This tool produces translated TEXT and subtitle files (SRT/VTT). For voice-cloned dubbed audio, see Maestra, ElevenLabs, or Vozo — different product category.
Audio Translator
Upload an audio file in any language and get it back as a translated transcript — text and subtitles, not a dubbed voice.
Drop your file here
or click to browse
Audio: MP3 · WAV · M4A · FLAC · OGG · AAC · AIFF · WMA · AMR · OPUS
Video: MP4 · MOV · AVI · MKV · WebM · FLV · WMV
Up to 200 MB · 5 min free preview
Up to 200 MB · first 5 minutes translated free
Free preview — the first five minutes of your file, transcribed and translated, no account. Output is text and subtitles, never a dubbed voice. URL paste works for YouTube, TikTok and Instagram; for anything else, upload the file.
Looking for something more specific?
- Translating a video file (MP4, MOV) rather than audio? → /video-translator
- Translating an existing .srt subtitle file while preserving timings? → /srt-translator
- Specifically translating Spanish audio to English? → /translate-spanish-audio-to-english
- Want a bilingual side-by-side transcript (source + target)? → /transcribe-and-translate-audio
Can Google Translate do this?
This is the most common question on this topic, and the answer is narrower than people expect. Google Translate's audio feature is live conversation. You hold up a phone, someone speaks, it translates aloud. It is good at that. What it does not do is take an audio file — there is no upload box for an MP3 of a meeting you recorded last week.
| If you want to… | Use |
|---|---|
| Talk to someone in front of you, right now | Google Translate's conversation mode. It is built for exactly that and it is free. |
| Translate a recording you already have | This page. Upload the file and get a transcript in the target language. |
| Keep a written record you can search or quote | This page — live translation is spoken and leaves nothing behind. |
| Subtitle a recording with timings | This page. SRT and VTT export with the original cue times. |
Worth saying plainly: our free translation tier routes through Google's translation engine, so on a short clip in a common language the wording you get from us and from them will often be similar. What differs is the shape of the job — we take the file, handle length, keep timings, and hand back something you can edit and export.
Translate any audio file format
| Format | Source usage | Translation supported |
|---|---|---|
| MP3 | Podcasts, dictaphones | ✅ Direct |
| WAV | Studio, professional recording | ✅ Direct |
| M4A | iPhone Voice Memos, iTunes | ✅ Direct |
| OPUS | WhatsApp voice notes | ✅ Direct |
| OGG | Discord, Signal | ✅ Direct |
| FLAC | Lossless audio | ✅ Direct |
| AAC | Streaming, YouTube extract | ✅ Direct |
| AIFF | Apple pro audio | ✅ Direct |
For format-specific detailed workflows, see MP3 to text or M4A to text.
Popular language pairs — but honest
VexaScribe supports 99 source languages transcribed to 133 target languages. Common language pairs users search for:
- Spanish → English: dedicated page (1,600 monthly searches, our best-fit language pair with KD=2)
- French → English: 720 monthly searches — supported via this tool
- German → English: 390 monthly searches — supported
- Chinese → English: 170 — supported
- Japanese → English: 170 — supported
- Russian → English: 110 — supported
- Arabic → English: 90 — supported
- Italian, Korean, Portuguese, Hindi → English: all supported
- English → Spanish, French, German, Chinese, Japanese: supported
For "translate [language] audio" queries where the search intent is "get spoken audio in another language," the SERP is dominated by AI voice dubbing tools (ElevenLabs, Maestra, HeyGen). VexaScribe produces text output only — see our comparison in the vs dubbing section below.
How well is your language supported?
Coverage tracks how much recorded speech a language has in the training data, which is why the groups below are uneven. We do not publish an accuracy percentage for any of them — the honest signal is the grouping, not a number we cannot source.
Best supported
English, Spanish, French, German, Italian, Portuguese (BR + PT), Dutch, Japanese, Mandarin (Simplified + Traditional), Korean, Russian, Polish
Well supported
Arabic (MSA + regional), Turkish, Hindi, Vietnamese, Thai, Indonesian, Ukrainian, Czech, Hungarian, Romanian, Greek, Hebrew, Persian
Supported — check the output
Swahili, Bengali, Tamil, Punjabi, Welsh, Gaelic, Basque, Bulgarian, Croatian, Serbian, Slovak, and 60+ more
These work, but they are where a review pass is most worth the time. Run the free preview on your own file before committing to a long recording.
What this tool does — and what it doesn't
Setting expectations honestly before you upload. If you need something in the "does not" column, another product category (voice-dubbing SaaS) is a better fit.
| ✅ This tool does | ❌ This tool does NOT |
|---|---|
| Translate audio to TEXT in 133 target languages | Produce voice-cloned dubbed audio |
| Output SRT and VTT subtitle files synced to the original audio | Sync English speech to the original speaker's voice |
| Transcribe 99 source languages with Whisper Large-v3 Turbo | Do real-time voice translation (conversation mode) |
| Handle upload files up to 5 GB | Translate live phone calls or streaming audio |
| Show bilingual side-by-side (source + target) | Replace the original audio track in the exported video |
If you wanted a dubbed voice track
“Audio translator” is used for two different products, and this is the one that produces text. If what you need is the audio itself re-spoken in another language, that is AI dubbing — a separate category, and not something we do or plan to. A translated transcript from here is still a usable input script for a dubbing tool, which is a reason to run the free preview either way.
