Genspark Speakly: How to Set Up and Use Its Voice Dictation

Speakly is Genspark's voice dictation app for Mac, Windows, iPhone, iPad and Android, with a dictation key on the desktop and a keyboard on the phone. Free accounts get 4,000 words a week, reset every Monday at 00:00 UTC, while Plus, Pro, Team and Enterprise have no weekly word limit ^[1]. New users start with a 7-day dictation trial before that Free limit applies. Recognition and cleanup run in Genspark's cloud, so Speakly needs a network connection and does not work offline. Our verdict is simple. If you already pay for Plus or Pro, Speakly dictation costs you no credits and is worth installing today. On Free, 4,000 words a week suits short messages and notes, not long drafts.

Pick the right build before you install

Speakly ships as a separate build for each platform. On a Mac it is a .dmg file, with versions for Apple Silicon and Intel; on Windows it is a .zip file for x64 or ARM64 processors; on iPhone and iPad it comes from the App Store, and on Android from Google Play. Check your processor first. A Windows laptop with an ARM chip needs the ARM64 file, not the x64 one. Because every build sends audio to the cloud for recognition, there is no offline fallback for flights, trains or patchy office Wi-Fi.

Sign-in uses your Genspark account, and the weekly word allowance follows the plan on that account. Company SSO and QR login are both supported, which saves typing a long password on a phone. On a managed office network, IT has to allow traffic to *.genspark.ai on port 443. Ask for that before rollout. If the port is blocked, dictation simply fails, since nothing is recognized on the device itself. A short test on the office network, with one person speaking one paragraph, tells you more than any setup guide.

Dictating on the desktop versus the phone keyboard

On the desktop there is no hard cap on a single recording: a session stays open while you keep speaking, which suits long emails, project updates or a first draft spoken in one go. The phone works differently. Speakly's phone keyboard commits one dictation at 60 seconds, so longer thoughts arrive in pieces. Plan for that. A minute-and-a-half voice reply becomes two inserts, and stopping at the end of a sentence before the minute runs out keeps each piece readable instead of splitting a clause in half.

The personal dictionary holds up to 5,000 words, and on the web a single entry can be 40 characters long. Load it with client names, product names, acronyms and team jargon before your first real day of dictation. Genspark is explicit that the list is a recognition hint, not a guarantee of spelling. A rare surname can still come out wrong. Proofread names before sending. That matters most in client email, where a misspelled name costs more goodwill than the seconds you saved by speaking instead of typing.

Agent Mode: speaking a task instead of a sentence

Double-pressing the dictation key switches Speakly from typing to Agent Mode. What you say then runs as a Super Agent task billed to your Genspark account, not as text dropped into the app in front of you. On Plus and Pro, a Standard-mode task draws first on the weekly Free Quota, and Ultra always uses credits. Team and Enterprise seats have no Free Quota. There, every Agent Mode task is paid in credits. For scale, Genspark says a Standard Super Agent message typically costs 10-60 credits and an Ultra one 50-500.

A text task's cost appears only after the turn finishes, in your credit history, and monthly plan credits do not roll over, so a habit of speaking quick Ultra requests can drain a balance faster than you notice. Keep Agent Mode on Standard for routine requests, such as a short summary or a list, and switch to Ultra deliberately. The full credit rules, including what expires when, sit in our guide to whether Genspark is free.

How many words and minutes each plan gives you

New Speakly users get a 7-day trial of dictation, after which the Free weekly limit applies. An invite reward extends it: when you invite someone, both of you get 30 days, but only after the new person makes a first voice input. If the invitee installs and never speaks, nobody gets anything. On Free, meeting audio first uses a one-time 60 free minutes per account, shared with notes on the web and Meeting Bot. After that every minute is paid in credits from the daily 100, and those daily credits stop once the account reaches a lifetime cap on free credits. Heavy meeting users hit that wall first.

Removing the weekly cap means a paid plan. Plus runs from $24.99 a month billed monthly, or $19.99 a month billed annually ($239.99 a year), and its entry tier includes 10,000 credits a month ^[2]. By our arithmetic, monthly billing on that entry tier costs less than the annual payment up to nine months ($224.91) and more from the tenth month ($249.90), ignoring refunds, tier changes and taxes. Annual pays off late. Choose it only if dictation will be part of your work for most of a year.

Meeting notes and Live Translation inside Speakly

Speakly also turns meetings into notes. On Plus, Pro, Team and Enterprise those notes are credit-free for up to 24 hours of audio a day, counted per UTC day, and that allowance is shared with notes recorded on the web and by Meeting Bot ^[1]. On Free, the same notes are charged by duration in credits. Live Translation, with 39 target languages, follows exactly the same billing rule: credit-free inside the daily allowance on paid plans, credits on Free. Speakly is a dictation and meeting tool, not a language course.

You can also bring recordings in. A file import accepts formats such as M4A, MP3, MP4, WAV and MKV, up to 1 GB a file, and a note recorded on a phone stops at 2 hours. For a full-day workshop, start a fresh note at a natural break. Meeting Bot, which draws on the same allowance, joins Zoom, Google Meet, Microsoft Teams, Webex and GoToMeeting calls once a Google or Outlook calendar is connected and Auto-join is turned on. It leaves 60 seconds after everyone else has gone.

Regenerating notes is the hidden cost. Each regeneration, including switching to another template, counts the meeting audio against the daily allowance again, so a 1-hour meeting regenerated 3 times uses 3 hours ^[3]. On a paid plan that only matters on long meeting days. On Free it means paying in credits for the same audio more than once. Pick the template before generating, not after reading the first version.

Exports, retention and what leaves your device

A Speakly note exports as TXT, Markdown or DOCX, and its transcript as TXT, SRT or DOCX. There is no PDF export for either. If a client or manager wants a PDF, export the DOCX and save it as PDF from your word processor. Markdown is the practical choice when the note is headed for a wiki or a project tool that reads plain text, while DOCX keeps the familiar layout for colleagues who live in Word. Decide the destination before the meeting, so the export step takes one click instead of a cleanup session.

Transcripts and audio are kept on your device, and you choose how long: Don't keep, 24 hours, 1 week, 1 month or Forever, with 30 days as the default. For HR conversations, legal calls or client negotiations, shorten that window. One action sends a stored transcript back out. Sending feedback on a transcript uploads the transcript and its audio, so skip the feedback button on anything confidential, even when the recognition was poor and you want to report it.

For companies, Speakly activity appears in Team Admin as Speakly history, not through the Organization Audit and Analytics API, so a compliance team that pulls audit data by API will not see dictation there. Plan the rollout around that, along with SSO and the port 443 rule; our guide to Genspark for small business covers seats, admin roles and billing. If you want meetings and mail to become one searchable memory across Gmail, Outlook and project history, that is the job of Genspark SecondBrain, not Speakly.

When Speakly is worth switching on

Speakly earns its place for people who already pay for Genspark: on Plus and Pro, dictation costs no credits, meeting notes are credit-free up to 24 hours a UTC day, and Agent Mode can draw on the weekly Free Quota. On Team and Enterprise the dictation and meeting allowance is the same, but Agent Mode always spends credits. On Free, treat it as a trial. Four thousand words a week, a 60-second phone limit and meeting audio paid in capped daily credits after a one-time 60 free minutes are enough to judge the recognition quality, not to run a working week on it.