What Is Dialect Library? How Voice Training, Consensus Scoring & DL Token Rewards Work

If you just found this site and are wondering what it actually does, you are in the right place. Dialect Library is a platform where everyday speakers get paid to help build voice and language data for underrepresented dialects and languages — no linguistics degree, no special equipment, just your voice and a phone or laptop microphone.

The problem we exist to solve

Most speech-recognition and translation systems are trained on a handful of "high-resource" languages and accents. If your dialect is not one of them, voice assistants mishear you, translation tools mangle your sentences, and your language slowly gets left out of the AI systems everyone else relies on. Fixing that requires large volumes of real, verified speech data in those dialects — data that does not exist yet, because nobody has paid ordinary speakers to record it. That is the gap Dialect Library fills.

What Dialect Library actually is

Dialect Library is a crowdsourced voice-data platform. Trainers (that is you, once you sign up) complete short recording and validation tasks in their own dialect. Every submission is checked for quality through a consensus process before it is accepted, and trainers are rewarded with Dial tokens (DL) for verified work. The resulting dataset is what makes speech tools — voice assistants, transcription, translation — actually work for languages and dialects that are normally ignored.

Our methodology, in plain terms

We do not just accept anything you record. Every task type is designed so that quality can be checked by comparing multiple independent submissions, not by trusting any single recording blindly. Three task types make up the core of the platform:

• Word training — you record yourself saying a single prompted word or short phrase in your dialect. These are the fastest tasks and the building blocks for everything else.
• English-to-dialect translation — you are shown an English word or phrase and record its equivalent in your dialect, helping build translation pairs.
• Reverse validation and sentence rebuild — you listen to or read existing submissions and either confirm/correct them, or help reconstruct full sentences from smaller verified pieces, strengthening the dataset's accuracy over time.

Because multiple trainers independently complete the same or related prompts, the platform can compare submissions against each other rather than relying on one person's word for it.

How consensus scoring works

Every recording goes through a scoring window (currently up to 20 minutes) during which the system looks for agreement across submissions for the same prompt. When enough independent recordings line up, that agreement becomes the score — this is what we mean by "consensus scoring." It is a crowd-verification method, not a single reviewer's opinion, which is what keeps the dataset trustworthy at scale.

We also run a no-fail safety net: if a valid recording genuinely cannot reach consensus in time (for example, you were an early or only submitter for a rare prompt), the system still settles it fairly with a synthetic score rather than leaving you unpaid for honest work. You are never penalized for being early.

Dial tokens (DL) — what they are and how you earn them

Dial, abbreviated DL, is Dialect Library's internal utility token. It is not a cryptocurrency and it does not live on any blockchain — there are no private keys or public ledgers involved. DL is a permissioned balance inside our own database, pegged to a fixed USD rate, that you fund and cash out through USDT/USDC. Think of it as platform credit with a transparent, fixed exchange rate rather than a speculative coin.

You earn DL by completing training tasks. Each task has a small DL stake and a scoring window; once your submission reaches consensus (or is settled through the no-fail safety net), your stake is returned along with any bonus earned for accuracy — so genuine, honest effort is never a losing proposition. Trainers can also earn additional DL through the referral program by inviting other trainers to the platform.

Funding, spending, and cashing out

To take on paid tasks, trainers fund their DL balance using USDT or USDC. From there:

• Each task costs a small DL stake, which is returned (plus any bonus) once your submission settles.
• DL can be withdrawn back to USDT or USDC once your balance passes the platform minimum, subject to a small withdrawal fee.
• Trainers can also trade DL directly with each other through the built-in peer-to-peer market, with tokens held safely in escrow until both sides confirm the trade.

All of this — funding, task stakes, payouts, and withdrawals — is tracked in an auditable internal ledger, so every DL movement in your wallet has a clear, traceable reason behind it.

Why this matters

Every verified recording you contribute helps train speech technology that finally understands your dialect — while you get paid fairly for your time. It is a small, honest exchange: your voice for real income, and a more inclusive voice-AI ecosystem for everyone who speaks like you.

Ready to start?

Create a free account, complete a quick verification step, and you can start picking up training tasks right away. Head to the Learning Center from your dashboard for a short guided course on how tasks, scoring, and DL all fit together.