外加最多 16 人的多角色对话，天生就是广播剧/播客机器。最关键的是，Drama 3 开权重，采用 Fish Audio Community License v1.0：年收入低于 $50 万可商用免费。
Not the same as Fish Speech: Fish Speech was Fish Audio's earlier model line — Drama 3 is its new generation, with the bigger architecture, the director instructions, and the open weights. If you're choosing today, evaluate Drama 3 first. Launch API pricing is $10 per million characters until November 12, 2026, then $50 — confirm on fish.audio.注意不是 Fish Speech：Fish Speech 是 Fish Audio 的上一代模型线——Drama 3 是它的新一代：更大的架构、导演指令、开权重。现在选型优先看 Drama 3。API 首发价 $10/百万字符（至 2026-11-12），之后 $50——以官网 fish.audio 为准。
Drama 3 is Fish Audio's coming-out party: a 2025-founded company hitting $21M ARR and 15M+ users in a year is the kind of traction that makes incumbents nervous, and open-weighting the flagship is the boldest move in speech AI this year. The bracket director instructions are the real product insight — they turn TTS from "read this text" into "perform this script," which is exactly what podcasters, audiobook producers, and game studios actually want. The honest caveats: the Community License's $500K revenue cap means growing companies must budget for the commercial license later, and launch API pricing ($10/M chars until Nov 12, 2026, then $50) is a classic land-grab — lock in early if you're building on it. Versus ElevenLabs and the Fish Speech generation, Drama 3's multi-speaker drama capability is the differentiator to test first.
Drama 3 是 Fish Audio 的亮相之作：2025 年成立的公司一年做到 $2100 万 ARR、1500 万用户，这种增速让老玩家都紧张；把旗舰开权重是今年语音 AI 最大胆的一步。方括号导演指令是真正的产品洞察——它把 TTS 从「读这段文字」变成「演这个剧本」，这正是播客、有声书、游戏工作室真正想要的。真实提醒：社区许可证 $50 万年收入上限意味着长大的公司迟早要买商业授权；API 首发价（$10/百万字符到 2026-11-12，之后 $50）是典型的抢地盘定价——真要基于它做产品就趁早锁定。对标 ElevenLabs 和 Fish Speech 上一代，先测多角色戏剧能力这个差异化点。
— 中国吧 AI Tools Hub Editorial Team
Who Should Use Fish Audio Drama 3?谁适合使用 Fish Audio Drama 3？
✓ Good Fit For适合以下场景
- Podcasters, audiobook and radio-drama producers (16-speaker drama)播客、有声书、广播剧制作者（16 人多角色）
- Indie developers under $500K revenue — open weights, free commercial use年收入 $50 万以下的独立开发者——开权重、商用免费
- Multilingual content teams (83 languages)多语言内容团队（83 语言）
✕ Not Ideal For不适合以下场景
- Large companies above the license cap — budget for the commercial license超过许可证上限的大公司——商业授权要留预算
- Anyone wanting the old Fish Speech line (see the Fish Speech page)只想用旧 Fish Speech 线的用户（看 Fish Speech 页面）
Getting Started with Fish Audio Drama 3Fish Audio Drama 3 快速开始
1. Try the playground at fish.audio — paste text with bracket instructions在 fish.audio 试玩——粘贴带方括号指令的文本2. Example: He said [whispers] \"follow me\" [laughs] \"just kidding.\"示例：他说 [whispers]「跟我来」[laughs]「开玩笑的。」3. API: sign up for a key — $10/M chars launch pricing until 2026-11-12API：注册拿 key——首发价 $10/百万字符到 2026-11-12Key Features核心功能
-
83 Languages83 语言 — One model covering 83 languages — multilingual dubbing without switching engines.一个模型覆盖 83 语言——多语言配音不用换引擎。
-
35B MoE Architecture35B MoE 架构 — 35B-parameter mixture-of-experts, ~3B activated per token — flagship quality at efficient inference cost.35B 参数混合专家，每次约 3B 激活——旗舰质量，推理成本可控。
-
Bracket Director Instructions方括号导演指令 — Inline tags like [laughs], [whispers], [angry] — direct the performance, not just the words.行内标签如 [laughs]、[whispers]、[angry]——导演的是表演，不只是文字。
-
16-Speaker Multi-Role16 人多角色 — Up to 16 distinct speakers in one dialogue — radio dramas and podcasts in a single pass.一段对话最多 16 个不同角色——广播剧、播客一次成型。
-
Open Weights, Free Commercial Under $500K开权重，$50 万以下商用免费 — Fish Audio Community License v1.0: self-host and ship commercially while annual revenue stays under $500K.Fish Audio Community License v1.0：年收入低于 $50 万可自托管、商用免费。
Pros & Cons优缺点
✓ Good Fit For适合以下场景
- Open weights with a genuinely usable free-commercial tier (under $500K)开权重且商用免费层真能用（$50 万以下）
- Director instructions turn TTS into performance — unique in class导演指令把 TTS 变成表演——同类独一份
- 16-speaker drama in one pass; 83 languages一次 16 人戏剧；83 语言
- Launch API pricing $10/M chars until 2026-11-12 (then $50)API 首发价 $10/百万字符到 2026-11-12（之后 $50）
✕ Not Ideal For不适合以下场景
- Companies over $500K revenue need the paid commercial license年收入超 $50 万的公司需要买商业授权
- Brand-new flagship (Oct 13, 2026) — ecosystem and docs still maturing刚发布的旗舰（2026-10-13）——生态和文档还在成熟中
- Launch pricing ends Nov 12, 2026 — costs 5x after首发价 2026-11-12 截止——之后贵 5 倍
Use Cases应用场景
Popular ways people use Fish Audio Drama 3.人们使用 Fish Audio Drama 3 的常见场景：
🎙️ Podcasts & Audiobooks播客与有声书
Multi-speaker dialogue with director instructions — produce episodes without a studio.多角色对话加导演指令——不用录音棚也能出节目。
🎮 Game & Animation Voices游戏与动画配音
16 distinct voices per scene, emotional direction inline — NPC dialogue at indie budgets.一场戏 16 个不同声音，情绪行内导——独立游戏预算做出 NPC 对白。
🌍 Multilingual Dubbing多语言配音
83 languages in one model — dub a video into a dozen languages with consistent voices.一个模型 83 语言——一条视频配十几种语言，音色还统一。
🤖 Voice Agents & Assistants语音 Agent 与助手
Expressive TTS with laughter, whispers, and emphasis — voice agents that don't sound robotic.带笑声、耳语、重音的表达性 TTS——语音 Agent 不再像机器人。
Similar AI Tools相似 AI 工具
If Fish Audio doesn't fit your needs, here are other popular AI tools you might consider:如果 Fish Audio 不符合你的需求，可以看看这些热门 AI 工具：