How to Get Better Transcripts from Bad Audio

Not every recording happens in a soundproof studio. Maybe you recorded a meeting in a noisy cafe, a phone interview over a spotty connection, or a lecture from the back row. Bad audio means worse transcription. Here's what you can do about it — both before and after recording.

Why audio quality matters for transcription

AI transcription models work by converting sound waves into text. When background noise, echo, or low volume obscures the speech signal, the model has to guess — and guesses introduce errors. Studies show that clean audio can achieve 95%+ accuracy, while noisy recordings may drop to 70-80%.

The good news: you don't need professional equipment to get good results. A few simple techniques make a big difference.

Before recording: prevention is better than cure

Choose your environment

• Avoid open spaces. Cafes, co-working areas, and open-plan offices are the worst environments for recording. Even if the noise doesn't bother you, it will affect the transcript.
• Small rooms are better. A small meeting room or office reduces echo and ambient noise.
• Close windows and doors. Traffic, construction, and hallway noise are common culprits.

Microphone placement

• 6-8 inches from the speaker. Too far and you pick up room noise; too close and you get distortion.
• Between speakers in a conversation. For interviews, place the recorder at equal distance from both people.
• Don't put the recorder on a vibrating surface. Table taps and vibrations travel through the recorder and create thumping sounds.

Equipment that helps

You don't need expensive gear, but a few cheap upgrades make a real difference:

• $15-20 lapel microphone: Clips to your shirt, dramatically reduces background noise compared to your phone's built-in mic.
• Phone voice recorder app: Built-in recorders work fine. For better quality, apps like Voice Memos (iOS) or Easy Voice Recorder (Android) let you choose higher bitrates.
• For podcasts: A USB condenser microphone ($40-80 range) is the single best investment for both audio quality and transcription accuracy.

After recording: fixing what you have

Sometimes you're stuck with imperfect audio. Maybe the recording already happened, or conditions were beyond your control. Here's what you can do before transcribing.

Noise reduction with Audacity (free)

Audacity is a free, open-source audio editor that can clean up recordings:

1. 1. Open your file in Audacity
2. 2. Select a quiet section (where nobody is speaking but background noise is audible)
3. 3. Go to Effect → Noise Reduction → "Get Noise Profile"
4. 4. Select the entire audio (Ctrl+A)
5. 5. Go to Effect → Noise Reduction again, set reduction to 6-12 dB, click OK
6. 6. Export as MP3 or WAV

Don't overdo noise reduction — too aggressive settings make speech sound robotic, which can actually hurt transcription accuracy.

Volume normalization

If speakers are at very different volumes (common in phone recordings or when one person sits farther from the mic), normalize the audio:

• In Audacity: Effect → Normalize → check "Normalize peak amplitude to -1 dB"
• This brings quiet speakers up and loud speakers down to a consistent level

How speaker labels help with multiple voices

One of the biggest challenges with multi-speaker recordings isn't the transcription itself — it's figuring out who said what. When two or three people talk in a meeting, a plain transcript becomes a wall of unattributed text.

TranscribeCat's speaker labels automatically detect different voices and label them Speaker 1, Speaker 2, etc. This is especially useful for:

• Meeting notes where you need to attribute action items to specific people
• Interviews where you need to separate your questions from the answers
• Group discussions where multiple participants contribute

Common audio issues and solutions

When to invest in better recording vs. better transcription

If you're regularly transcribing and regularly unhappy with accuracy, the problem is almost always the recording, not the transcription service. Spending $20 on a lapel microphone will improve your results more than switching to a more expensive transcription service.

AI transcription accuracy has largely converged across services for clean audio — the gap between a $2/hr service and a $15/hr service shows up mainly in how they handle badaudio. But even the best AI can't recover words that are drowned out by noise.

Bottom line

Good transcription starts with good audio. A quiet room and a decent microphone solve 90% of accuracy problems. For recordings you already have, try noise reduction in Audacity before uploading. And use speaker labels to make sense of multi-speaker recordings — even imperfect ones.