Otter.ai: How to Fix Speaker Names and Boost Accuracy

You open the transcript and find a conversation between Speaker 1, Speaker 2 and someone Otter could not place at all. Your company name is spelled three different ways. The meeting happened, but the record is not usable yet.

The good news is that this is mostly a one time cost. Otter treats the corrections you make as training data, so tagging a colleague today means they are recognised automatically next week. The bad news is that nobody tells you that, so most people retag the same people forever.

Why does Otter label people as Speaker 1 and Speaker 2?

Because it can hear that they are different people but has never been told who they are. Otter's documentation describes this as a new speaker: it distinguished a unique voice in the recording but could not match it to anyone already identified in your account.

There is a separate case for an unknown speaker, which is what you see when Otter could not isolate a distinct voice at all. Those two look similar in the transcript but behave differently, and one of the repair tools only works on one of them.

One timing detail explains a lot of confusion: Otter states that speaker identification processing only begins after the recording has stopped or a file has finished importing. You cannot tag anyone mid-recording. The exception is Zoom, where using Otter Notetaker tags speakers in real time from the Zoom participant names.

How do you tag a speaker in Otter?

Click the speaker label in the transcript and pick a person. Otter offers suggestions, lets you search people you have tagged before, or lets you create a new speaker on the spot.

Tag a speaker

1. Open the transcript — Go to the conversation and click the Transcript tab.
2. Click the speaker label — Click the icon showing Speaker 1, Speaker 2 and so on. A pop-up appears with suggested speakers.
3. Choose the person — Click a suggestion, search for an existing speaker, or create a new one.
4. Wait for the green check — Otter confirms the tag and applies it to every paragraph with that same speaker number. A green check mark means it finished.

You only need to do this once per speaker number in a conversation. Otter's documentation is explicit that all other paragraphs with that number update automatically, so there is no point clicking through the whole transcript.

Retagging works the same way when Otter guessed wrong. The catch is that retagging only fixes the paragraph you clicked on, so if the same person was mislabelled in three separate places you have to correct all three.

Renaming is different again, and more powerful. Click any speaker icon to open the list, find the person, open the three dot menu next to them and choose Rename. Otter documents that this updates that speaker's tag across all conversations, not just the one you are looking at.

What is Rematch, and when does it help?

Rematch goes back to an old conversation and applies speaker knowledge you have gained since. Open a conversation that still shows Speaker 1 or Speaker 2, click the three dot menu in the upper right, and select Rematch.

Otter's own example makes the use case clear. You record a call in February with two new people and leave them untagged. In March you meet them again and tag them properly. Going back to the February conversation and clicking Rematch fills in the names using the newer data.

Untagging exists as well, but Otter recommends retagging instead so the audio stays associated with the right person. Like retagging, it only affects the paragraph you clicked.

How does speaker sharing work across a team?

This is the part that saves the most time and is barely advertised. Otter documents that in a workspace, each member's speaker profile is shared, so a colleague gets tagged automatically in your recording without you ever tagging them.

The same applies inside a Channel. If someone in the channel tags an external contact and shares that conversation to the channel, another channel member's future recordings can recognise the same voice. Otter's documentation gives the counter example too: a person outside that channel records the same speaker and has to tag them from scratch.

How do you stop Otter misspelling names and jargon?

Add them to custom vocabulary. Otter lets you teach it proper names, acronyms, product names and industry jargon so it stops guessing at the spelling every time.

Add a term to your vocabulary

1. Open account settings — Sign in to Otter, click your name in the top left, then go to Account settings and the General section.
2. Open the vocabulary manager — Click Manage Vocabulary.
3. Pick the right bucket — Under Personal Vocabulary, choose Names of people for a proper name, or Other vocabulary for a general term.
4. Add team terms if you have them — The same two options exist under Team Vocabulary. Otter documents that team terms apply to every team member's conversations.

How much you can add depends on your plan, and the free allowance is small enough that you have to choose carefully. These figures come from Otter's pricing comparison table, read on 17 August 2026.

With five slots on the free plan, spend them on the things that appear in every single call: your company name, your product name, and the two or three acronyms your industry cannot function without.

What actually causes most transcription errors?

Audio. Otter's own accuracy guidance names background noise as generally the biggest factor in inaccurate transcriptions or missing audio, and that matches what most people find in practice. No amount of vocabulary tuning rescues a recording made in a cafe.

• Cut background noise first. If Otter cannot make out what was said, it may transcribe it wrongly or not at all.
• Avoid people talking over each other. Otter states that overlapping dialogue may not be distinguishable and can be dropped entirely.
• Use an external microphone. Otter recommends omnidirectional for a room with several speakers, and unidirectional pointed at you if you are recording only yourself.
• Get the mic within three feet. That is Otter's stated guidance, along with not moving or touching the device while recording.
• Speak at a normal pace. Otter's advice is to speak clearly and naturally at your usual conversational tone, not to over enunciate.

One more thing worth checking before you blame the model: language coverage. Otter's pricing page lists AI transcription in English, Spanish, French, German, Japanese and Chinese, so a meeting held in a language outside that list is a different problem entirely.

When the cleanup is the actual work

Add it up honestly. You tagged four speakers, renamed one, ran Rematch on two old calls, added five vocabulary terms, and moved your microphone. All of that produced one accurate transcript. Nobody has been emailed and no record has been updated.

Cleanup work like this is exactly what we removed at Sistava. Rather than tuning a capture tool and then working the output by hand, you hire an AI Employee and tell it in plain English what you want handled after a call. It knows who was on the call because it has your calendar and your inbox, it drafts the recap, updates the record, and asks you when it is unsure.

To be fair to Otter, the speaker training model is a genuinely good design. Tagging a person once and having them recognised across your workspace forever is better than what most transcription tools offer, and the vocabulary manager is easy to find. If a high quality searchable archive is your goal, the steps above are the right ones.

Sources

• Otter.ai Help Center, Speaker Identification Overview
• Otter.ai Help Center, Tagging speaker names in a conversation
• Otter.ai Help Center, Manage vocabulary
• Otter.ai Help Center, Tips on improving speech and transcript accuracy
• Otter.ai, Pricing, plan comparison table read on 17 August 2026
• ClickUp Blog, How to Use Otter AI with Zoom, an independent review of the practical limits