Conversation transcription is the process of converting spoken dialogue between two or more people into written text. It is different from single-speaker dictation in almost every important way: there are multiple voices, speakers interrupt and overlap each other, the audio quality is rarely controlled, and the output needs to clearly indicate who said what. Getting it right requires understanding both the technical constraints and the practical preparation that makes transcription go smoothly.
Whether you are a journalist transcribing an interview, a researcher working with focus group recordings, a podcaster creating episode transcripts, or someone who needs a written record of a recorded meeting, this guide covers the full workflow from recording to final edited transcript.
Recording Quality: The Most Important Factor
The quality of your recording determines the ceiling for transcription quality. No transcription tool can recover audio that was not captured clearly. Investing in recording quality is always worth more than spending on premium transcription software.
Microphone Placement
For in-person conversations, each speaker should ideally have their own microphone, or a central microphone should be placed equidistant between speakers on a table surface. The most common recording mistake is placing one device on the far side of the room, which results in one speaker sounding clear and the other sounding distant. Lavalier (clip-on) microphones are the gold standard for interview recording because they capture each speaker independently at consistent volume.
Remote Conversations
Video call recordings via Zoom, Teams, Meet, or similar platforms give each participant their own audio channel by default, which makes transcription significantly easier. When you record a Zoom call to the cloud, the recording includes separate audio tracks per participant in some configurations, which allows transcription tools to cleanly separate speakers. Local recordings (MP4 files on your computer) mix all audio into a single track, which makes speaker separation harder but still possible with modern diarization tools.
Controlling the Environment
Background noise is the enemy of accurate transcription. HVAC systems, open-plan office chatter, traffic outside a window, and music playing in the background all reduce transcription accuracy significantly. If you have any control over the recording environment, choose a quiet room, close windows and doors, and ask participants to silence their phones.
Speaker Diarization: Who Said What
Speaker diarization is the process of automatically identifying which speaker is talking at any given point in the audio. It is one of the harder problems in speech recognition, and most tools that offer it do so with variable accuracy depending on how clearly distinguishable the speakers' voices are.
For a two-person conversation where the speakers have clearly different voices and speaking styles, diarization usually works well. For a group conversation with three or more speakers, or with speakers of similar age and gender, diarization accuracy drops. You will typically need to manually correct speaker attribution for at least some portions of the transcript.
Some transcription platforms ask you to provide speaker names upfront and attempt to learn each person's voice pattern from the recording. This improves accuracy for longer recordings where the system has time to calibrate. For short recordings under ten minutes, you often get better results by transcribing without diarization and adding speaker labels manually afterward.
Transcription Tools for Conversations
Dedicated Interview Transcription Services
Tools like Otter.ai, Fireflies.ai, and similar services are purpose-built for conversation transcription. They handle multi-speaker audio, produce searchable transcripts with speaker labels, and integrate with calendar and video conferencing tools to automatically capture meetings. These are the best tools if conversation transcription is a core part of your workflow — though their free tiers are limited and paid plans can add up quickly if you transcribe frequently.
Manual Upload Services
If you have an existing audio file to transcribe, web-based services let you upload MP3, M4A, WAV, or other formats and return a transcript. Quality varies significantly by service. Most charge per minute of audio above their free tier. For occasional use, this can be cost-effective compared to a monthly subscription.
Timestamped vs. Plain Transcripts
Some workflows need a simple plain-text transcript where you can search for what was said. Others need timestamps at every sentence, or at regular intervals, so that the transcript can be used as closed captions or synchronized with video. Choose your tool based on which output format your downstream workflow requires — converting between formats after the fact is tedious.
Editing the Transcript: The Step Everyone Underestimates
Automated conversation transcription is never perfect. Even excellent tools will make errors on proper nouns, technical terms, and moments where speakers overlap or speak quietly. Planning for an editing pass is not optional if the transcript needs to be accurate.
Time Estimates for Editing
A common rule of thumb is that editing a transcript takes one to one and a half times the length of the original audio. A 30-minute interview takes 30 to 45 minutes to edit to publication quality. This is dramatically faster than transcribing from scratch by ear (which typically takes three to five times the audio duration) but is still a real time investment. Factor this into any workflow planning.
Cleaning vs. Verbatim Transcription
Spoken language is full of false starts, filler words ("um," "uh," "like," "you know"), incomplete sentences, and recursive self-correction. Whether to include these in the final transcript depends on your purpose. For legal or academic research where the exact speech matters, verbatim transcription is required. For journalism, podcast show notes, or business summaries, a cleaned transcript that removes hesitations and represents what the speaker meant to say is more readable and more useful.
Dictating Your Corrections
An underrated approach to transcript editing is using live voice-to-text to dictate your corrections. If you are replacing a misheard phrase in a document, it is often faster to hold a hotkey, speak the correct phrase, and release — rather than typing it. Tools like Steno, which work in any application on Mac, make this kind of hybrid workflow natural. You use automated transcription for the bulk of the work and live dictation to efficiently apply corrections.
Legal and Ethical Considerations
Recording conversations involves legal and ethical obligations that vary by jurisdiction. In the United States, some states are one-party consent states (only one participant needs to consent to recording) while others are two-party consent states (all participants must consent). Many other countries have different rules. Always understand the applicable law before recording any conversation, and always inform participants that they are being recorded unless you have confirmed it is legal not to do so in your jurisdiction.
Transcripts of conversations may also contain confidential or sensitive information. Consider your data handling obligations before uploading recordings to a third-party transcription service, especially for conversations involving clients, patients, or legally privileged communications.
Practical Workflow for a Research Interview
Here is a concrete end-to-end workflow for a 45-minute research interview:
- Schedule a quiet room or use a video call platform with cloud recording.
- Ask for consent to record at the start, and capture consent on the recording itself.
- Record with a quality microphone for each speaker if in person, or use cloud recording for video calls.
- Export the audio as a WAV or high-quality MP3 file.
- Upload to your preferred transcription service.
- Review the automated output while listening to the audio at 1.5x speed, correcting errors as you go.
- Add speaker labels and clean up the most egregious filler words.
- Export as a Word document or PDF for your records.
This workflow reliably produces usable transcripts in under two hours for a 45-minute interview, compared to five or more hours if you transcribed by ear.
The difference between a good conversation transcription and a great one is almost always in the recording quality — not in the transcription tool.