What is audio translation?
Audio translation is the process of turning speech recorded in one language into a usable version in another language. The source can be an MP3 interview, a WAV voice-over, a voice memo, a podcast episode, the narration of a training course or the soundtrack of a video. The output can take three forms: translated text (a transcript), translated subtitles timed to the original speech, or a new spoken track in the target language.
What makes audio different from every other file type is that there is no text to work with until someone, or something, transcribes it. A document translator can read a DOCX; it cannot read a recording. That is why every serious audio translation workflow has the same backbone: transcribe, correct, translate, then export as text, subtitles or voice. Smartcat’s AI audio translator follows exactly that sequence, and the transcript you correct once becomes the single source for every language and every output.
Audio translation vs. video translation
audio is video without the picture. If your recording lives inside a video file, the same transcript-first workflow applies, plus subtitle timing and on-screen considerations. Those are covered in our parent guide, Video Translation: The Complete Guide. This guide focuses on the spoken track itself: transcription, voice-over, AI dubbing, AI voices and podcasts.
Four terms people mix up
Transcription converts speech into text in the same language. Audio translation takes that text into another language. Voice-over lays a new narration in the target language over (or instead of) the original, typically for narrated content where the speaker is not on screen. Dubbing replaces the original speaker’s voice with a new voice timed to the original speech, so the result plays like a native-language version. AI now handles each of these steps: speech recognition produces the transcript, machine translation produces the target text, and synthetic voices produce the new spoken track.
The practical consequence is that “translate audio” is no longer a single service you buy from a studio. It is a pipeline you can run yourself, with a human reviewer added wherever the stakes justify it. For a walkthrough of what that looks like in practice, read how to translate audio from a video automatically.
Transcript, subtitles or voice: what should you deliver?
One upload can come back in several forms. The right one depends on where the audio will be used and how the audience will consume it. Smartcat’s MP3 translator sums up the options: a translated transcript (a timestamped document per language), subtitles (SRT or VTT timed to the original speech), an AI voice track, or any of those checked by a human reviewer before export.
| Output | Best for | Languages on Smartcat | Watch out for |
|---|---|---|---|
| Translated transcript | Interviews, meetings, research, compliance records, voicemails | 280+ languages | Speaker labels and timestamps make it usable; check them |
| Subtitles (SRT / VTT) | Audio that belongs to a video or podcast video, social clips | 280+ languages | Reading speed and line length |
| AI voice-over / AI dubbing | Training, product and marketing content people listen to | 35 AI voice-over locales | Pronunciation of names; emotional or comedic delivery |
| Human-reviewed version | Regulated, customer-facing or high-visibility content | Any language with a reviewer | Plan review time into the schedule |
The language column is the decision most teams miss. Text and subtitles are available in any of Smartcat’s 280+ languages, but synthetic speech covers a narrower set: 35 voice-over locales, powered by ElevenLabs, according to Smartcat’s AI dubbing page. Where a target locale has no AI voice, export translated subtitles or a transcript instead — or route that language to a human voice talent.
Subtitles and voice are not mutually exclusive. Several Smartcat workflows deliver both at once: the AI voice and subtitle translation page describes uploading a single-language video and receiving synchronized voice-overs and embedded subtitles from one workflow. If you already have a translated WebVTT file and want it burned into a video, the VTT to MP4 converter translates the subtitle file and embeds it into the MP4.
How audio translation works on Smartcat
Audio translation used to mean sending a file to an agency, waiting for a transcriber, then waiting again for a translator. On Smartcat it is one project: the file goes in as it is, and every step — transcription, correction, translation, voice and review — happens in the same place.
Upload the recording as it is
No conversion to text first. The audio translator accepts MP3, MP2/M2A, M4A, AAC, OGG, FLAC and WMA, and there is a dedicated WAV translator for uncompressed studio files. You can drop in one file or a whole folder; there is no cap on the number of files per project.
Check the transcript once
Smartcat transcribes the speech into a timed, speaker-labelled transcript. Fix a misheard product name or acronym here, before it is translated — one correction flows into every target language.
Pick languages and output
The AI translates the corrected transcript into any of 280+ languages, reusing the glossary and translation memory attached to the project so names and terms stay consistent across every recording. Choose translated text, SRT/VTT subtitles or, in the 35 voice-over locales, an AI voice track.
Review, then export
Download the result, or hand any language to a professional reviewer inside the same workflow. For a batch, select the projects and download once: Smartcat returns a single ZIP with every file in its original format. Need specialists? Hire from the audio translation services marketplace.
