Transcribe and Translate Audio
One upload, both languages: an accurate transcript in the original language plus a translated version, with timestamps and speaker labels kept in sync. Paste a YouTube, TikTok or Instagram URL, or upload a file — the translated SRT drops straight into an editing timeline.
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, transcribed and translated, no account. Upload a file or paste a YouTube, TikTok or direct audio URL. Output is text and subtitles, not a dubbed voice track.
VexaScribe does both halves of the job in one pass: it transcribes audio in 99 languages, then translates that transcript into the language you need. Upload a Spanish podcast, export an English transcript and English subtitles. Record a German interview, get French, Japanese and Portuguese versions for international publication. Timestamps and speaker labels survive the translation, which is what makes the translated SRT drop straight into an editing timeline. Translation runs on one of two engines. The free one is Google Translate, included with any account and fine for understanding what was said. The paid one reads each line as speech rather than text: proper names stay names instead of being translated into words — the Turkish surname Yıldız comes back as “star” from a general-purpose translator — idioms become natural equivalents, and the speaker's register is preserved. It also has no language list at all, so it will attempt languages Google does not cover. An hour-long interview costs about eleven minutes of your plan, and re-translating that same file later costs nothing.
Paste a URL — no download step
New for July 2026: paste a YouTube, TikTok, Instagram Reel, or direct audio/video URL — no downloading, no re-uploading. Perfect for translating source video without touching the file. Feed a Spanish YouTube URL, get an English transcript and an English SRT for your international upload. Same minute cost as a file upload.
Also works with Google Drive share links and any HTTPS URL that directly serves an audio or video file (podcast episode MP3, Dropbox share, S3 link) — see all URL sources.
Two AI Models, One Workflow
Most tools stop after transcription. We continue through translation — automatically.
You never copy-paste the transcript into Google Translate, never sign up for a translation API, never manage two accounts. The whole workflow happens automatically when you click “Translate to [language]” in the editor.
Common Workflows
Real use cases where transcription + translation in one tool saves hours per week.
Spanish podcast → English subtitles
Reach English-speaking audiences without recording twice. One source episode, two language versions.
German interview → 4 European languages
Transcribe a German interview once, publish French, Italian, Portuguese, and Dutch versions.
Japanese course → English materials
Open Japanese course content to global learners. Transcribed lecture notes, translated study guides.
Multilingual sales calls → English summary
Customers speak their language; sales reps read transcripts in English. No interpreter needed.
Documentary footage → unified script
Editors working across language barriers. Transcribe each language, translate to a common language for editing.
Conference talk → 8 languages
One recorded talk, eight translated transcripts. Make a single conference session globally accessible.
99 Languages for Transcription, 133 for Translation
Audio is transcribed in 99 languages with automatic detection. Transcripts can be translated into 133 target languages.
Transcription languages
+ 73 more including Welsh, Swahili, Filipino, Bengali, Punjabi
Translation languages
All 99 transcription languages plus 34 additional translation-only languages:
Powered by Google Translate. Translation quality is highest for major Western European, East Asian, and Arabic pairs.
Two translation engines, and when each one is right
Translation runs on one of two engines. The free one is fine for understanding what was said. The paid one is for when the wording itself matters — and it costs about a third of a dollar for an hour-long interview, so the decision is rarely about money.
| Standard — free | Premium | |
|---|---|---|
| What it is | Google Translate — the same free engine behind translate.google.com, not a paid enterprise integration. | Each line sent through a language model prompted for spoken conversation rather than written text. |
| Proper names | Translated like ordinary words. The Turkish surname Yıldız comes back as “star”. | Names, places, dates and numbers are kept verbatim. Yıldız stays Yıldız. |
| Idioms | Rendered literally, which reads stiffly in the target language. | Turned into the natural equivalent rather than word-for-word. |
| Ambiguous words | Takes the most common reading. Italian fratelli becomes “brothers”. | Read in context — fratelli becomes “siblings” unless the context is specifically male. |
| Real speech | Stumbles and repetitions pass straight through: “the, the, ah so…” | Minor disfluencies smoothed, without flattening how the person actually spoke. |
| Languages | Around 133 — whatever Google Translate currently supports. | No fixed list. Name the language and it will attempt it, including ones Google does not cover at all. |
| Cost | Included free with any account. | One minute of your plan per 5,000 characters — about 11 minutes for a 60-minute interview. |
Languages Google does not have
This is the one place the gap is absolute rather than a matter of degree. Premium has no language allow-list, so it will attempt Cherokee, Aramaic, Church Slavonic, Latin, or a named regional variety like Levantine Arabic or Rioplatense Spanish. Quality on very low-resource languages is noticeably weaker than on Spanish or French — but Standard returns nothing for them at all.
