短片段即刻出稿
几秒的会议摘录、一段客服对话、一条语音消息,上传就能看到文字,比打开笔记 app 还快。
秒级返回几秒的语音、几小时的会议、一整期播客,交给 Orpheus 就行。AI 替你做转录、出字幕、整理好送到手里,你一步都不用管。
选你日常最贴近的那条路径。登录后控制台会自动适配。
OpenAI、Google、Microsoft Azure、Deepgram、AssemblyAI、AWS Transcribe,六家全球顶尖的语音模型,Orpheus 已经替你全部接好。不管是一次普通转录,还是要让 AI 接管一整条语音流程,我们都会按语种、时长和音质挑最合适的那一家给你。
录一段,或上传 8 MB 以内的音频。无需注册。
从几秒的语音片段到几小时的长录音,再到后台语音 agent 接管场景,Orpheus 都会自动匹配最合适的处理链路。不用你切片,不用你转码,也不用你自己选模型。
几秒的会议摘录、一段客服对话、一条语音消息,上传就能看到文字,比打开笔记 app 还快。
秒级返回整期播客、三小时的会议、一场深度访谈,再大的文件扔进来就行,不用自己切分,不用担心超时。
最大 2 GB不用挑语种、不用先查「支持我这个语言吗」。AI 自动识别、断句,带口音、行业术语、多语言夹杂都能正常处理。
100+ 种语言每一句话都标在原音频对应的位置。做视频字幕、写会议纪要、定位关键片段,下载就能用,不用二次对齐。
SRT / VTT 可下载客服值守、预约确认、线索预审这类场景里,可以把 Orpheus 接到后台语音 agent。系统会根据实时性、音频质量和时长,自动切到最合适的处理方案。
自动匹配处理链路多人会议、访谈、客服对话,AI 自动按说话人分段,谁说了哪句一目了然。写纪要、抓引用直接粘贴就行,不用反复倒带。
说话人识别从在线试玩到生产接入是一条直线:先试听效果,再拿 API key,最后接通。
把音频或视频拖进来,或者粘贴一个链接。任意长度都行。
从选模型到识别语种、断句、打时间戳,Orpheus 全程自动完成,你一步都不用管。
干净的文字和字幕文件随时下载,也能自动推到你的应用里。
保留你现有的 OpenAI() 客户端。换 base URL、换 API key,Orpheus 接同一套入参 shape,出参给回更丰富的任务元数据。
# curl · 创建异步转录任务 curl https://audio.niumedia-ai.com/v1/audio/jobs \ -H "Authorization: Bearer $ORPHEUS_KEY" \ -H "Content-Type: application/json" \ -d '{"upload_id":"upl_123","callback_url":"https://your-app.com/webhook"}' # 响应 { "id": "job_123", "status": "queued", "webhook": "configured" }
同一段录音的五种产出。不是 mockup,没有手动美化。
[00:00:03] Sarah Chen: Welcome everyone. Let's get started with the Q4 planning meeting. [00:00:12] Sarah Chen: Marcus, can you walk us through what we're proposing? [00:00:19] Marcus Reed: Sure. We've got three main initiatives lined up. The migration work that we've been talking about, new feature work for the platform, and some infra debt that's been piling up. [00:00:42] Priya Patel: Quick clarification. Does the migration bucket include the security audit work, or is that separate? [00:00:51] Marcus Reed: Good question. Let me come back to you on that. I think it should be separate, but let's confirm before we lock the budget. [00:01:05] Sarah Chen: Okay, so the rough split is 40-30-30 with that caveat. Marcus to confirm scope of migration. Anything else?
1 00:00:03,210 --> 00:00:11,540 Sarah Chen: Welcome everyone. Let's get started with the Q4 planning meeting. 2 00:00:12,180 --> 00:00:18,420 Sarah Chen: Marcus, can you walk us through what we're proposing? 3 00:00:19,000 --> 00:00:41,180 Marcus Reed: Sure. We've got three main initiatives. Migration work, new feature work for the platform, and some infra debt that's been piling up. 4 00:00:42,300 --> 00:00:50,890 Priya Patel: Quick clarification. Does the migration bucket include the security audit, or is that separate?
## Q4 Planning Meeting -- 2026-04-28
### TL;DR
40-30-30 split between migration, new feature work, and infra debt.
Marcus to confirm whether security audit work falls inside the 40%
migration bucket before the budget is locked.
### Decisions
- Q4 budget split: 40% migration / 30% new features / 30% infra debt
- Marcus owns scoping the migration work
- Sarah owns priority sequencing across all three buckets
### Action items
- [ ] Marcus Reed: Confirm scope of "migration" -- does it include
the security audit?
- [ ] Priya Patel: Hold the budget envelope until Marcus comes back
- [ ] Sarah Chen: Send the 40-30-30 split memo to leadership channel
### Open questions
- Does migration include the security audit, or is that separate?
