The most reliable way to translate an audio file to text is a two-step workflow: an AI transcription tool turns the speech into text in the original language, then the transcript is translated. Many tools now do both in one pass. What separates good results from bad ones is mostly the audio quality, whether the tool tells speakers apart, and whether you can export the result in a usable format.
Below are free and paid options, a workflow that works for interviews, lectures and meetings, and the mistakes that cost the most time.
How to transcribe audio to text for free
Free options exist, each with limits. At the time of writing, these are the common ones:
- Microsoft Word for the web. Microsoft 365 subscribers can use Transcribe in Word on the web to upload an audio file and get a transcript with speakers. It is included in the subscription rather than free on its own.
- Google Docs voice typing. Free, but it listens through the microphone. Playing a file out loud into it works poorly and has no speaker labels.
- Phone dictation. Built into iPhone and Android keyboards. Fine for your own short notes, not for a recording.
- Free tiers of online transcription sites. Usually a limited number of minutes. Read the privacy terms before uploading anything confidential.
- YouTube captions. If the audio is already a YouTube video, its auto-generated captions are a free transcript of sorts.
Once you have a transcript, pasting it into Google Translate or DeepL gives a free translation. The cost is time: splitting long text, re-aligning speakers, and fixing formatting.
A step-by-step workflow: audio file to translated text
This workflow uses Record meeting in the SpeakFlow phone app, which imports a video or audio file and does the transcription, speaker separation and translation together.
- Get the file onto your phone. AirDrop, a cloud drive, email or a messaging app all work. Video files are fine; the app uses the audio track.
- Open Record meeting and import the file.
- Wait for the transcript. AI writes it out and tells the speakers apart.
- Check the translation. Read it next to the original. Fix any names that came out wrong.
- Use the minutes. The app drafts a summary, key points, decisions and action items, which is often what you actually need from a meeting recording.
- Ask questions. For a long file, ask what was said about a topic instead of scrolling.
- Export. Save as a Word (.docx) document or share the text.
Recordings stay on the phone. There are free minutes every day with no account or card, so test a short clip before committing a two-hour file. You can also record a meeting directly, up to 5 hours with the screen locked. See meeting transcription apps and free AI meeting minutes.
How to get a more accurate transcript
AI transcription is good, but the input decides most of the outcome.
Before recording
- Put the microphone close to the speakers. Distance hurts more than a cheap microphone.
- Avoid rooms with echo, fans or music. Background music is especially hard.
- Ask people not to talk over each other. Overlapping speech is where speaker labels break.
Before transcribing
- Use the original file, not a compressed copy forwarded through a chat app several times.
- Trim long silences or unrelated sections if the tool charges by the minute.
- Know the main language. Mixed-language recordings work better with tools that auto-detect.
After transcribing
- Check names, numbers and technical terms first; they are the usual errors.
- Read the original and the translation side by side for anything you will quote.
- Rename speaker labels (Speaker 1, Speaker 2) to real names before sharing.
Comparing ways to transcribe and translate audio
| Method | Cost | Speaker labels | Translation | Export | Main drawback |
|---|---|---|---|---|---|
| Voice typing (Docs, phone keyboard) | Free | No | No | Copy text | Live microphone only, not files |
| Word for the web Transcribe | Microsoft 365 | Yes | Separate step | Word | Needs a subscription |
| Transcript + Google Translate | Free | Depends | Yes, manual paste | Copy text | Slow for long files |
| SpeakFlow phone app, Record meeting | Free minutes daily, then Pro or hour packs | Yes | Yes | Word (.docx), share text | Phone app, not a web upload |
Does AI audio-to-text work for every language?
Modern speech recognition covers dozens of languages, and the SpeakFlow phone app supports 60+ for speech. Quality is highest for widely spoken languages and clear audio, and lower for strong dialects, heavy slang or very noisy recordings. If your recording mixes two languages, check both parts of the transcript carefully.
Translation of a full transcript is usually better than live translation, because the tool sees whole sentences and surrounding context. If you only need to follow audio as it plays, see how to translate audio to English, which covers live options.
Exporting your transcript and keeping it private
Which export format should you use?
The right format depends on what happens to the text next.
- Word (.docx) for minutes, reports and anything someone will edit or comment on. Headings, speaker names and paragraphs survive, and it opens in Word, Google Docs and Pages.
- Plain text for pasting into an email, a chat, a ticket or a notes app.
- Subtitle files (SRT) if the goal is captions on a video rather than a document. That is a different workflow; see translating a video into English subtitles.
Whatever you export, keep the original-language transcript alongside the translation for anything important. If someone questions a translated sentence later, you can show what was actually said.
What happens to your recording?
Recordings of meetings and interviews often contain names, numbers and opinions that should not leave your control. Before you upload a file anywhere, check three things: where the file is processed, whether it is stored afterwards, and whether it may be used to train models. With the SpeakFlow phone app, recordings stay on the phone. For anything confidential, ask permission from the people recorded and follow your organization's rules.
Next step
Pick a short recording, import it into Record meeting in the SpeakFlow phone app, and check the transcript, speaker labels and translation before you trust it with a long file. Export to Word when it looks right.
If the audio is a live meeting on your computer
If the audio is a meeting you attend on a computer, you do not have to record and import. SpeakFlow Pro on Mac or Windows translates the meeting live, keeps a full bilingual transcript on the computer and writes an AI summary on demand. That suits regular calls; the phone workflow suits files and in-person meetings.
Either way, the principle is the same: get clean audio, let the tool separate speakers, and check the translated text against the original before you rely on it.
Frequently asked questions
Can I transcribe audio to text for free online?
Yes, with limits. Many online tools offer a few free minutes, and Google Docs voice typing is free but only listens to a live microphone. Check privacy terms before uploading sensitive recordings.
Can AI tell different speakers apart in a recording?
Many AI transcription tools label speakers, including the SpeakFlow phone app. Accuracy drops when people talk over each other or sound similar, so check labels before sharing.
Can I transcribe a video file, not just audio?
Yes. Most transcription tools, including Record meeting in the SpeakFlow phone app, accept video files and use the audio track.
Should I translate audio directly or transcribe first?
Transcribing first and then translating is usually more accurate and lets you check the original text. Tools that do both in one step still follow that order internally.