Inside the product, audio files run through the same project flow as documents: you upload the file, AI translation runs, and you review the result in the CAT editor. The help center article translate audio files in Smartcat walks through each screen. Direct upload of video and audio files for automatic transcription has been part of Smartcat since 2023 — see upload video and audio files directly to Smartcat for the original announcement and format list.
How audio translation is priced
Audio is metered by the words in the transcript, not by the length of the recording. The audio translator and MP3 translator pages state one Smartword per source word of transcript, so a 60-minute walkthrough with sparse narration can cost less than a 10-minute scripted ad. AI dubbing is metered differently, at ten Smartwords per source word, so budget a dubbed version at roughly ten times the text-translation cost of the same script. New accounts can try everything free for 15 days with 15,000 Smartwords, no credit card. Current plans are on the pricing page.
Transcription: the step that decides quality
Every audio translation starts with speech-to-text, and the transcript sets the ceiling for everything downstream. If speech recognition hears your product name wrong once, that error is translated faithfully into every target language, and voiced by every AI voice. That is why Smartcat puts a human check on the transcript before translation rather than after.
Three things break recordings before translation even starts, according to Smartcat’s MP3 translator page: noisy or muddy audio (background music, wind, heavy compression), sung vocals with instruments behind them, and several speakers talking over each other. Bitrate and encoding make no difference; the quality of the recording does. Clear, single-speaker audio transcribes best.
From transcript to translated text
A translated transcript is the most underrated output. For meetings, research interviews, customer calls and compliance records, a clean, timestamped document per language is what people actually need. The audio-to-text translator page describes the flow: pick languages and upload, the Media Agent transcribes the speech, the transcript is translated and can be refined in the editor, optionally with subject-matter experts or professional reviewers.
When you need a human transcript
Some recordings justify a verbatim human transcript: legal depositions, research interviews with heavy accents, or audio in a language where speech recognition is weaker. Smartcat’s marketplace lists linguists offering multilingual transcription services for audio and video, and the transcript can then go through the normal AI-plus-review translation workflow.
Translating audio from the browser
If your team works in the browser, the audio translator extension page explains how Smartcat automatically transcribes audio, translates the text and prepares it for use as subtitles or voice-over scripts.
Voice-over translation
Voice-over translation replaces or overlays narration with a new voice in the target language. It suits content where the narrator is off-screen or not the focus: explainer videos, product walkthroughs, eLearning modules, presentations and internal communications. Traditionally that meant booking voice actors per language; with AI voices, the narration is generated from the translated script and timed to the original.
Voice-over vs. dubbing
The two terms overlap, and many vendors use them interchangeably. A useful working distinction: voice-over is narration (one voice telling the story), while dubbing replaces each on-screen speaker’s voice line by line. In Smartcat both are produced from the same editable transcript, so the choice is mostly about how many speakers you have and how closely the new voice needs to follow the original timing. The AI speech translator page shows the voice-over flow: upload, AI transcription and translation, choose a voice, review and download — and the Media Agent learns from your edits.
Voice-over for presentations
Narrated slide decks are a common blind spot. The PowerPoint audio translation page covers translating the sound in a presentation and generating multilingual narration and transcripts, for language pairs such as Spanish, French, German, Chinese and Japanese to English.
eLearning voice-over
Training is where voice-over translation earns its keep: courses are long, updated often and rolled out to many regions at once. Smartcat’s eLearning voice-over page describes generating narration from scripts, adjusting pacing, emphasis and pronunciation in the editor, and localizing that audio into 280+ languages, with professional reviewers available from the marketplace. For the full course workflow — SCORM packages, authoring tools and LMS delivery — see our eLearning & SCORM translation guide.
AI dubbing: how it works and when to use it
AI dubbing replaces a recording’s spoken audio with a synthetic voice in another language. On Smartcat, the dub is generated from an editable transcript rather than straight from the audio: fix a name or rewrite a sentence, and the voice says exactly that. Edit a single subtitle line and only that line re-renders — you don’t regenerate the whole video. Speech is separated from the background audio, so music and sound effects carry over under the new voice instead of disappearing.
Transcribe
The spoken track becomes timed, editable cues. Dubbing works from speech, not from text baked into the picture.
Correct
Fix names, acronyms and product terms once, before anything is translated.
Translate
The AI translates the cues into your target languages, reusing your glossary and translation memory, so episode 12 uses the same terms as episode 1.
Voice
An AI voice track is generated, timed to the original cues, in one of 35 locales. Editing a cue changes what the voice says.
Review and export
Assign any language to a professional reviewer in the same workflow, re-generate individual lines, then download the dubbed video or the voice track on its own.