How to translate an audio file (3 steps)
- 1
Upload the audio or video
MP3, M4A, WAV, OPUS, OGG, FLAC, MP4, MOV, WebM. Up to 5 GB. Source language auto-detects.
- 2
Pick the target language
133 target languages — English, Spanish, French, German, Italian, Portuguese, Japanese, Mandarin, Arabic, Hindi, plus 123 more.
- 3
Download translated text or SRT
TXT, DOCX, SRT, VTT, JSON, or bilingual side-by-side DOCX. All formats included on every plan.
5 output formats
| Format | Best for |
|---|---|
| TXT | Copy-paste translated text into email, article, notes |
| DOCX | Editable Word doc, bilingual side-by-side option |
| SRT | Video subtitle track synced to original audio |
| VTT | HTML5 <track> element, browser web video |
| JSON | Developers — per-word timestamps + metadata |
Pricing compared (July 2026)
| Tool | Free tier | Paid | Notes |
|---|---|---|---|
| VexaScribe | 30 min (no card) | $2-20/month | Text + SRT output, EU hosting |
| HappyScribe | 10 min | €15-72/month AI | Text side; also has dubbing tier |
| Maestra | 10 min | Per-minute | Text + dubbing hybrid; focuses on video translation |
| Rev (human) | None | $1.99/min | Highest text accuracy, 12-24h turnaround |
| Google Translate | Live only | N/A | No file upload; conversation mode only |
What one file actually costs
Transcription and translation draw on the same pool of minutes, so the useful number is what a finished file costs end to end. A 60-minute interview runs to roughly 9,000 words: 60 minutes to transcribe, plus about 11 more if you translate on the premium engine. Call it 71 minutes. Free translation adds nothing on top.
| Plan | Minutes | Hour-long files a month | Each |
|---|---|---|---|
| Starter — $2 | 200 | 2 | $1.00 |
| Basic — $5 | 1,000 | 14 | $0.36 |
| Pro — $10 | 2,500 | 35 | $0.29 |
| Studio — $20 | 6,000 | 84 | $0.24 |
Translating a file into a second language does not re-charge you for the transcription, and a language already generated on the premium engine stays generated — returning to it later costs nothing.
Who this is for
Journalist / researcher
Foreign-language source interview → translated transcript for quoting or coding.
International business team
Multilingual meeting → English minutes for HQ. Bilingual DOCX for the record.
Podcast creator
Non-English podcast → English SRT for YouTube. Reach English-speaking audience without re-recording.
Privacy
- EU hosting — AWS eu-west-2 (London).
- Deletion is immediate when you ask for it — files stay in your library until you delete them; deletion removes the audio, transcript, translations and summaries within seconds, from every storage layer, with no soft-delete tombstone.
- Not used for training — Your files are not used to train AI models. Speechmatics and pyannoteAI, which handle premium transcription and speaker labelling, are contractually prohibited from training on customer data. OpenAI and Replicate, which handle summarization, premium translation and standard transcription, do not train on API inputs under their default terms. One exception worth knowing: free translation routes through Google's public translation endpoint, which is not covered by an enterprise data-handling agreement — use premium translation if that matters for your material.
- Not a certified translation service — for court, medical, or legally binding translation, use a certified translator.
See our full privacy policy.
Frequently asked questions
Does audio translation produce voice or text?
Text only. VexaScribe transcribes your audio in the source language (Whisper Large-v3 Turbo), 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.
How do I translate an audio file to text in another language?
Three steps: (1) Upload your audio file (MP3, WAV, M4A, OPUS, FLAC, or other common formats — up to 5 GB), (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.
Can I translate MP3 files to text in another language?
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.
How much does it cost to translate audio file to text?
The first five minutes of one file are free without an account. Signing up raises that to 30 minutes free with no card. Beyond that, plans start at $2/month equivalent. Fixed monthly pricing rather than per-minute — better for consistent volume than per-minute APIs.
Is machine audio translation accurate for professional use?
Whisper Large-v3 Turbo transcription: 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.
What's the difference between an audio translator and a transcript translator?
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.
Does this tool produce dubbed audio or text?
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.
What audio file formats does the translator accept?
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 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 Turbo supports.
How does the translation actually work?
Two stages. (1) Speech-to-text: Whisper Large-v3 Turbo transcribes the source-language audio ( accuracy on clean best-supported languages per FLEURS benchmark; well-supported). (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.
How many source and target languages are supported?
99 source languages (any language Whisper Large-v3 Turbo 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.
How much does it cost to translate an audio file?
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.
How accurate is the translation?
Two accuracy floors. Transcription (source language): word accuracy on clean audio in best-supported languages (English, Spanish, French, German, Italian, Portuguese, Dutch, Japanese, Mandarin, Korean, Russian, Polish). well-supported (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.
Can I get both the original transcript and the translation?
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.
Is this the right tool for translating a video?
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.