How Standard fails
Worth knowing so you can recognise it. Standard sends text in batches of up to 4,000 characters. If a batch fails, those segments keep their original untranslated text rather than the whole file failing. So an occasional untranslated passage in a long free translation is expected behaviour, not a broken file.
What neither engine does
Both translate one line at a time. Neither builds a glossary and carries it through the file, so a term translated one way in minute two is not guaranteed to be translated identically in minute forty. Premium is more likely to get each line right, but it is not reading the surrounding conversation when it does. For anything you are going to publish, check proper nouns and recurring technical terms across the whole transcript — that is the one review pass worth doing, and neither engine removes the need for it.
What you get back
Both languages, kept in sync. The translation inherits the timings and speaker labels from the original, so every line stays tied to the moment it was said — which is what makes a translated SRT drop straight into an editing timeline, and what lets you check a translated quote against the audio in a couple of seconds.
Original-language transcript
Full transcript in the language the audio was recorded in. Export to TXT, DOCX, SRT, VTT, or JSON.
Translated transcript
Same export formats, in your target language. Pick one or more — generate English, French, Spanish from a single source.
Side-by-side editor
View original and translation in two columns. Edit either version. Useful for verifying translation accuracy or polishing wording.
Translated subtitles
Word-level timestamps preserved through translation. Upload directly to YouTube as a translated caption track. SRT and VTT supported.
Cross-language AI Chat
Ask the transcript questions in your working language and get answers in it, with original-language quotes preserved. A Turkish meeting transcript can be queried in English; a Spanish podcast queried in Japanese. Citations link back to exact moments in the original audio. Available on paid plans.
A transcript is not just a document
Once the text exists in both languages it stays queryable. You can summarise it in either language, or ask it questions directly — “what did she say about the deadline?” — and get an answer with the timestamp attached, so you can hear the original line rather than trusting the translation of it. That last part matters more on translated material than on anything else: when a quote is load-bearing, being able to jump to the audio is how you verify it.
What it costs to translate a real file
Transcription and translation draw on the same pool of minutes, so the useful number is what one finished file costs end to end. Take a 60-minute interview — roughly 9,000 words, or 54,000 characters. Transcribing it uses 60 minutes; translating it on Premium uses 11 more. Call it 71 minutes for the finished bilingual transcript.
| Plan | Minutes | Hour-long interviews a month | Each |
|---|---|---|---|
| Starter — $2/mo | 200 | 2 | $1.00 |
| Basic — $5/mo | 1,000 | 14 | $0.36 |
| Pro — $10/mo | 2,500 | 35 | $0.29 |
| Studio — $20/mo | 6,000 | 84 | $0.24 |
The comparison that matters
A professional translator working from that same transcript, at ten cents a word — the low end of the range — would charge around $900for the file. That is not an argument that the two are interchangeable: a human reviewer catches things neither engine will. It is an argument about what to spend the review budget on. Machine-translate first, then pay a person to check the parts that matter.
You only pay once per language
Premium-translate a file into Spanish today and the result is stored against it. Coming back next month for that Spanish version costs nothing — no minutes, no second charge. It also means you can run Standard first, look at it, and upgrade that same file to Premium without paying for the transcription again.
YouTuber workflow: from URL to translated SRT
Content creators publishing to international audiences need SRT subtitle files in each target language — not just plain-text translations. Paste the URL and the whole workflow happens in one pass — no download, no re-upload, no rebuilding timestamps by hand.
The 4-step workflow
- Paste the YouTube URL — your own video, or a source video you have rights to use. Instagram Reel, TikTok, and direct URLs work too.
- Whisper Large-v3 transcribes in the original language with word-level timestamps. A 10-minute video takes ~1-2 minutes.
- Choose target languages — around 133 on the free engine, or name any language at all on Premium. English → Spanish, Portuguese, French, German, Japanese, Korean, and Arabic is a common set for global YouTubers.
- Export SRT per language — each translated SRT preserves the original timestamps, ready to upload to YouTube's subtitle editor as “Language: Spanish (subtitles from another language, with timing)”.