That is the flow described on Smartcat’s AI dubbing page, and the step-by-step version with screenshots is the help article add multilingual AI dubbing to videos. Other pages cover the same capability from different angles: AI voice dubbing (transcription, translation, voice generation and audio sync in one workflow, for L&D, training and marketing video), affordable multilingual dubbing for marketing campaigns, training programs and product launches, and AI video voice translation, which turns one upload into multiple language versions with AI voice-over and burned-in subtitles.
Where AI dubbing works — and where it doesn’t
AI dubbing works best on training, product and marketing video: scripted, information-dense content where clarity matters more than performance. Smartcat’s own AI dubbing page is candid about the limits:
- Synthetic voices read; they don’t act. For emotional or comedic material, use subtitles or human voice talent.
- Timing is matched, not mouth movement. Lip-sync is time-matched rather than mouth-matched, so tight close-ups on a speaker’s face are dubbing’s worst case and subtitles’ best.
- On-screen text is not dubbed. Titles, lower thirds and graphics embedded in the picture must be changed in the source video files.
- Noise degrades everything downstream, which is why the editable transcript step matters.
Three of the four common reasons dubbed videos never ship — mispronounced names, a voice rushing through a long translation, and music that vanishes — each have a control in the editor: set a pronunciation rule, retime the cue, separate the background audio.
Dubbing from subtitle files
If you start from an existing SRT file rather than a video, the dubbing video maker page covers translating the subtitle file with AI and reviewer oversight, then downloading the SRT or integrating it into the video.
Choosing a tool and managing risk
The AI dubbing market moves fast. Smartcat’s blog compares the field in the 10 best AI dubbing tools for enterprise teams, evaluating language coverage, voice quality and cloning, translation accuracy and editability, human review workflow and enterprise security. It is candid that dedicated voice and avatar tools lead on raw voice fidelity and lip-synced talking heads, while a translation platform wins on terminology, review routing and repeatable multilingual output. For the governance side — confidentiality, access control, data handling and human quality checks — read is dubbing AI safe to use in your business videos?
AI voices, voice selection and voice cloning
The quality of a dub depends as much on the voice as on the translation. A perfect script read by a voice with the wrong accent still sounds foreign to local listeners. Smartcat’s AI voices run on ElevenLabs, and voice selection happens in the Subtitle Editor.
Choosing the right voice
In February 2026 Smartcat introduced smarter voice selection and multi-speaker control. The AI Voice Recommendation System shows voices created from the same native language as your target language, plus “optimized” alternatives trained in other native languages that still suit the project. The help article use recommended voices for AI dubbing explains how to preview and apply them, and notes that recent voices are shared across the workspace so teams stay consistent.
Choice of voice has been part of AI dubbing since the female and male voice varieties were announced in 2024 — see Smartcat enhances AI Dubbing with female and male voice varieties, which also introduced side-by-side previews of both language versions in the editor.
Multiple speakers
Interviews, panels and podcasts have more than one voice. Manual multi-speaker support lets you assign different AI voices to different speakers in the same file, so a two-person conversation doesn’t come back as one narrator.
Pronunciation control
Brand names, abbreviations and domain terms are where synthetic voices stumble. Pronunciation control for AI voice-over lets you tell the voice how to say a word using transliteration, without changing the visible subtitle text, and reuse those rules through the project glossary.
Voice coverage vs. translation coverage
Smartcat translates into 280+ languages, but AI voice-over covers 35 locales, including Japanese, Chinese, German, Hindi, Korean, French (France and Canada), and English in US, UK, Australian and Canadian variants. For languages outside that list, deliver translated subtitles or a transcript.
What about voice cloning?
Voice cloning creates a synthetic voice that sounds like a specific person, usually the original speaker, so the dubbed version keeps their identity. Some Smartcat pages, such as AI video and voice translation, describe voice cloning alongside AI dubbing. In the product today, the help center states that voice cloning requests are handled manually through Smartcat Support rather than self-serve, so talk to your account team if you need a cloned or custom brand voice. Before cloning anyone’s voice, get their explicit written consent and agree on where the cloned voice may be used.
Podcast and long-form audio translation
Podcasts are long, conversational and multi-speaker — the hardest combination for any translation workflow. The upside is that podcasts are also repetitive in structure: the same hosts, the same show name, the same sponsor reads. That makes them ideal for a transcript-first workflow with a glossary and translation memory.
Three ways to translate a podcast
- Translated show notes and transcripts. The cheapest option and good for search: publish the translated transcript next to the episode. Any of 280+ languages.
