AI note takers for multilingual meetings that follow a switch from English to Spanish

If you want an AI note taker for multilingual meetings that can follow a switch from English to Spanish mid-call, look for a setting that names several languages in one meeting, because a long list of supported languages does not promise it. On the vendor pages we opened in October 2026, Fireflies (Multi-Language Mode, a beta on Business and Enterprise plans), Granola's desktop app (its Multi-language setting) and Notta (bilingual transcription of two languages you choose) say they handle a switch inside one meeting. tl;dv says it supports several languages in a single meeting with its Whisper model, on Business or Enterprise. Otter's pricing page lists six languages, and we could not confirm how it treats a switch.

Everything here about Fireflies, Notta, tl;dv, Otter and Granola comes from their own help, feature and pricing pages, read on October 16, 2026. We ran no hands-on tests and took no screenshots, so treat each line as the vendor's statement and open the linked page before you decide. The help centers of Otter and Notta refused our page reader, so where only they would have the answer, we write that we could not confirm it.

Many languages or a switch inside one meeting: what is the difference?

An AI note taker that can auto detect the language of a bilingual meeting does not always follow a switch inside it. Language support comes in three levels, and a feature list rarely says which one you get:

• One language, chosen by you. You set the meeting language before the call, and the tool writes the meeting down in that language.
• One language, detected. The tool listens and settles on the main language for the whole meeting. Fireflies' language settings guide says its Auto-detect "can identify only one primary language per meeting", and Granola says its phone apps choose a single language to transcribe in.
• Several languages in one meeting. The tool keeps transcribing when people move from one language to another, sometimes inside one sentence. A bilingual team that switches mid-call needs this level, and vendors often put it behind a separate setting, a beta label or a higher plan.

A page that says "30+ languages" or "60+ languages" tells you how many languages a tool can write down, and on its own says nothing about a change of language inside one meeting. What a bilingual team needs to find is a line saying the tool keeps going when the language changes. Fireflies is direct about the gap: its multi-language guide says you should not pick Auto-Detect for multilingual meetings, "as it transcribes only the dominant language".

Which AI note takers follow a switch from English to Spanish mid-call?

Among the AI note takers for multilingual meetings we checked, three say on their own pages that they follow a switch from English to Spanish mid-call, or between other languages they support: Fireflies with Multi-Language Mode, Granola on desktop with its Multi-language setting, and Notta with bilingual transcription. tl;dv says it handles multiple languages in a single meeting with its Whisper model. The table sums up each page as of October 2026.

Read by need, the choice looks like this:

• People mix English and Spanish inside one sentence and you need every word: Fireflies' Multi-Language Mode describes word-level detection, and Granola's desktop Multi-language setting says it should transcribe all the languages you swap between.
• Your meetings always use the same two languages: Notta's bilingual transcription asks for the two languages up front, which suits a fixed pair once you have checked that your pair is offered.
• Each meeting uses one language, but it changes from meeting to meeting: a detection setting is enough, such as Fireflies' Auto-detect, Granola on a phone, or Barside's Detect automatically.
• You want the speech turned into text on your own computer: Barside, with the limits described further down.

What Fireflies, Notta, tl;dv, Otter and Granola say about switching languages

If you are comparing AI meeting note takers for multilingual English and Spanish meeting transcription in 2026, Fireflies, Otter, Notta and tl;dv each describe a switch of languages differently, and Granola splits its answer between desktop and phone. Each section below keeps to what the vendor's own page says.

Fireflies

Fireflies spells this out in detail. Its guide to Multi-Language Mode, marked beta, says the mode detects and transcribes all spoken languages in the same meeting, "even when speakers switch languages mid-sentence", because it performs word-level language detection. The page says Fireflies supports 60+ languages and lists them, English and Spanish among them, and says the feature is for Business and Enterprise plans only.

To turn it on, the guide sends you to Settings, then Recording & Privacy, the Recording section and Meeting Language, where you pick Multi-Language (Auto-Detect). The language settings guide adds that you can set the language for a single meeting from the Upcoming tab. Two notes matter for planning: the summary is written in the dominant language spoken, and the guide warns against changing the meeting language after a meeting is processed, because that can reprocess the transcript and summary. The default meeting language, per the language settings guide, is English (Global).

Notta

Notta's bilingual transcription page describes the feature as "Transcribe two languages at once" for meetings live or recorded, says Notta can handle code-switching, and tells you to select two languages under the Bilingual transcription tab. That page says 10+ languages are supported, while Notta's step-by-step guide for bilingual business meetings says 23 and tells you to choose the primary and secondary languages before recording starts. The guide also names a Notta Bot that joins Zoom, Microsoft Teams, Google Meet or Webex calls.