Why this beats YouTube's auto-translate
- ■ Control over the source text: YouTube captions its video automatically and you translate whatever it produced. Here the transcript is yours to correct first, which matters most on accented speech and proper nouns — see how ASR accuracy is measured
- ■ Edit before publishing: YouTube auto-translate publishes verbatim errors; VexaScribe's editor lets you fix mistakes before export
- ■ Speaker labels persist through translation (multi-host podcasts, interviews)
Honest limitations
- ■ Well-resourced pairs (English ↔ Spanish, French, German, Italian, Portuguese) read far more naturally than long-tail ones like Khmer, Amharic or Welsh
- ■ Highly technical or slang-heavy content benefits from human polish
- ■ Cultural jokes, puns, wordplay don't translate — edit these manually per target language
Translation Included on Every Paid Plan
No upcharge, no separate translation account, no usage-based fees. Just unlimited translation across 133 languages.
Starter
200 min/month
Translation included
Basic
1,000 min/month
Translation included
Pro
2,500 min/month
Translation included
Frequently Asked Questions
How does VexaScribe transcribe and translate audio in one step?
VexaScribe (formerly NovaScribe) combines two AI models in one workflow. Step 1: OpenAI's Whisper Large-v3 transcribes your audio into text in the original spoken language (99 languages supported, with automatic detection). Step 2: Google Translate translates the resulting transcript into your target language (133 languages available). Both happen automatically — you upload once, get both versions in your dashboard within 5-10 minutes for a one-hour file.
How many languages does VexaScribe support for translation?
Free translations run through Google Translate, which currently covers around 133 languages — including all 99 we transcribe plus 34 additional translation-only languages — Welsh, Bengali, Punjabi, Tamil, Urdu, Albanian, Armenian, Esperanto, Hawaiian, Khmer, Lao, Latin, Macedonian, and many more.
Is translation included free with the transcription plan?
Yes. Translation is included on every paid plan at no extra cost — Starter ($2/month), Basic ($5/month), Pro ($10/month), and Studio ($20/month) all include unlimited translation. The 30-minute free trial also includes translation. There's no separate translation account, no API key, no usage-based fees.
Can I translate to and from any language pair?
Yes. You can translate from any of the 99 transcription languages. The free engine covers around 133 targets; Premium has no fixed list and will attempt any language you name, including ones Google Translate does not cover. Common pairs (Spanish to English, English to French, German to Italian) work very well because Google Translate has been trained on these for over 15 years. Less common pairs (Khmer to Welsh, for example) have lower fluency.
What's the translation quality like for less common languages?
Translation quality depends on the language pair and the content type. Major Western European languages (Spanish, French, German, Italian, Portuguese) translate to and from English at very high quality. Asian languages (Japanese, Korean, Chinese) and Arabic perform well. Less common pairs and technical content with industry jargon may need human review. VexaScribe doesn't claim to improve on Google Translate's output — our value is eliminating the copy-paste workflow and the second-tool subscription.
Can I export translated subtitles (SRT/VTT)?
Yes. Translated transcripts can be exported as SRT or VTT subtitle files with the original word-level timestamps preserved. Upload a Spanish video, transcribe to a Spanish SRT, then export an English translated SRT for the same video — both files have correctly synced timestamps. Perfect for international YouTube uploads.
Does the speaker detection carry over to the translated version?
Yes. Speaker labels (Speaker 1, Speaker 2, or your custom names) are preserved through translation. If you rename speakers in the original transcript, those names appear in the translated version too. Useful for multilingual interviews where you need consistent speaker attribution across language versions.
What's the difference between transcription and translation?
Transcription converts spoken audio into text in the same language — a Spanish recording becomes Spanish text. Translation converts text from one language to another — Spanish text becomes English text. VexaScribe does both: transcribe Spanish audio to Spanish text, then translate that Spanish text to English (or any of 132 other languages).
Why not just use Google Translate separately?
You can. The downside: you have to transcribe in tool A, copy-paste the transcript into Google Translate, paste back into your subtitle tool, and rebuild timestamps manually. With VexaScribe, it's one upload and the entire workflow happens automatically — speaker labels and timestamps preserved through translation, exportable as SRT directly. If you do this once a year, the manual workflow is fine. If you do it weekly, VexaScribe saves hours.
Can I edit the translation manually?
Yes. The built-in editor shows the original-language transcript and the translated transcript side-by-side. You can edit either version, add corrections, fix machine-translation errors, or refine wording. Both versions update independently so you can preserve the accurate original while polishing the translation.
Can I paste a YouTube or TikTok URL for translated subtitles?
Yes. As of July 2026, VexaScribe accepts YouTube, TikTok, Instagram Reel, and direct audio/video URLs — no download step. Paste a Spanish YouTube video URL, transcribe to Spanish, translate the transcript to English, then export the English SRT. Ideal for creators making international captions from other people's videos (within copyright fair-use limits) or from their own multilingual back catalog. The URL step happens on our side; no software to install.
