Translate Audio File to Text — Any Language, 30 Min Free

Text only. VexaScribe transcribes your audio in the source language (Whisper Large-v3), then translates the text to your target language. Output is TXT/DOCX/SRT — no synthesized voice. For voice-cloned dubbed audio, see ElevenLabs, HeyGen, or Rask AI.

Three steps: (1) Upload your audio file (MP3, WAV, M4A, OPUS, FLAC, or other common formats — up to 5 GB / 10 hours), (2) Select source language (or auto-detect) and target language, (3) Download the translated text in TXT, DOCX, or SRT format. Total processing time: ~5-10 minutes for a 1-hour file.

Yes. Upload the MP3, select source and target languages, download translated text. Works for MP3, WAV, M4A, OPUS, OGG, FLAC, AAC. See our MP3 to text page (/mp3-to-text) for MP3-specific details including podcast and dictaphone workflows.

Free for the first 30 minutes of source audio, no signup required. Beyond that, plans start at $2/month equivalent. Fixed monthly pricing rather than per-minute — better for consistent volume than per-minute APIs.

Whisper Large-v3 transcription: 93-95% accuracy on clean English audio, 85-92% on other Tier-1 languages (Spanish, French, German, Portuguese, Italian, Chinese, Japanese, Russian, Arabic). LLM translation is production-ready for creator content, internal training, e-learning, and personal use. For legal, medical, broadcast, or contract-critical content, human review is recommended.

An audio translator takes an audio file as input. A transcript translator takes existing text as input. If you already have a transcript in the source language, use ChatGPT/Claude directly or our SRT translator (/srt-translator) for subtitle files. If you have the audio file itself, use this page.

Text only — plus subtitle files (SRT, VTT). VexaScribe outputs the translated text in TXT, DOCX, SRT, VTT, and JSON formats. We do NOT produce voice-cloned dubbed audio in the target language. For voice-cloned dubbing (translated speech in the original speaker's voice, or a synthetic voice), use specialized tools like Maestra, ElevenLabs, or Vozo — different product category with different pricing and ethics considerations. Choose our tool when you need translated text or subtitles; choose a dubbing tool when you need translated spoken audio.

Nine common audio formats: MP3, M4A, WAV, OPUS, OGG, FLAC, AAC, AIFF, WMA. Plus common video formats (MP4, MOV, WebM) — we extract the audio track automatically. Files up to 5 GB and 10 hours per upload. If your file is in an unusual format, most modern audio tools can export to MP3 or WAV in seconds. Handles all Spanish, French, German, Italian, Portuguese, Japanese, Mandarin, Arabic, Hindi, Turkish, and 89 other source languages that Whisper Large-v3 supports.

Two stages. (1) Speech-to-text: Whisper Large-v3 transcribes the source-language audio (93-95% accuracy on clean Tier 1 languages per FLEURS benchmark; 88-92% Tier 2). (2) Text-to-text translation: the source transcript is translated to the target language. General-content translation lands around 90% BLEU on major language pairs. The two stages run automatically; you get the translated output plus (optionally) the original-language transcript in a bilingual side-by-side view for verification.

99 source languages (any language Whisper Large-v3 supports) × 133 target languages (translation stack). Common pairs — English ↔ Spanish, French, German, Italian, Portuguese (BR and PT), Dutch, Japanese, Mandarin, Korean, Russian, Polish, Turkish, Arabic (MSA and regional variants), Hindi, Bengali, Vietnamese, Thai, Indonesian — all work with strong accuracy. Lower-resource pairs (Swahili ↔ Tamil, Welsh ↔ Bengali) work but expect quality drops on both stages.

Free for 30 minutes, then billed by minute after that. VexaScribe: 30 min free (no card), then $2-20/month ($0.01-$0.003 per minute effective). Competitors: HappyScribe 10 min free then €15-72/month; Maestra 10 min free then per-minute; Rev human-review $1.99/min. For a one-off audio file under 30 minutes, the free trial covers it end-to-end. For ongoing translation workflows (multilingual podcasts, international meeting archives), the $5/month Basic tier at 1,000 minutes handles most single-creator use.

Two accuracy floors. Transcription (source language): 93-95% word accuracy on clean audio in Tier 1 languages (English, Spanish, French, German, Italian, Portuguese, Dutch, Japanese, Mandarin, Korean, Russian, Polish). 88-92% Tier 2 (Arabic, Turkish, Hindi, Vietnamese, Thai, Indonesian). Translation (source → target): ~90% BLEU on general content for common language pairs. Where accuracy drops: dense domain jargon (medical, legal, technical), heavy idioms, poor source audio, uncommon language pairs. Budget 5-10 minutes of manual review per audio hour for publication-quality output.

Yes — every job produces both. VexaScribe shows a bilingual side-by-side view (source language on the left, target translation on the right) in the browser editor. Download options: source transcript alone (TXT/DOCX), target translation alone (TXT/DOCX), bilingual side-by-side (DOCX), timing-synced SRT/VTT for both languages (great for shipping a video with dual caption tracks), JSON with both texts and per-word timestamps. Useful for verification, correcting errors, or shipping a bilingual video accessibility experience.

Yes — upload the video file (MP4, MOV, WebM) and we extract the audio automatically. The translated SRT output is timed to the original video's timeline, so you can attach it to the video as a subtitle track without re-syncing. For creators translating YouTube videos, TikTok clips, Instagram Reels, or podcast episodes: this is the workflow. For voice-cloned dubbed video (translated speech in place of the original audio track), you'll need a dedicated dubbing tool afterward.