Focus Group Transcription: 8 Tips for Multi-Speaker Audio (2026)

The fastest path to a usable focus-group transcript is fixing the audio before the group starts, not shopping for a better AI tool afterward. Every diarization engine, from the cheapest to the most expensive, degrades when speakers overlap. A well-recorded 8-person group processed by a mid-tier AI tool almost always beats a poorly-recorded group processed by a premium service.

This checklist walks through what actually matters, in the order it needs to happen.

The Multi-Voice Problem: Why Focus Groups Break Transcription

Pre-Group Checklist: Recording Setup

Pre-Group Checklist: Administrative Setup

Two administrative steps done before the group take hours off the back end.

• Place name tents. A folded card with each participant's first name or pseudonym in front of them lets the facilitator call people by name naturally. Hearing "So David, what do you think of that?" in the audio gives you a reliable anchor when you map speaker labels to people later.
• Record a roundtable introduction. Have each person state their name and one sentence at the start. This 60-second segment of clear, individual voice audio gives the diarization model something to anchor on for the rest of the session. It also gives you a reference when you map Speaker 1 through Speaker N manually.

During the Group: Facilitation Habits That Help or Hurt

Some moderator choices produce cleaner audio. Others produce unusable sections.

Practices that improve the transcript:

• Repeat a participant's short answer back before the next probe ("So you're saying the packaging matters more than the price?"). This inserts a clear moderator voice between turns.
• Address participants by name when directing a follow-up. "Maria, does that match your experience?" is a free speaker anchor in the audio.
• When cross-talk starts, acknowledge it and return to one voice: "Let's hear from one person at a time. Go ahead, James."
• Pause 2 to 3 seconds between speakers when you can. It helps the model detect turn boundaries.
• Keep latecomers from entering silently. If someone arrives late, pause and have them introduce themselves briefly.

Practices that hurt the transcript:

• Letting participants finish each other's sentences or build on partial phrases.
• Allowing prolonged cross-talk because the energy feels productive.
• Holding the group in a room with hard walls and no soft furnishings.
• Skipping the roundtable introduction because you're running short on time.

These facilitation habits are worth discussing in your moderator guide before the session. A moderator guide that includes a note like "pause at least 2 seconds between probes" is a cheap insurance policy on transcript quality.

Moderator-Guide Alignment: Setting Up for Faster Coding

One step that most transcription guides skip: align your transcript to the moderator guide before you start coding.

A focus group moderator guide is organized into sections, typically 4 to 8 topic areas with probes under each. The transcript, as it comes out of the AI, is a time-stamped stream of speaker turns. Neither your qualitative software nor your brain wants to code a raw stream.

Before coding, mark the transcript with the section headings from your guide. Find the timestamp where the moderator shifted from warm-up to Topic 1, from Topic 1 to Topic 2, and so on. Insert those section markers as plain-text headers in the transcript document.

This means your coder can search the document by topic section rather than scrolling through 90 minutes of turns. It also means that if two coders are splitting the work, they can divide by section rather than by time range.

If you export to NVivo, MAXQDA, or ATLAS.ti, those tools can import SRT or VTT files with timestamps and auto-sync to the audio for clip-based coding. A well-sectioned transcript with moderator-guide headers drops cleanly into any of those workflows.

Post-Group Workflow

Step 1: Choose a Tool That Handles Multi-Speaker Audio

For online-recorded groups, any major AI transcription platform that supports diarization will work. The quality difference between tools at 2 to 4 speakers is modest. At 6 to 10 speakers, it widens.

For offline focus groups, look for tools that accept multi-track audio files. If your tool requires a mono mix-down, mix to mono rather than stereo. Stereo mix-downs can confuse a diarization model when the same voice appears on both channels at different volumes.

Platforms commonly used for qualitative research transcription include Sonix (pay-as-you-go at $10/hr or subscription starting at $25/month), Trint (subscription-based, per vendor documentation pricing as of mid-2026), and Rev's AI tier ($47.99/seat/month for the Pro plan). Human transcription through Rev starts at $1.99/minute, which is what most researchers cite as the market reference rate for human-verified focus group work as of July 2026.