What you can actually translate for free
Free tiers in this category are advertised in transcription minutes, and translation is usually a paid feature sitting behind them. That makes the headline numbers hard to compare, so here is what each one gets you for translating — checked against each vendor's pricing page on 22 September 2026.
| Tool | Free allowance | Translation on free? |
|---|---|---|
| VexaScribe | First 5 minutes, no account. One preview per week, shared with the transcription tool. 30 minutes on signup. | Yes — transcribed and translated in the same preview. |
| Notta | 120 minutes a month, 3 minutes per recording, 50 uploads. | No. Transcript translation is a Pro feature (€11.69/mo). The free plan also cannot export transcripts. |
| HappyScribe | 10-minute trial, no file size limit. Paid from $8.50/mo for 120 minutes. | Yes — the trial covers transcription, subtitling and translation. |
| Clideo | Free tier limits not published on the tool page. | Requires sign-in before you can try it. |
Reading this honestly: if you want the single biggest free transcription allowance, Notta's 120 minutes a month beats everything here — just not for translating, which is what this page is about. If you want to see a translated transcript before deciding anything, our five minutes and HappyScribe's ten both do that; theirs is longer and ours needs no account. Our real constraint is the one-per-week limit, which is the thing to know before you spend it on a file you are not sure about.
Text and subtitles, not a dubbed voice track
Worth stating plainly, because several tools ranking for this search do something different. You get the speech as text in your target language — a transcript with timestamps, exportable as TXT, DOCX, SRT or VTT. We do not generate a synthetic voiceover and we do not replace the audio in your video.
Use this when you need the words
Subtitling a foreign-language interview, quoting a source, searching a recording, or captioning a video for people who will read it.
Use something else when you need the voice
If the deliverable is a video that now speaks another language, that is dubbing — a different product, and we do not compete with it. Clideo and Maestra both generate a spoken translation (verified 22 September 2026). Note that several tools which look like dubbing are not: HappyScribe, for instance, translates and subtitles but leaves the voiceover to you, which is why exporting a reviewed SRT is the normal handoff into a dubbing workflow.
The two do combine: a reviewed SRT from here is a better dubbing script than a raw machine transcript, so some people run this first and a dubbing tool second.
Translating from a specific language?
This page handles any source language. These four have their own pages because each carries problems the others do not — regional varieties, writing systems, or the messaging apps people actually record in.
| Source language | What is specific to it |
|---|---|
| French | Quebecois joual, Belgian, Swiss and African French; verlan and argot translate literally rather than idiomatically. |
| Spanish | Regional dialects across Spain and Latin America, and what that does to a transcript before translation starts. |
| Chinese | Mandarin versus Cantonese — genuinely different coverage — plus Simplified/Traditional output and WeChat voice notes. |
| Japanese | Keigo politeness levels that English flattens, Kansai and Tohoku dialects, and LINE voice messages. |
Any other language pair works here on this page — those four simply have enough language-specific detail to be worth reading before you upload.
Learn more
Video Translator
Same pipeline for video files + YouTube URLs, with translated SRT/VTT export
Transcribe audio to text
Supported formats, 99 languages, accuracy
Audio translator (translation-only workflow)
Skip the bilingual workflow — just translate to a target language
Best transcription tools for researchers
For foreign-language interviews → English NVivo/ATLAS.ti coding. IRB compliance, verbatim accuracy, dissertation-cost tables.
Translate Spanish audio to English
Language-pair page with regional Spanish handling and specific accuracy figures
Translate French audio to English
France, Quebecois, Belgian, Swiss and African French — with the varieties that need care called out
Translate Chinese audio to English
Mandarin and Cantonese scoped honestly, plus WeChat voice notes
Translate Japanese speech to English
Tokyo, Kansai, regional dialects, keigo + LINE voice notes
MP3 to text
Convert MP3 podcasts and voice memos
Transcript to summary
Translate the summary, not just the transcript
Podcast transcription
Show notes + multilingual subtitles
SRT generator
Create .srt subtitle files with timestamps
SRT translator
Translate an existing SRT to 99 languages — timings preserved
Video to SRT
Generate multi-language SRT from video for global audiences
Transcribe Spanish audio
All dialects + Spanish-to-English translation included
Captions vs subtitles
Subtitles for translation, captions for accessibility
How to add subtitles to a video
Step-by-step for YouTube, Premiere, CapCut, iPhone
How accurate is Whisper?
WER benchmarks by language and condition
Whisper transcription
Hosted Whisper Large-v3 — 99 languages, no GPU
Sermon transcription
Multilingual church services — Spanish, Korean, Mandarin