ElevenLabs(@elevenlabsio)
We’re releasing Scribe v2 Medical, our Speech to Text model fine-tuned for clinical audio. Scribe v...
7.0内容质量
TL;DR · AI 摘要
ElevenLabs发布Scribe v2 Medical，临床语音转录错误率降低35%。
核心要点
- Scribe v2 Medical在临床音频上降低35%词错误率(WER)
- 模型在普通语音测试中保持5.3% WER与基础模型一致
- 专门优化药物名称、解剖学术语等医疗领域专有名词识别
结构提纲
按章节快速跳转。
- §产品发布
ElevenLabs宣布推出针对临床场景优化的Scribe v2 Medical模型。
- ·性能提升
该模型在临床音频测试中将词错误率(WER)降低35%。
- ›技术验证
在6000条普通语音样本测试中保持与基础模型相同的5.3% WER。
- ·应用场景
专门优化医疗领域术语识别，包括药物名称和病理学词汇。
思维导图
用一张图看清主题之间的关系。
查看大纲文本（无障碍 / 无 JS 友好）
- Scribe v2 Medical
- 临床音频优化
- 35% WER降低
- 性能验证
- 普通语音5.3% WER
- 医疗场景应用
- 药物名称识别
- 病理学术语处理
金句 / Highlights
值得收藏与分享的关键句。
Scribe v2 Medical在临床音频上降低35%词错误率(WER)
在6000条普通语音样本测试中保持5.3% WER
临床语音转录错误可能导致患者记录错误
#Speech-to-Text#医疗AI#ElevenLabs#模型优化
打开原文ElevenLabs on X: "We’re releasing Scribe v2 Medical, our Speech to Text model fine-tuned for clinical audio. Scribe v2 Medical reduces word error rate (WER) on clinical audio by 35% compared to our base Scribe v2 model, improving recognition of medication names, anatomy, and pathology." / X
[](https://x.com/)
[](https://x.com/)
ElevenLabs on X: "We’re releasing Scribe v2 Medical, our Speech to Text model fine-tuned for clinical audio. Scribe v2 Medical reduces word error rate (WER) on clinical audio by 35% compared to our base Scribe v2 model, improving recognition of medication names, anatomy, and pathology."
-  ElevenLabs @ElevenLabs We’re releasing Scribe v2 Medical, our Speech to Text model fine-tuned for clinical audio. Scribe v2 Medical reduces word error rate (WER) on clinical audio by 35% compared to our base Scribe v2 model, improving recognition of medication names, anatomy, and pathology.  [](https://x.com/ElevenLabs/status/2102762665995325741/photo/1) 2:11 PM · Sep 23, 2026·12K Views 11 12 116 26
-  ElevenLabs @ElevenLabs Sep 23 Clinical speech is one of the hardest tests for transcription. Drug names are long and sound alike (hydroxyzine vs. hydralazine). Dosages, units, anatomy and pathology arrive in rapid sequence. The stakes are higher because a transcription error could end up in a patient’s Show more  [](https://x.com/ElevenLabs/status/2102762669636005931/photo/1) 2 2 8 830  ElevenLabs @ElevenLabs Sep 23 Scribe v2 Medical improves clinical transcription without regressing on ordinary speech. On 6,000 samples of everyday, non-medical speech derived from Common Voice, Scribe v2 Medical scores the same 5.3% WER as base Scribe v2.  [](https://x.com/ElevenLabs/status/2102762671850557818/photo/1) 1 3 326  ElevenLabs @ElevenLabs Sep 23 Scribe v2 Medical is HIPAA-eligible for enterprise customers with a BAA and Zero Retention Mode enabled, which ensures that audio and transcripts are deleted the moment each request completes.  [](https://x.com/ElevenLabs/status/2102762673964536000/photo/1) 1 6 1.5K  ElevenLabs @ElevenLabs Sep 23 Scribe v2 Medical is available now via ElevenAPI. Pricing starts from $0.22 per hour, decreasing with scale. From elevenlabs.io [](https://x.com/ElevenLabs/status/2102762675944194297) 4 1.4K
-  minime · fireply.ai @minimesoy Sep 23 35% wer drop just from a clinical fine tune, that's a real number not a marketing one [](https://x.com/minimesoy/status/2102763767109239169) 1 59
See all the replies
Log in or sign up for X
See what’s happening and join the conversation
Continue with phoneContinue with Apple
Continue with Google
通过 Google 继续操作 通过 Google 继续操作。在新标签页中打开
or