If you just need a clean transcript without meeting-bot software or a per-seat subscription, ConvertAudioToText accepts uploaded audio directly and returns a speaker-labeled transcript without requiring a login.

See the AI vs. human transcription comparison for a fuller breakdown of when each approach makes sense.

Step 2: Map Speaker Labels to People

The AI gives you Speaker 1 through Speaker N. You convert those to actual participant names or pseudonyms. This is the slowest manual step in focus group transcription, and it determines whether the transcript is usable for analysis.

Use the first 5 minutes of the transcript, where each participant introduced themselves, to anchor the labels. Then move to moments where the facilitator addressed someone by name to verify the mapping holds. If your labels drift in the second half (a common issue with long sessions), re-anchor at the midpoint.

For an 8-person group, expect this step to take 30 to 60 minutes.

Step 3: Handle Cross-Talk Sections

Step 4: Anonymize Before Sharing

Tool Comparison for Focus Group Transcription

Time Budget

A typical project includes 4 to 8 focus group sessions at 90 minutes each.

For an 8-session project, budget 16 to 28 hours of focused post-session work before you open your coding software. Researchers who underestimate this stage miss deadlines or rush the review in ways that introduce coding errors.

The discipline of clean focus group transcription is front-loaded. Get the mics right, run the introductions, brief your moderator on pacing, and the post-processing is a manageable workflow. Skip those steps and no amount of expensive transcription will save a garbled recording.

FAQ

How many speakers can AI transcription handle in a focus group?

Most major platforms technically support 8 to 20 speakers in a single file. The practical reality is that accuracy in speaker labeling degrades as the count rises. AssemblyAI's own documentation notes that accuracy is highest at 2 speakers and decreases as more are added. For a 6-10 person focus group, plan to manually review and correct speaker labels rather than treating the AI output as final.

What is the best microphone setup for recording a focus group?

Two boundary microphones placed at opposite ends of a 6-foot table, each recording to a separate track on a multi-track recorder like the Zoom H5 or H6, plus a lavalier on each facilitator. This setup captures all seats evenly and guarantees clean facilitator audio even when participants cross-talk. Budget around $70-$200 per boundary mic depending on the model.

How do I handle cross-talk in a focus group transcript?

Mark overlapping sections with a plain-text tag like [CROSS-TALK] or [MULTIPLE SPEAKERS] rather than attempting to reconstruct them word-for-word. If the content of a cross-talk section is analytically important, return to the audio for that specific segment. In most 90-minute groups, only a fraction of total cross-talk contains content that matters for your analysis.

How much does professional human transcription for a focus group cost?

Rev, which is the most widely-cited market reference rate, charges $1.99 per audio minute for human transcription as of July 2026. A 90-minute focus group runs approximately $179 at standard turnaround. Rates at volume-focused services like GoTranscript start from approximately €1.05 per minute for 5-day turnaround (checked July 2026). Human transcription costs more per session than AI, but produces higher accuracy on cross-talk and multi-speaker overlap.

Do I need to anonymize the transcript before sharing it with my team?

Yes, before the transcript leaves your immediate research team. Strip names, employer details, locations, and any other identifying information. Replace them with pseudonyms or participant codes. Store the code-to-name mapping separately from the transcript in a password-protected file. If your research is IRB-approved, document your anonymization procedure as many IRBs now request a written protocol describing what identifiers were removed and how the mapping is secured.

Sources

• Rev pricing (human transcription, $1.99/min): https://www.rev.com/pricing (checked July 2026)
• Rev subscription plan details: https://www.rev.com/services/subscription (checked July 2026)
• GoTranscript focus group rates (from €1.05/min, 5-day): https://gotranscript.com/focus-group-transcription-services (checked July 2026)
• Sonix pricing ($10/hr pay-as-you-go; $25-$80/month subscription): https://sonix.ai/pricing (checked July 2026)
• Otter.ai pricing (Free / $8.33 / $19.99 per user/month): https://otter.ai/pricing (checked July 2026)
• AssemblyAI speaker diarization documentation (accuracy by speaker count): https://www.assemblyai.com/blog/top-speaker-diarization-libraries-and-apis (checked July 2026)
• Trint pricing (subscription tiers, see current plans): https://app.trint.com/plans (checked July 2026)