"Comfy 打造了一个全新而强大的创意生态,同时毫不牺牲创作掌控力。看着技术型艺术家和充满好奇的创意人借助 Comfy 探索创意的每一个维度,令人惊叹。"
Scott Belsky,
Behance 创始人
更新于 2026 年 10 月
首 30 天享 5 折优惠。这是 Fish Audio 迄今最富情感、最可控的模型。把多角色场景写成一段对话,再在同一个工作流中连同音频一起构建。
(8 秒) · 声音: Friendly Women (en)
[excited, hushed tones] Did you hear, Fish Drama 3 is in Comfy now?! [barely containing a squeal] I know, I know, keep your voice down. <whisper>It does whispering too.</whisper>
(18 秒) · 声音: Dramatic Character Male (en)
[deep, gravelly movie trailer voice] In a world... where every node sat in silence... [pause] one model dared to speak. [building to an epic shout] FISH! AUDIO! DRAMA THREE! [suddenly flat and casual] It's in Comfy now. You can just, like, use it.
(21 秒) · 声音: Energetic Male (en)
[groggy, just woke up, mumbling] Mmm, what time is it... [yawn] why is my phone blowing up. [reading slowly, confused] Fish Drama 3... is in Comfy? [pause] [wide awake, voice cracking with excitement] WAIT. It's in Comfy?! [talking very fast] Okay okay okay, nobody talk to me, I'm opening my laptop, I have forty ideas and no time. [laugh] <emphasis>This is the best morning of my life.</emphasis>
(34 秒) · 声音: @Voice1 Sarah (en), @Voice2 Adrian (en)
@Voice1 [leaning in, conspiratorial whisper] Okay, don't freak out. But I have news. @Voice2 [bored, not looking up] If this is about your sourdough starter again, I'm leaving. @Voice1 [giddy, speeding up] Better. Fish Drama 3 is in Comfy now. One node. Both of us. In the same take. @Voice2 [long pause, then slowly, suspicious] ...Wait. Are we in it right now? @Voice1 [delighted] We are literally in it right now! @Voice2 [gasp] [dramatic, theatrical despair] Then everything I have ever said was a prompt. [sigh] @Voice1 [laugh] [warm] Yeah. And you sounded great.
(44 秒) · 声音: Narrator (en)
[weary noir detective, low and slow] It was three in the morning. The queue was empty, and so was my coffee. [sigh] Then she walked into my workflow. A text to speech node, with a model I'd never seen before. [suspicious, quiet] Drama 3, the dropdown said. Fish Audio. [pause] <whisper>I'd heard the rumors.</whisper> Voices that could laugh. [dry chuckle] Voices that could cry. Voices that could change their mind halfway through a sentence. [voice rising, alarmed] I typed one line and hit Run, and the thing talked back like it meant it! [long exhale, calm again] I've worked a lot of cases in this town, kid. <emphasis>This one's in Comfy now.</emphasis> [soft, almost fond] And I'm not closing it.
步骤 01
打开 Comfy Cloud
登录或免费创建账号
步骤 02
选择 Drama 3
从合作伙伴节点中添加 Fish Audio Text to Speech 节点,并将 model 设为 drama-3
步骤 03
执导场景
添加声音,写下对白,运行即可
Fish Audio 的最新模型,也是 Fish Audio 称为迄今最富情感、最可控的一款。你写下多角色场景并给出表演指示,Drama 3 会把它作为一段对话完整演绎。它在 Fish Audio Text to Speech 节点中运行,支持 Comfy Cloud 和 0.39.3 及以上版本的本地 ComfyUI。
打开 Comfy Cloud,在模板库中搜索 Fish Audio Drama 3,或从合作伙伴节点中添加 Fish Audio Text to Speech 节点并将 model 设为 drama-3。添加声音,写下对白,运行即可。
Drama 3 在 Comfy Cloud 中前 30 天半价。打开 Comfy Cloud 即可开始。
Drama 3 在 Fish Audio Text to Speech 节点中运行,因此和其他节点一样融入 Comfy 工作流。在同一个工作流中构建场景和它的音频。改一句台词,再运行一次,下游的每个环节都会用上新的演绎。
不是。Drama 3 通过 Fish Audio API 运行,无需下载任何内容,也无需配置 GPU。
Text to Speech 还可运行 S2.1 Pro 和 S1,支持在文本中加入情绪提示。Speech to Text 自动检测语言并转写音频,Instant Voice Clone 用 1 到 20 段录音生成私有声音。Voice Selector 用于从 Fish Audio 声音库中挑选声音。这四个节点都收录在 Fish Audio 文档中。
可以,使用 Fish Audio Instant Voice Clone 节点:连接 1 到 20 段录音,总时长不超过 270 秒,每段以 10 到 30 秒为佳,克隆出的声音会直接接入 Fish Audio Text to Speech 节点,无需等待训练。声音克隆免费。请只克隆你有权使用的声音。
通过你的 ComfyUI 账号按使用量计费。文本转语音按文本长度计量,语音转文字按分钟计量。当前费率见 Fish Audio 文档。
通过简单透明、按使用量计费的方式,访问云端 ComfyUI 工作流。
首发优惠:Fish Audio Drama 3 前 30 天半价
在真实 GPU 上免费运行 5 次 — 无需信用卡。
按月计费
包含内容:
Comfy Agent
Comfy API
约可生成 380 个 5 秒视频*
按月计费
包含标准版的全部功能,另加:
Comfy Agent
Comfy API
约可生成 670 个 5 秒视频*
按月计费
包含创作者版的全部功能,另加:
Comfy Agent
Comfy API
约可生成 1,915 个 5 秒视频*
为协作开发工作流的团队打造。
按月计费
约可生成 13,405 个 5 秒视频*
包含专业版的全部功能,另加:
Comfy Agent
Comfy API
即将推出…
需要更多成员?想要更多灵活性或定制功能?
*基于默认设置(81 帧、18fps、640x640、4-step sampler)使用 Wan 2.2 图生视频模板生成 5 秒视频估算
你的 AI 助手可以接入整个生态、构建工作流,并生成图像、视频、音频或 3D 内容。
今天就在 Comfy Cloud 中运行 Drama 3,通过 API 批量生成场景,或在你自己机器上的 ComfyUI 中运行。