How to transcribe a recording that mixes Spanish and English
Short answer
Dolcenota detects Spanish and English on its own. It can work well when one language dominates, but the other-language phrases and names are the most likely to be missed. If you switch mid-sentence, those stretches can come out wrong in our tests. Try a one-minute clip first and check them.
If you speak both languages at work, at home or in class, you have probably watched a transcript fall apart at the exact moment you switch. A recording that mixes Spanish and English is one of the harder things to transcribe, and no tool gets it right every time.
This page says what to expect from Dolcenota, including where it falls short, so you can decide with a short test instead of after an hour of lost minutes. The test results below are small and our own, and we say exactly how small.
What to expect, by how the languages mix
You don't pick a language: it is detected automatically, and the transcript keeps each language as it was spoken. The catch is that the transcriber tends to settle on one main language for a recording, and the other language is the one that suffers.
| Your recording | What to expect |
|---|---|
| All in Spanish, or all in English | Works well when the audio is clear. |
| One language dominates, with a few words or phrases from the other | In our one short test of this, the result came out almost clean. Errors are most likely in the other-language phrases, names and terms, so check those. |
| Each speaker keeps their own language, or an interpreter alternates turns | We haven't tested this case on its own. Try a one-minute clip before you rely on it. |
| The languages alternate inside sentences (Spanglish) | The hardest case. In our test, stretches in the minority language came out wrong or as nonsense words. |
What went wrong in our test
In October 2026 we ran seven short clips through Dolcenota's transcription. We made them with computer-generated voices that switch between Spanish and English: one speaker, two, three, a noisy room, and a WhatsApp-style voice note. It is a small test with synthetic speech, not a benchmark, and we are not publishing an accuracy percentage from it. What we saw:
- Spanish stretches inside a recording the system treated as English came out as English-sounding nonsense. The English stretches were fine.
- A short voice note where one language dominated came out almost clean. In another clip where one language also dominated, most of the words in the other language came out wrong.
- A Spanish clip packed with English product names came back with no text at all.
- In two clips with several speakers, the transcript labeled everyone as one speaker. We can't tell how much of this comes from the voices being synthetic.
An example from the test
In one clip the voice said: “Ayer tuve una reunión con el equipo de ventas y la verdad…” (Yesterday I had a meeting with the sales team and honestly…). The transcript wrote: “aertu mil marion con nelequipo de menda silla berda”.
That is what can happen when the transcriber takes a Spanish stretch for English. It is why we suggest testing before you trust a long mixed recording.
How to get a better transcript
- Test first. Upload one minute of your real audio, with the names and terms that matter. It uses about a minute of your plan and shows you the problem before you commit an hour.
- If you can, keep each language to its own stretch and avoid switching mid-sentence. We haven't tested whether uploading each language as a separate transcription helps; if you try it, don't use the multi-file option, which joins the recordings before transcribing.
- Keep the microphone close and let one person talk at a time. Overlapping voices plus a language switch is the hardest combination.
- It may help to say and spell names and product terms early, in the language you want them written. We haven't tested this.
- Check the switched passages against the audio. Each line has a timestamp, so you can find the spot without listening to everything.
- Fix the text in the Word or TXT export. The transcript text isn't edited inside Dolcenota.
What the summary and questions do with a mixed recording
The summary and the answers to your questions are built from the transcript, so a garbled passage carries over into them. The summary is written in the language you have active in Dolcenota, and you can translate it to the other language with one tap. Translating does not repair a transcript that came out wrong.
When Dolcenota fits, and when to compare first
- Fits: recordings that are mostly in one language with only a few words of the other, WhatsApp voice notes where each note is mostly one language, and any case where you also want a summary and answers to your questions.
- Compare first: most of your audio switches language mid-sentence, it involves more than two languages, or you need a certified transcript.
- Run the same one-minute clip through every tool you are considering, including ours. Free plans make that cheap.
What we haven't measured
We haven't yet tested real people switching languages, and we haven't compared tools on real mixed recordings, so we make no claim to be the best at this. We also haven't tested whether recording each language separately or spelling out names improves the result. We will revise the page if our results change.
Frequently asked questions
Can Dolcenota transcribe Spanish and English in the same recording?
Yes, it detects both automatically. Accuracy can be good when one language dominates and the other shows up in short phrases, but those phrases and names are the most likely to be missed; it is lowest when the languages alternate inside sentences. Test a short clip first.
Do I have to choose the language?
No. The language is detected automatically, and the transcript keeps each language as it was spoken.
Why do some passages come out as nonsense words?
The transcriber tends to settle on one main language for the recording. Stretches in the other language can be written as if they were the main one, which gives words that sound similar but mean nothing.
In which language will my summary be written?
In the language you have active in Dolcenota. You can translate it to the other language with one tap. The transcript stays in the languages spoken.
What if the transcript comes back empty?
It means no words were found. Dolcenota tries not to use your minutes in that case, and the transcript tells you whether they were returned. Check that the audio isn't silent, then try a shorter clip.
Is the transcript certified for legal use?
No. AI transcripts aren't certified. For legal material in two languages, have a person check the transcript against the audio.
Try it on one minute of your audio
Upload a short mixed-language clip, read the switched passages and decide with your own recording. Free to start, no card needed.
Test one minute free