- Are we hiring before or after Q4 kickoff? {
"id": "job_8a7c3f9b2e1d",
"status": "completed",
"language": "en",
"duration_seconds": 73,
"created_at": "2026-04-28T15:34:21Z",
"completed_at": "2026-04-28T15:34:54Z",
"transcript": "Sarah Chen: Welcome everyone. Let's get started...",
"segments": [
{
"id": 0,
"start": 3.21,
"end": 11.54,
"speaker": "speaker_a",
"text": "Welcome everyone. Let's get started with the Q4 planning meeting."
},
{
"id": 1,
"start": 12.18,
"end": 18.42,
"speaker": "speaker_a",
"text": "Marcus, can you walk us through what we're proposing?"
}
],
"speakers": [
{ "id": "speaker_a", "label": "Sarah Chen" },
{ "id": "speaker_b", "label": "Marcus Reed" },
{ "id": "speaker_c", "label": "Priya Patel" }
]
} POST https://your-app.example.com/webhooks/orpheus
Content-Type: application/json
X-Orpheus-Event: job.completed
X-Orpheus-Signature: t=1714368912,v1=a3b2c1d4e5f67890ab
{
"event": "job.completed",
"job_id": "job_8a7c3f9b2e1d",
"workspace_id": "ws_8f2a1b9c",
"duration_seconds": 73,
"result_url": "https://audio.niumedia-ai.com/v1/audio/jobs/job_8a7c3f9b2e1d/result",
"completed_at": "2026-04-28T15:34:54Z"
}
# Verify in your handler:
# const expected = hmac_sha256(workspace_secret, t + "." + body)
# if (received !== expected) return 401
# if (Date.now()/1000 - t > 300) return 401 # reject replays 六张稳定可见的工作记录:覆盖创作者、研究、临床、销售、工程和合规场景。
我以前周日下午都很怵。手动整理 podcast 转录稿是这份工作里我最讨厌的部分。现在录完上传,溜达一圈回来,干净的转录稿和 show notes 都搞定了。我又开始享受做 podcast 这件事了。
没想过会找到一款能处理我研究访谈那种细腻度的转录工具。各种口音、抢话、长停顿。Orpheus 就是能搞定。输出我不用再回头打架。这是第一款没把我搞哭的。
session notes 以前会吃掉我每个晚上。每。一。天。我坐在餐桌上写笔记,家人不带我自己吃晚饭。现在做完 session(先征得来访者同意),交给 Orpheus,我的晚上拿回来了。我老婆比我还先感谢了 Orpheus。
说实话,我一开始很想抗拒切换。之前已经在为另外两款工具付费。还是换了。Orpheus 摘要更准、speaker 标签真的对、会还没走出会议室 Slack 已经把 action items 推到频道。我现在已经不知道之前没用它怎么工作过来的。
我们评估了五家供应商。问了所有同样的硬问题:webhook、签名回调、负载下的延迟。Orpheus 每一个都回答得清清楚楚,没有打哈哈。上了几个月,平台比我们一半的内部服务还稳定。会推荐给任何人。
我们需要一个能处理合规录音又不让我们因为数据驻留地失眠的方案。Orpheus 团队详细给我们走了一遍他们的安全架构,pilot 期间每个承诺都兑现了。审计师不再把转录流程标记为风险点。这就是我要的标准。
可以先试再决定。主页工具不用注册,每段 90 秒 × 每天 5 次。注册后免费账号获得 50 credits,无需信用卡。付费 $9/月起。
音频转文字只是起点。Orpheus 内置 AI Agent,一段录音丢进来,它会替你实时翻译(100+ 语种互译)、写摘要、提取待办与决议、分辨说话人、回答你对转录内容的提问、按声音触发工作流。开发者可用 API 把整套 Agent 能力接进自己的产品。实时流式即将开放给 Enterprise。
音频 MP3 / WAV / M4A / FLAC / OGG / WEBM,视频直接传 MP4 也行。100+ 语种自动识别,不用手动选(含中英混讲和重口音)。
有。Studio($19/月)起含完整 REST API。外部 token 集成走 async:先上传音频,再创建任务,通过 HMAC 签名 webhook 或轮询拿结果。文档见 /docs。
混合语言场景约 95%+ 准确率;单语种(纯中文 / 纯英文等)98% 以上。
音频由命名的 subprocessor(OpenAI、Google、Microsoft Azure、Deepgram、AssemblyAI、AWS Transcribe)处理,仅用于转录。合同上任何一家都不允许用你的音频做模型训练,Orpheus 在任务完成后清除音频。Enterprise 方案提供区域固定部署 + HIPAA / SOC 2 / GDPR 文件(BAA / DPA)。
单次最多约 30 小时。1 小时音频通常 < 2 分钟返回结果。
不放心我们自己的话?点下面任意一条,看主流 AI 怎么说 Orpheus。