Powered by
Fluents records any call (Meet, Teams, Zoom, Slack), transcribes with diarization, and turns it into a structured Markdown ata — paste straight into Obsidian or Notion. Live copilot drafts replies from your project context. (Live mode coming soon.)
— Micael, founder AraraHQ. Still using it daily.
"We're still seeing latency spikes on the webhook delivery."
"Ainda estamos vendo picos de latência na entrega dos webhooks."
Agreed to investigate queue depth. You shipped fix on Nov 20.
Shipped the fix 2 days ago — partition rebalance on high-lag tenants. Let me pull the metrics.
macOS (recommended — fully invisible on screen share) · Windows · Linux
Works invisibly with
Invisible on screen share (ScreenCaptureKit). System audio via BlackHole. Meeting title auto-fill. Native TTS. Recomendado.
Core features rodam. Screen share invisibility via SetWindowDisplayAffinity (não é 100% garantido em todo capture path). System audio requer VB-Audio Cable. Validado em Zoom/Meet padrão.
Transcrição + rascunhos funcionam. Screen share invisibility é bloqueado pelo Wayland por design — só X11. Sem TTS nativo. Use por sua conta.
Why this exists
AraraHQ serves 140+ tenants. Half our clients are in the US. Every Tuesday we had the same problem: our senior engineer — fluent in Portuguese, brilliant in code — would freeze for 3 seconds when a client asked something technical in English.
Not because he didn't know. Because his brain was translating.
I tried Otter, Granola, Fathom. All post-call. Useless for the moment that mattered.
So I built Fluents. Whisper on-device for transcription, local model for draft responses, memory per company so the next call picks up where the last one ended.
It worked. The team asked for access. I figured if it solved our problem, it probably solves yours too.
Pricing reflects that: $19/mo covers infra. Lifetime $249 for the first 300 founders covers my weekends. I'm not building a business here — I'm sharing what works.
What Fluents does
Every company you meet with has a memory file. Decisions, action items, who said what. Before the call, Fluents loads it so you walk in knowing exactly where the last call left off.
Their question translated in your language. Draft response from your actual context — project history, previous calls, your CV if it's an interview. Under 1 second, often 100ms on cache.
Summary, action items, tags — saved locally, indexed by company. Next call with them picks up where this one ended. No manual work, ever.
How it works
Fluents writes an ATS-tuned CV PDF and pre-caches 300+ likely questions from your actual experience.
Overlay pops, invisible on screen share. Question, translation, and draft answer from your CV.
Summary, action items, context for next round — already saved locally before you close the laptop.
The 3-second freeze
Notetakers summarize after. Too late. The moment that matters is 2:47pm, when the VP asks and your brain — fluent in Portuguese — buys time with "that's a great question."
vs. the notetakers
| Capability | Fluents | Granola | Otter | Fathom |
|---|---|---|---|---|
| Live translated suggestions | ✓ Mid-call | — | — | — |
| Memory per company (auto-loads next call) | ✓ Indexed | — | — | — |
| Answers drafted from your context | ✓ < 100ms cache | — | — | — |
| Invisible on screen share | ✓ Mac · Windows | — | — | — |
| ATS-tuned CV builder | — | — | — | |
| Audio stays on device | ✓ Local by default | Server | Server | Server |
| Post-call recap | ✓ Auto | ✓ | ✓ | ✓ |
| Entry price | 3 free/month, then $19/mo | $18/mo | $17/mo | $15/mo |
Pricing
3 free meetings every month, no card. If it saves you once, Pro pays for itself forever.
Prove it works, then pay.
30s install · macOS recommended · Windows beta · Linux X11-only
Everything unlocked. AI included.
or $19/mo · $179/yr
Prefer recurring? $19/mo · $179/yr
300 spots · 14-day refund
The team, the clients, the founder
Built this for my team. We run 140+ tenants on WhatsApp API, most clients in US/EU. Before Fluents, client calls were a bottleneck — even senior engineers would freeze on technical English. Now Fluents runs on every call. The memory-per-company feature is what makes it stick: I walk into call #8 with a client already knowing what we decided on call #7.
Recruiter from a US fintech pings me for a "casual 30-min chat." Twenty minutes in she's grilling me on Swift concurrency. Fluents caught the question in PT before I found the English. Moved to onsite — first time I didn't walk out sure I'd blown it.
Dutch client goes full British accent when he's annoyed. "The bloody endpoint isn't returning the expected payload" shows up translated before he finishes. I look more competent now. Maybe I am. Either way it's working.
Standup with the Austin team. Someone drops "we gotta babysit this rollout" and I'm stuck on babysit — what, manage? while the conversation has moved on. Fluents puts the translation in front of me before he's done. Week one: stopped saying "sorry, can you repeat" ten times a day.
FAQ
Install Fluents and run it on any real meeting — 1:1, client call, standup, interview, whatever. Every user gets 3 free meetings per month, up to 30 minutes each, all features unlocked. No card, no setup, no bait-and-switch. The counter resets on the 1st of every month. When you want unlimited or longer sessions, upgrade: $19/mo, $179/yr, or $249 once (lifetime).
A meeting-based cap protects you and us. It stops the "one 3-hour interview then disappear" exploit without punishing habit users: 3 meetings/month is enough to use Fluents on your weekly 1:1 or a few interviews per month. Paid tier has no cap — run as many and as long as you want.
Literally unlimited. No monthly meeting cap, no 30-minute session cap. Every feature, every modality, as long as you want. We don't rate-limit real users.
Yes. In Settings, toggle "Use my own key" and paste your OpenRouter, OpenAI, or Anthropic key. You get 30% off on renewals / lifetime. Typical cost on your own key: ~$2/mo for 40h of meetings.
Fluents is a meeting copilot — interviews are one use case. Same category as Granola, Otter, Fathom. We translate in real-time and draft responses from your own context. Whether you read the suggestion aloud or ignore it is your call.
By default, transcription runs on-device via Whisper (Apple Silicon: MLX · Intel/Windows/Linux: CPU). Your voice never leaves your computer. If you paste your own Deepgram key in Settings for premium accuracy, audio goes to Deepgram (zero-retention available on their end). Either way, only transcribed text (plus your CV context) reaches the LLM — through our managed Vertex/Gemini or your own key.
On macOS — yes, fully. The overlay uses NSWindowSharingNone, same flag 1Password uses. Tested on Zoom, Meet, Teams, Slack Huddles. On Windows, SetWindowDisplayAffinity — hides in most share modes but the OS can't guarantee it against every capture path. Linux (X11) is experimental. If hiding during live demos is the feature you're paying for, go Mac.
14 days, no questions asked — monthly, annual, or lifetime. Email refund@fluents.app.
Install Fluents tonight. Press ⌘ I before the call. Run it free. Worst case: uninstall.
Best case: the 3-second freeze stops happening — and nobody on the call knows why.
Interviewing soon? Start with a free CV — ATS-tuned, 20 seconds, no signup needed. Then Fluents handles the call itself.
No card · 14-day refund on any paid plan