- Subtitles for video podcasts. If the episode is published as video, export SRT or VTT timed to the original speech for YouTube and social clips.
- A dubbed audio episode. The translated transcript becomes synthetic speech, so the episode ships as audio, not a document — as Smartcat’s MP3 translator page puts it. Available in the 35 AI voice locales.
Drop in one episode or a whole folder: one project, one glossary and one translation memory keep recurring names identical across every file, and you pay by transcript words rather than episode minutes. Spanish↔English is the most-requested pair; the Spanish to English audio translator covers recorded interviews and podcasts through to full-length video, with output as a text document, time-synced subtitles or a dubbed audio file.
Internal and learning podcasts
Audio-first learning is growing in L&D. Smartcat’s eLearning voice-over page describes creating and translating eLearning podcasts — write the script, generate the narration, add background music, then use voice-over translation to produce multilingual versions.
Songs and music-heavy shows
sung vocals with instruments behind them transcribe far worse than speech. Audio translation is a speech tool; for lyrics you already have as text, translate the text file instead.
Audio translation use cases
Most organizations translate audio for a handful of recurring reasons. Each one points to a different output.
Training and onboarding
Narrated courses, onboarding modules and compliance training are the classic voice-over job. AI voice-over lets L&D teams launch every language at once instead of region by region. Smartcat’s 2024 AI dubbing launch was aimed at teams producing training and marketing videos — see introducing Smartcat AI Dubbing.
Marketing and product content
Product demos, explainer videos and campaign assets need a voice that matches the brand in every market. Use AI dubbing for the bulk of languages and human review for flagship markets. The AI dubbing launch post on the blog explains how marketing teams use it for localized campaigns.
Meetings, interviews and research
For recorded meetings, user interviews and research calls, a translated, speaker-labelled transcript is usually the right deliverable — it can be filed as meeting minutes, quoted in a report or kept as a compliance record.
Customer and partner voice messages
Support and sales teams receive voice notes and voicemails in languages they don’t speak. The voice message translator transcribes and translates a voice message in seconds and shows the result in a side-by-side editor; the voicemail translator applies the same speech recognition and AI translation to voicemail audio such as customer feedback and partner messages.
Live speech, calls and events
Translating speech as it happens is a different problem from translating a recording. Smartcat describes a real-time voice translation API for speech-to-speech translation in applications, platforms and live events, and a real-time audio translation page for meetings, calls and messages. For high-stakes live events, a human interpreter is still the standard: the marketplace lists interpreting services for conferences, meetings and other events.
AI vs. human audio translation
The right mix depends on audience, risk and budget. Most teams use AI for transcription, translation and voice generation, and bring humans in for review — or for the voice itself when performance matters.
| Factor | AI only | Smartcat AI + human review | Studio / human only |
|---|---|---|---|
| Speed | Minutes | Hours to days | Weeks per language |
| Cost basis | Per transcript word | Per word, plus review | Per minute, per language |
| Names and terminology | Depends on transcript | Glossary, TM and pronunciation rules | Depends on brief |
| Emotional or comedic delivery | Weak | Subtitles or human voice for those parts | Strong |
| Updating one line later | Regenerate | Re-render that cue only | Re-book the studio |
| Language coverage for voice | Limited by voice library | 35 AI voice locales; human talent beyond | Wherever talent exists |
Human expertise slots in at three points. A transcriber can produce a verbatim source transcript; a translator or reviewer can check the target script; and, where needed, a voice artist can record the final track. All three are available through the Smartcat marketplace: transcription services, professional audio translators (filterable by language pair), and interpreters for live sessions. Reviewer corrections train your translation memory, so the next recording starts from a better baseline.
Audio translation tools & resources
Every Smartcat page for audio, voice and dubbing, grouped by what you are trying to do.
Translate audio files
Voice-over and AI dubbing
- AI dubbing from an editable transcript
- AI voice dubbing for global video content
- AI voice-over translation for video
- Multilingual voice-over with AI
- AI voice-over and subtitle translation
- AI speech translator for video and voice
- AI voice translation for videos
- AI video & voice translation
- Affordable AI video dubbing
- Dubbing video maker for SRT localization
- VTT to MP4 — convert and translate subtitled video
Voice messages, live speech and APIs
Training, presentations and human services
Audio and AI dubbing guides, news & product updates
In-depth articles and the history of AI dubbing and audio translation at Smartcat, from the first announcement to today’s voice controls.
Smartcat product documentation
Step-by-step guides for translating audio and producing AI voice-overs in Smartcat. The full video workflow, including subtitle export, is documented in the video translation guide.