免费在线音频转文字
使用AI驱动的转录将音频转为文字。上传音频文件、通过麦克风录音或粘贴URL。支持100多种语言、10多种模型，准确率超过98%。
1. 上传音频
上传MP3、WAV、M4A、FLAC、OGG或任何音频格式。最大2GB。
2. AI处理音频
AI从您的音频中提取语音，包含说话人检测和时间戳。
3. 获取转录文本
查看、编辑、下载或分享。导出为TXT、SRT、VTT、DOCX或PDF。
音频转文字模型
选择适合您需求的AI模型——或让我们为您选择最佳模型。
准备好将音频转为文字了吗？
免费开始 →常见问题
Upload your audio file or paste a URL, pick an AI model, and click Transcribe. STT.ai returns editable text with timestamps and speaker labels — most files finish in under five minutes.
MP3, WAV, M4A, FLAC, OGG, AAC, AMR, and 10+ more are all supported. You don't need to convert between formats first — upload whatever your recorder or app produces.
A little. Lossless formats like WAV and FLAC carry bit-perfect audio, so accuracy is bounded only by the model and speaker clarity. Lossy formats (MP3, M4A) at 128 kbps or higher are effectively identical; very low bitrates under 64 kbps can cost a few points.
Yes. STT.ai includes 600 free minutes per month with no signup for your first file. Paid plans starting at $5/month add longer files, private transcripts, and priority processing.
On clean audio our best models reach 95-97% accuracy (3-5% Word Error Rate). Background noise, overlapping speakers, and strong accents are the main factors that lower accuracy.
Yes. Free users can transcribe up to one hour per file; paid plans extend that to 8+ hours, which covers full-length podcasts, interviews, and audiobooks in a single pass.
Yes. Speaker diarization labels each voice (Speaker 1, Speaker 2, ...) and you can rename them in the editor — works on every supported audio format and model.
Export to TXT, DOCX, PDF, JSON, or SRT/VTT subtitles. JSON keeps machine-readable timestamps and speaker labels; DOCX and PDF are best for sharing and archiving.
Yes. 100+ languages with auto-detection, plus the option to set the language manually. Mixed-language audio is handled by switching mid-file, and you can translate the result afterwards.
Yes. Audio is processed and deleted by default, and Pro plans add client-side encryption so transcripts are unreadable without your key. Nothing is used for training without explicit opt-in.
Yes. Paste a link from any of 1,300+ supported platforms — podcast hosts, SoundCloud, YouTube, and more — and STT.ai fetches the audio directly. DRM-protected sources can't be transcribed.
Yes. The REST API accepts audio files directly, with Python and Node.js SDKs and a free tier of 100 minutes/month. Per-second billing applies beyond the free tier.