As of October 2026, Notta's pricing page lists up to 4 free uses of bilingual transcription and translation a month on every plan. More uses come with the Bilingual Transcription & Translation add-on, sold only on Pro, Business and Enterprise, at US$9 a month billed annually, or US$15 month to month, for up to 200 uses per account a month on Pro and Business. Neither page lists the language pairs. The only pair the guide names is in a screenshot's description, "English-to-Spanish bilingual transcription", and Notta's help center refused our page reader, so check in Notta that your pair is offered before you rely on it.

tl;dv

tl;dv's help article on setting your preferences, dated January 12, 2026, says "tl;dv now supports multiple languages spoken in a single meeting when the transcription model is Whisper", and that Whisper is available on the Business or Enterprise plans. The same article says the transcript language setting will not translate from the language spoken, and that AI notes are written in the language detected for the transcript. Choosing a language for one meeting needs the Chrome extension and a recording you start by hand, and the article says this is not yet supported for Zoom and Teams with the desktop app.

tl;dv's home page says it transcribes in 30+ languages and lists English and Spanish (Mexico, Spain) among them. Neither page we read describes a switch inside one sentence, so we can only report that tl;dv supports several languages in one meeting with Whisper.

Otter.ai

Otter's pricing page lists AI transcription in English, Spanish, French, German, Japanese and Chinese, and "Multi-language support" on the free Basic plan. The page says nothing about a switch between languages inside one conversation or about automatic detection. Otter's help center has an article on its transcription languages, but it refused our page reader, so we could not confirm whether one Otter conversation can hold English and Spanish together. Read that article yourself before a bilingual call that matters.

Granola

Granola's multi-language help page lists 31 languages on desktop (macOS and Windows) and 17 on mobile, with English and Spanish on both. On desktop, with the Multi-language setting on, it says Granola lets you swap between any of the supported languages throughout your meeting and should transcribe all of them. On iOS and Android it picks a single language for the meeting, so the page advises sticking to one language there, and the language cannot be changed once the meeting is transcribed.

You choose the languages for transcripts and notes in Settings → Preferences → Language: English writes the notes in English whatever was spoken, and Auto writes them in the majority language. The page adds that you can change the language during a meeting in the meeting controls, with a brief pause while the new model loads, and that the setting only works for new meetings.

How bilingual teams can get a cleaner transcript

The tips below are general advice for any note taker, drawn from how the vendors above describe their settings. They are not a promise from any vendor, and none of them comes from a test we ran.

1. Set the main language before the meeting. Pick the language most of the call will use, or the multi-language mode if the call truly mixes them. Fireflies warns against changing the language after a meeting is processed, and Granola says its setting only works for new meetings, so a choice made afterwards may come too late.
2. Match the setting to how people talk. If each speaker keeps to one language for long stretches, a two-language mode such as Notta's fits. If people mix languages inside a sentence, look for a mode that says it handles that, such as Fireflies' word-level detection.
3. Check names and numbers. Prices, dates, product names and people's names carry the most weight in a recap, so an error there costs the most. Read them against the recording before you send anything.
4. Say key figures twice, once in each language, or type them in the meeting chat, so the notes have a second record of them.
5. Decide the language of the recap. Fireflies writes the summary in the dominant language, Granola's Auto setting writes notes in the majority language, and tl;dv writes AI notes in the language it detected. If half your readers need the other language, plan to translate or rewrite the recap.
6. Try a short practice call in both languages before a meeting that matters, and read the result line by line.
7. Tell everyone you are recording. Recording rules differ by country and state, and some places require everyone's consent.

How Barside's transcription handles other languages

Barside transcribes calls on your own computer with Whisper, a speech model that runs locally, and its English model comes with the first speech download. An optional multilingual model adds automatic detection and 34 more languages, among them Spanish, French, German, Portuguese, Chinese, Japanese, Hindi and Arabic. No bot joins the call: Barside records your microphone and the sound your computer plays, on Windows, Linux and Mac.

1. Install speech once. Open the assistant with Ctrl+Shift+A, select Assistant settings (the gear at the top) and choose Install speech. It downloads Whisper and its English model, about 150 MB, and needs no AI key.
2. Add the other languages. Under Speech language, choose Install more languages · 141 MB.
3. Pick a language or detection. Choose a language such as Spanish, or Detect automatically, then choose Save assistant settings.
4. Open the call setup. Choose Transcribe a call at the bottom of the assistant, click the microphone on the desktop bar, or press Alt+Shift+Space and type call. The card reminds you: "Let call participants know before recording."
5. Press Start recording. Every 30 seconds, Barside turns the new audio into text on your computer and adds it to a Call transcript note, with the time and a label: You for your microphone and Others for the sound your computer plays.
6. Press Stop & save when the call ends.

