🇳🇱 Dutch - English Bilingual Speaker (Conversation Recording)

The essentials

• Who we need: you, if you're a native Dutch speaker who naturally slips English into your Dutch
• Where you can be: anywhere. Most conversations happen remotely, and you can also record in person if you're in the same place as your partner
• How you're paid: freelance, project-based. Up to $40 per hour. $20 per hour of conversation + $20 per hour for conversations that meet all requirements
• Each conversation: 10 to 30 minutes
• Your whole mission: up to one recorded hour in total.
• What you need to know about AI: nothing
• What you need: a headset, a laptop and a quiet room

What you'll be working on

Voice AI still breaks on something every Dutch bilingual does without thinking: sliding into English mid-sentence. "Ik had gisteren echt geen tijd, dus ik heb de hele onboarding maar even uitgesteld."

Systems mishear it, force it back into Dutch, or lose the sentence entirely. You record the conversations that teach them better. This data cannot be scraped and cannot be scripted. It exists only if people like you create it.

What this takes

• Staying genuinely relaxed and spontaneous while working inside a defined protocol. That is the real skill here.
• Mixing English inside your sentences, not only between them: "Waarom hebben ze jou tot sales representative gemaakt?"
• Your own English, unforced. No performance, no affected accent.
• A real exchange: everyone takes turns, nobody monologues.
• Your own regional Dutch. We are recruiting across varieties, Hollands, Noord-Hollands, Brabants, Limburgs, Gronings, Twents, Zeeuws. Regional speakers especially welcome.

How it works

1. Apply alone and we pair you up, or apply with people you already have in mind.
2. We brief you and give you access to the recording tool. All you need is a headset.
3. Record 10 to 30 minutes in a quiet room, together or remotely.
4. Talk mainly around work and study, tech and media, travel and life abroad, popular culture, everyday life.
5. We review the conversation, then pay it.

Requirements

• Native speaker of Dutch as spoken in the Netherlands. Near-native or classroom-learned Dutch doesn't qualify.
• Fluent, confident English you actually use when you talk, not just when you write.
• 18 or older. No exceptions.
• A headset or wired headphones with a mic. Not optional: each voice has to be recorded on its own channel, and speakers in a room leak the other person's voice into your track. Earbuds you already own are fine.
• A laptop or computer and a stable connection. The recording tool runs in your browser, there's nothing to install.
• A quiet room where nobody will walk in on you for half an hour.
• Comfortable following written guidelines closely while still sounding like yourself.
• Willing to sign a consent form before recording, covering commercial use of your recorded voice for speech-recognition training.

Nice to have, not required: previous AI audio recording work. Transcription or subtitling experience is a significant plus, so tell us about it when you apply.

What makes a conversation validated

1. Enough English, and the right kind.

A set share of each conversation has to be genuinely in English, and not everything that sounds English counts: iPhone or email are just Dutch now. We show you exactly what counts during the briefing, with examples. This isn't a style preference. It's measured on every conversation, and it's the most common reason a recording gets rejected.

2. No real personal details.

No real names, addresses, phone numbers, medical or financial information, yours or anyone else's. These recordings train models that will later talk to the public, and a model that memorised a real address will repeat it. Whenever you need a concrete detail, invent one.

3. Be yourself, not impressive.

We're not after the cleverest conversation. We're after the way you actually talk: hesitations, interruptions, boring tangents and all. Spontaneous beats polished every time. Reading, rehearsing, or running your lines through an AI tool makes a recording unusable, and it gets rejected.

A few things get a conversation rejected outright: other voices audible, switching into a third language, political, religious, violent or offensive content, and commenting on the recording itself. Or course, the briefing covers all of it before you start.

FAQ

Why only one hour per person?

We want models to learn how Dutch speakers in general move between languages, not how one person does it. Feed a model too much of a single voice and it learns that voice instead of the pattern, which is called overfitting and makes the model worse for everyone. A little from many speakers beats a lot from a few. So yes, this is a short, one-off mission rather than ongoing income.

Do I need to find a partner?

No. Apply alone and we match you. If you already know people who would be good at this, bring them.

What are we talking about, concretely?

A moderator sets up the session and steers topics from the outside. They won't interview you, and they stay off the recording.

Ready? Aan de slag!

Hit Apply and tell us your Dutch regional variety, your city, whether you're applying alone or with others, and any transcription or AI audio experience.