Choose the right microphone

A microphone selection mistake costs a session. The speech recognition engine is resilient to accent, cadence, and vocabulary. It is not resilient to a microphone that is too far from the speaker, too sensitive to ambient noise, or too inconsistent in its output. For the underlying signal requirements, see Audio requirements. This guide walks through the trade-offs so you can match a microphone to your room and your speaker’s movement.

Lavalier (lapel) microphones

Best for: Stationary or moderately mobile speakers in quiet to moderately noisy rooms.

A lavalier clipped to the speaker’s lapel or collar sits 15–30 cm from the mouth. At that distance, the speaker’s voice dominates the signal. Room noise, HVAC rumble, and distant conversations fall below the recognition engine’s threshold.

When to choose a lavalier:

• The speaker is seated or standing at a podium.
• The room has moderate ambient noise (a conference hall with air conditioning, a meeting room with corridor noise).
• The session runs longer than 30 minutes and the speaker should not be holding a microphone.

What to watch for:

• Clothing rustle against the mic capsule. Clip the lavalier to a rigid surface — a collar button, a tie bar, a lapel seam — not to soft fabric.
• The lavalier falling out of position as the speaker moves. A second clip on the cable, fastened to the inside of the shirt, provides strain relief.

Recommended at three price tiers:

• Budget: Boya BY-M1 (~$20). Omnidirectional, 3.5 mm, sufficient for quiet rooms.
• Mid-range: Rode Lavalier II (~$150). Omnidirectional, broadcast-grade clarity, Kevlar-reinforced cable.
• Professional: DPA d:fine 6066 (~$400). Sub-miniature capsule, exceptional off-axis rejection, industry standard for live speech.

Headset microphones

Best for: Highly mobile speakers, loud rooms, and outdoor sessions.

A headset microphone places the capsule at the corner of the mouth, 5–10 mm from the speaker’s lips. The signal-to-noise ratio is higher than any other common form factor. The engine will not miss a word.

When to choose a headset:

• The speaker gestures, moves across a stage, or walks through the audience (typical in a translated classroom or tour setting).
• The room is loud (a factory floor, a trade show hall, an outdoor stage with wind).
• The session involves significant audience interaction where the speaker must be heard while facing away from the microphone.

What to watch for:

• Fit and comfort. A headset that pinches or slides causes the speaker to adjust it mid-session, which creates handling noise. Test the fit before the session.
• Cable management. The cable runs down the speaker’s back. Secure it with a belt pack or a clip at the collar to prevent tugging.

Recommended:

• Budget: Shure SM35 (~$150). Cardioid, robust build, comfortable for up to 2-hour sessions.
• Mid-range: Sennheiser ME 3-II (~$300). Excellent rejection of ambient noise, slim boom arm.
• Professional: DPA d:fine 6066 Headset (~$500). Same capsule as the lavalier version, mounted on a lightweight headband. Wireless lavalier and headset microphones also work well for docent-led experiences — see Museums and tours.

Boundary (table) microphones

Best for: Fixed-seating meeting rooms with a single speaker position.

A boundary microphone sits on a flat surface — a conference table, a podium, a desk. It captures audio across a wide pickup pattern. These are the most convenient option (no wearable hardware) but the most vulnerable to room acoustics.

When to choose a boundary mic:

• The room has multiple speakers who rotate through a single speaking position (a common pattern when hosting a multilingual meeting).
• The speaker prefers not to wear any hardware.
• The session is in a small meeting room with controlled acoustics (soft furnishings, carpet, non-reverberant walls).

What to watch for:

• Distance. The speaker must remain within roughly one metre of the microphone. Every additional 30 cm reduces the signal-to-noise ratio noticeably.
• Surface coupling. The microphone relies on the table surface for acoustic coupling. A hard wooden or glass table transmits footsteps and paper rustling. Place the microphone on a silicone pad or felt mat to decouple it from the surface.

Recommended:

• Budget: Samson Go Mic (~$50). Compact, adjustable pickup pattern, USB.
• Mid-range: Shure MX418D (~$350). Cardioid condenser, 18-inch gooseneck, XLR.
• Professional: Audio-Technica ATDM-0608D (~$800). Automatic mixer with per-channel gating, Dante digital output, designed for conference rooms.

USB vs. XLR vs. Bluetooth

USB microphones contain a built-in audio interface. They connect directly to the presenter’s computer and require no additional hardware. For Loquira sessions, USB microphones are the simplest reliable option — plug, select in the browser, and the session proceeds. The limitation is cable length (typically 1.5–3 metres) and the lack of an upgrade path to a professional mixer.

XLR microphones require an external audio interface or mixer. They are the correct choice for permanently installed rooms, broadcast environments, and any setup where the microphone cable runs longer than 5 metres. The audio quality is identical to USB at the same capsule quality; the advantage is in infrastructure — balanced cables reject electrical interference, and a mixer provides per-channel gain control.

Bluetooth microphones introduce codec compression that reduces the signal available to the recognition engine. Bluetooth is acceptable for short, informal sessions in quiet rooms but should not be relied upon for any session where transcript quality matters. If Bluetooth is the only option, test the full chain before the session and verify that the live transcript reflects the spoken words accurately.

Evaluating an existing microphone setup

If you already have a microphone and want to know whether it will perform well, run this diagnostic:

1. Open a throwaway Loquira session from the presenter’s device.
2. Position the microphone as it will be used during the session.
3. Read a paragraph of mixed-content text — something with short words, long words, and numbers: “The 2026 quarterly report shows 14.7% growth in the Asia-Pacific region, driven primarily by the manufacturing sector.”
4. Check the live transcript. Count the errors.
5. Repeat the test from a different position — one metre further back, at a different angle, with ambient noise at a realistic level.

If the transcript shows more than one missed or substituted word per twenty words of cleanly read speech, the microphone or its placement is affecting recognition quality. Try a different microphone or a closer position before the session.