How to Translate Audio to English From a Recorded Conversation

Sarah Johnson
Writes about field sales, meeting notes and voice-first workflows at ParrotNotes. Every article is reviewed by the ParrotNotes product team before it goes live.

Table of Contents
- 1.First, which kind of recording do you have?
- 2.How to translate audio to English with ParrotNotes
- 3.Translate audio to English in two steps: transcribe, then translate
- 4.Translating a conversation as it happens
- 5.Record it so the English comes out right
- 6.Check the English before you rely on it
- 7.When a machine translation isn't enough
- 8.Get the English, then check it
Rafael runs a small import business in Newark. At 7:40 on a Monday morning his supplier in Porto sends a six-minute voice message in Portuguese about a delayed container. Rafael's Portuguese covers "bom dia" and not much more, and his customer wants a new delivery date by lunch.
If you work across languages, you've had a Rafael morning. You don't need an interpreter. You need to translate audio to English fast, with the dates, prices, and names right.
This guide starts with a route chooser, because a voice message you already have, a conversation happening now, and a recording for official paperwork each need a different tool. Then come the routes step by step, a script to say before you record, and a six-point sheet for checking the English.
If most of your recordings start on your phone, try ParrotNotes free and set it up while you read. Everything else below works with tools you may already have.
First, which kind of recording do you have?
The right way to translate audio to English depends on where the audio is right now.
| Your situation | Best route | What you get |
|---|---|---|
| You record client meetings or debriefs on your phone, in more than one language | Record in ParrotNotes Pro with English set as your transcript language | English transcript and summary, every time, without a second step |
| A file already exists (a voice message, an old recording, a video) | Transcribe it in its own language, then translate the text | Original transcript plus an English version you can compare |
| The conversation is happening right now | Google Translate's Transcribe or Conversation mode | Rough English on screen as people talk |
| The English will go to a court, an agency, or a contract | A human translator who certifies the translation | A translation that an official body will accept |
How to translate audio to English with ParrotNotes
This is the route for anyone who regularly records conversations in a language they don't read. Set it up once, and every recording comes back in English.
- Open ParrotNotes and go to Settings. Choose English as your transcript language. You can also turn on Auto detect, so the app works out which language is being spoken.
- Record as normal. Tap the microphone and talk, or put the phone on the table. Recording keeps going with the screen locked.
- Stop and let it transcribe. Once the phone has a connection, ParrotNotes detects the spoken language and writes the transcript in English.
- Read the summary. Every recording gets an AI summary, and Pro adds action items, a drafted follow-up email, and AI chat on every recording, so you can ask it a question.
If the app is new to you, getting started with ParrotNotes walks through the first recording.
Translation is part of multilingual transcription on ParrotNotes Pro, which covers 99+ languages and follows a conversation that switches language partway through. Pro is $19.99 a month, or $14.99 a month billed annually, with 3,000 minutes of recording a month and recordings up to 3 hours, on iOS, Android, and macOS.
For a recording you already have, share the file into ParrotNotes from your phone's share sheet. Imported and shared files are unlimited on Pro. The voice memo transcription guide shows the share steps for iPhone, Pixel, and Galaxy.
Picture Keiko, an executive assistant in Seattle who minutes a weekly call between her CEO and a partner in Osaka, half of it in Japanese. With English set once in Settings, her minutes start from an English transcript, and she spends her time checking names, not typing. (Keiko is an illustration, not a case study.)
Want your next multilingual meeting to come back in English? Download ParrotNotes free and upgrade to Pro when you need translation.
Translate audio to English in two steps: transcribe, then translate
When the file already exists, split the job: transcribe the audio in its own language, then translate the text.
Step 1: get a transcript in the original language. ParrotNotes' free audio to text converter takes an MP3, WAV, M4A, or MP4 up to 10 minutes and 100 MB, 3 files a month free. Leave the language on Auto-detect or pick it from the list, and it transcribes in the language spoken.
Step 2: translate the text. Two well-known free options:
- Google Translate. Paste the text into the box, or upload it as a document. Google's document translation accepts .docx, .pdf, .pptx, and .xlsx files up to 10 MB. Its help pages cover speech from a microphone and documents, not audio file uploads, so the transcript has to come first.
- DeepL. The free plan includes 50,000 characters a month of text translation and 1 file translation a month, up to 5 MB.
Keep both versions: the original is how you check a strange sentence later.
Transcription services can also do both steps in one place. Sonix charges $10 per audio hour for transcription and another $10 per audio hour for translation, with 30 minutes free. Happy Scribe offers the first 10 minutes free and automatic translation into more than 50 languages. Both were checked on their own sites on 2026-09-27.
Translating a conversation as it happens
Sometimes there's no file yet: a contractor is explaining a repair, or a visitor has questions at the front desk. Google Translate's app has two modes for that.
- Transcribe translates in near real time while someone speaks, and when they're done you tap Save transcript to keep it. Google's Transcribe help page notes that if the button is greyed out, the spoken language isn't supported.
- Conversation plays each translation out loud, so two people take turns speaking.
Live translation is for understanding in the moment. For anything you'll act on later, record the conversation too.
Record it so the English comes out right
Every tool above translates what it hears, so better audio is the cheapest way to better English.
- One speaker at a time. When voices overlap, transcripts drop words, and one dropped word can flip a sentence's meaning.
