You recorded yourself rehearsing a presentation or a job interview, played it back — and heard “um” and “uh” far more often than you expected. These verbal habits are called filler words, and they’re hard to notice in your own speech.
This guide shows you how to count filler words in a recording automatically with the free Filler Word Analyzer. Your audio is analyzed in your browser and never uploaded.
Listening back to your own voice is uncomfortable. I figured that if I could at least see the count as a number, I could track whether it went down with practice — so I built this.
What are filler words?
Filler words are the little sounds and words people use to fill pauses: “um”, “uh”, “er”, “like”, “you know”. In Japanese, the equivalents are words like “えー” and “あのー”.
A few fillers are a natural part of speech. Too many, though, can make you sound less confident and make your main points harder to follow. In presentations, interviews and online meetings — anywhere you want your message to land — they’re worth cutting down.
What this tool does
- Transcribes a recording, or a live microphone recording, with AI
- Counts how many times your chosen filler words appear
- Shows fillers per minute and your speaking rate
- Highlights the fillers in the transcript
- Lets you add or remove the words to count
- Works entirely in your browser — your audio is never uploaded
How to use it
Step 1. Provide your audio
Open the Filler Word Analyzer and choose a recording of your practice session. To speak and analyze on the spot, click Record with microphone.
Step 2. Choose your settings
Set accuracy to More accurate. Fillers are short sounds, so recognition accuracy directly affects the count. Pick the language and, if you like, edit Words to count. The English default is “um, uh, erm, er”.
The default list works fine as-is
Step 3. Analyze
Click Analyze. The AI transcribes your audio and then counts the fillers. The first time, the AI model (tens of megabytes) takes a while to download.
Step 4. Review the results
You’ll see Total fillers, Per minute and Speaking rate, with the fillers highlighted in the transcript. Notice where they cluster — at transitions, or at the start of answers — for clues on what to work on.
Rather than the count itself, watch how it changes as you rehearse the same talk. Seeing the number drop is a great sign your practice is paying off.
When this comes in handy
- Presentation practice: record a rehearsal and check your filler count and pace
- Interview prep: record answers to likely questions and cut the “um” at the start
- Online meetings: record your own contributions to learn your speaking habits
- Video and podcast recording: spot the filler-heavy parts before you re-record
FAQ
Why aren’t “like” and “so” in the default list?
Those words have perfectly normal meanings too (“I like it”, “so that”), and the tool can’t tell the difference from context. Counting them would inflate your score with false positives, so the default list sticks to sounds that are almost always fillers. Add them yourself if you know you overuse them — just read the result with that in mind.
How many filler words is too many?
There’s no fixed rule; what sounds natural depends on the setting and your style. Instead of aiming for zero, record your own numbers and check that they go down over time.
How can I use fewer filler words?
A common tip is to pause instead of filling the gap. A short silence sounds to listeners like you’re gathering your thoughts. Planning the structure of your talk in advance also helps, because you’ll hesitate less about what comes next.
Things to watch out for
- Results are an estimate: if the AI mishears something, the count changes too. Short sounds like “um” are sometimes left out of transcripts
- It doesn’t judge your speaking: it simply counts the words you specify
- Speaking rate is counted in characters per minute: use it to compare your own sessions rather than as an absolute benchmark
- The first use downloads data: the AI model is tens of megabytes, so use Wi-Fi on a phone
Summary
Verbal habits you can’t hear yourself become clear once you record yourself and see the count. The Filler Word Analyzer does it right in your browser, without sending your audio anywhere.
Start by recording a one-minute self-introduction and counting your fillers.