What Barside's articles say about languages, and what they leave out:

• One choice under Speech language. You pick a language or Detect automatically there, before the call.
• No claim about a switch. The articles say the multilingual model adds automatic detection and that a call is written down in 30-second pieces. They do not say whether detection follows a move from English to Spanish inside one call, so we make no such claim. If your calls mix the two, try both settings on a short practice call and compare.
• English until the model is in. If you choose a language before the multilingual model is installed, Barside transcribes in English and tells you so, with a pointer to Assistant settings.
• Compact models. Barside uses Whisper's compact base models (base.en for English), which can mishear names, numbers and crosstalk. Each passage keeps its Call audio clip next to it, so you can play it back and fix the words.
• No translation described. Barside's articles describe no translation of a transcript into another language.
• You and Others. Labels come from the channel, so on a group call everyone except you shows as Others.
• Live calls only. Barside transcribes a call as it records it, and it does not transcribe an audio file you already have.
• What differs by system. On Linux, the computer's sound comes from PulseAudio or PipeWire. On a Mac, recording the computer's sound needs macOS 14.2 or later, and macOS asks once to allow Barside under Screen & System Audio Recording.

In Barside's own video meetings for up to 6 people, each person's captions are written by the speech engine on that person's computer, so the meeting note shows who said what. The articles do not say whether each person's captions follow that person's own Speech language choice, so try it before you count on it for a bilingual meeting.

After the call, Meeting follow-through uses the AI you connect in Settings → AI (Claude, ChatGPT, Gemini or OpenRouter) to propose a summary, decisions, tasks with owners and a recap draft. Each decision and task comes with the transcript quote it rests on, so you can check that quote against its audio clip before you save. The full setup guide is transcribe calls on your computer.

Examples: three bilingual teams and what fits them

The people below are made-up examples, and what each one picks follows from the vendor pages quoted above.

• Example: Lucía, an account manager in Miami, runs a weekly call with a client in Bogotá that starts in English and moves into Spanish for the details. She needs both languages in one transcript, so she looks at tools whose pages say they follow a switch: Fireflies' Multi-Language Mode on a Business plan, or Granola's desktop app with Multi-language on. If her client asks that the audio stay on her laptop, Barside keeps it there, and she would set Spanish or Detect automatically and check the English stretches against the Call audio clips.
• Example: Daniel, a product lead, has engineers in Madrid and Austin. Meetings run in English with a Spanish aside now and then, and nobody needs the asides written down. One language set to English is enough in any of these tools, and he spends his checking time on names, dates and version numbers.
• Example: Ana, a recruiter, interviews candidates in Spanish on Monday and in English on Tuesday, one language per interview. A detection setting suits her, such as Fireflies' Auto-detect, Granola on a phone, or Barside's Detect automatically once the multilingual model is installed.

Where the audio goes

In Barside, speech turns into text on your computer. The speech engine is downloaded once from GitHub and its model from Hugging Face, and after that a transcript needs no network. The call audio and the transcript stay in the note, in your library, and they are part of your backups. With Barside Plus, cloud backup encrypts them on your computer before upload, and cloud sync, if you turn it on, copies them to your other computers through Barside's server.

A transcript leaves your computer when you give it to an AI. Meeting follow-through sends it to the AI company you connected, and that company's data policy applies. Read how recordings and transcripts work for the details. For the other tools, read each vendor's privacy page; Granola's language page, for one, says its languages are its transcription provider's supported languages.

How it compares

If your team mixes English and Spanish inside the same sentences and needs every word written down, a tool whose own page says it follows that switch fits better today: Fireflies' Multi-Language Mode or Granola's desktop app, with Notta for a fixed pair of languages. Barside fits when each call mostly uses one language, when you want transcription on your own computer with no minute count, and when you want transcripts in the same app as your notes and tasks.

For wider comparisons, read the lists of the best AI meeting assistants and the best transcription apps, or the guide to AI note takers without a bot. If you are leaving one tool in particular, the Otter.ai alternative and Granola alternative guides compare Barside with each of them feature by feature.

Works well with

A transcript is only the start. These guides cover recording a call, turning it into decisions and tasks, meeting in Barside with captions, and asking your notes about past calls.