- Phone close to the person who matters most. A phone on the table beats one in your pocket.
- Skip the noise-reduction filter. Deepgram, the speech-to-text company behind ParrotNotes' transcription, explains that noise reduction often removes audio cues that speech models rely on. Send the original file.
- Ask for names and numbers twice. Names of people, streets, and products are the words most likely to come out wrong.
The "say this first" script
When a conversation will need translating, spend the first 30 seconds of the recording on this, and let the other person answer in their own language.
- "Today is [day and date]. We're at [place]."
- "Could you say your name, and spell it for me?"
- "What's the company or project name?"
- "If we talk about prices or dates, could you say each one twice?"
- "Is it all right if I record this so I can have it translated?"
The last line matters. If a recording includes someone else's voice, tell them you're recording and why, and stop if they'd rather you didn't. Rules on recording conversations vary by country and by state, and this isn't legal advice.
Check the English before you rely on it
Machine translation reads well even when it's wrong. Before you forward, quote, or act on it, run the English through these six checks against the original transcript or the audio.
| Check | What to look for | How to confirm |
|---|---|---|
| 1. Names | People, companies, streets, and products spelled differently in each version | Compare with the spelled-out names from the script |
| 2. Numbers | Prices, quantities, and measurements, where one wrong digit changes the deal | Play the audio at that timestamp |
| 3. Negatives | "Can" that should be "can't," "will" that should be "won't" | Read the original sentence, or ask a fluent colleague |
| 4. Dates and times | Day-month order, 24-hour times, and "next Friday" versus "this Friday" | Confirm in writing with the other person |
| 5. Who said what | Promises attributed to the wrong speaker | Check where speakers change in the audio |
| 6. Idioms | A phrase that makes no sense word for word | Treat it as a question, not a fact |
Back to Rafael. His English version says "the container arrives on the 11th and the port can release it." Check three sends him to the Portuguese transcript, where the supplier said não pode: the port can't release it until the customs papers arrive. Check four catches a second problem, because "the 11th" was the 11th of next month. His customer gets a real date before lunch. (Rafael is an illustration, and so are his dates.)
For interviews you'll quote, the interview transcription guide covers how to check quote candidates against the audio with a fluent speaker.
When a machine translation isn't enough
Some English versions have to satisfy an official body. USCIS filing guidance says supporting documents must come with "a complete English translation that the translator has certified as complete and accurate," along with the translator's certification that they're competent to translate into English.
For paperwork like that, and for court or contract use, hire a professional translator. Your recording and machine transcript can still save them time.
Get the English, then check it
Translating a recorded conversation into English takes three moves: pick the route that fits, record so the words survive, and check the English before you trust it.
- Recurring multilingual meetings: record in ParrotNotes Pro with English set once
- A file you already have: transcribe in the original language, then translate the text
- A conversation happening now: Google Translate's Transcribe or Conversation mode
- Anything official: a certified human translator
Keep the original transcript and run the six checks before you forward anything. If you record on the go, the best AI voice recorder roundup and the guide to transcribing voice notes cover the rest of the workflow.
Ready for your next conversation in another language? Download ParrotNotes free, set English as your transcript language on Pro, and get the English summary before you've left the building.
Frequently Asked Questions
Can Google Translate translate an audio file to English?
Not directly. Google Translate's help pages cover speech from a microphone and documents (.docx, .pdf, .pptx, and .xlsx up to 10 MB), not audio file uploads. Transcribe the recording first in its own language, then paste or upload the text to Google Translate, and keep the original transcript for checking.
What's the most accurate way to translate audio to English?
Start with clean audio: one speaker at a time, the phone close to the speaker, and no noise-reduction filter. Then transcribe in the original language and check the English against it for names, numbers, negatives, and dates. For anything legal or official, use a certified human translator.
Can I translate a recorded phone conversation or voice message?
Yes, if you have the audio file. Transcribe it in its original language, then translate the text. With ParrotNotes Pro, you can share the file into the app and get the transcript in English. Tell anyone you record that you're recording, since consent rules vary by place.
How do I translate a conversation with two languages in it?
Use a tool that handles code-switching, where speakers change language partway through. ParrotNotes Pro transcribes 99+ languages, follows a conversation that switches between them, and writes the transcript in the language you set, such as English. Then check names and numbers against the audio, as with any translation.
Is there a free way to translate audio to English?
Yes, for short files. ParrotNotes' free audio to text converter transcribes files up to 10 minutes, 3 a month, in the language spoken. Paste the transcript into Google Translate or DeepL's free plan (50,000 characters a month). Sonix gives 30 minutes free and Happy Scribe the first 10 minutes.
Related Posts

RFQ vs RFP: The Difference and How to Respond to Each
RFQ vs RFP from the supplier's side: what each asks for, how to tell which one you've got, and how the response differs. Includes a triage test and example.

Meeting Cadence: How Often Your Team Should Actually Meet
Set a meeting cadence that fits your team: a map from daily to quarterly, a five-question calendar audit to cut meetings, and notes that let you meet less.

How to Quote an Interview: Find, Check, Edit and Cite
Learn how to quote an interview: find the line in a long recording, check it against the audio, edit it ethically, and cite it in APA or MLA. Free template.