Different tools for different people. Even's built-in Conversate and IRIS Pro both run on the Even G2 — both put a live transcript, cues, and a summary on your lens, and both start in one tap with nothing to set up. Conversate is clean, free and built in, and it now remembers across conversations and answers from the web too. IRIS is just as simple to open — and she remembers: people, decisions, and context across every conversation. Here's the honest side-by-side, so you can pick the one that fits.
Opens as simply as Conversate — tap Start, nothing to configure, first 3 sessions free with everything on. Then the depth is there when you want it: speaker identification, people pages, Obsidian & Google Drive export, your own AI agent, custom modes, audio recordings.
No subscription, built into the Even app, and privacy-first: no audio is ever stored, data stays on-device under EU protection. Simple and self-contained — listen, cue, summarize, and now translate inline — nothing to configure.
| Feature | IRIS Pro3rd-party · Even Hub | Even Conversate1st-party · built in |
|---|---|---|
| In the moment · on your lens | ||
| Live transcript in lensReal-time captions of the conversation | Yes | Yes |
| Captions-only accessibility modeTranscript on the lens with nothing else — no cues, nothing saved | Yes — ephemeral by design, speaker labels included | Transcription runs inside saved sessions |
| Contextual AI cuesSuggestions surfaced mid-conversation | 6 types — answers, follow-ups, explanations, fact-checks, advice, thoughts; plus recaps, and fallacy calls in Debate mode | 4 types — suggestions, answers, concepts, bios |
| Live web / current-events answersPulls fresh facts during the talk | Yes — live search (Tavily / SearXNG / own key) | Yes — live web answers |
| Real-time translationUnderstand another language live | Transcribes multilingual speech in the original language; no translated output | Inline translation, recently added |
| Prep notes before a talkFeed context ahead of time | Yes — multiple files, per mode | 5 files (5 MB each) + a 5,000-char text field |
| Ask your assistant mid-conversationDirect question, answer on the lens | Yes — chat + push-to-talk | Not offered |
| Spoken answers (beta)Ask out loud, hear the answer in the assistant’s own voice | Yes — on the glasses, or a spoken back-and-forth on the phone; a once-a-day good morning with what’s ahead, otherwise never speaks unless spoken to | Not offered |
| Custom action buttonsYour own one-tap prompts on the live talk | Yes — tap a button, IRIS runs your instruction on everything said so far, answer on the lens | Not offered |
| Voice-following teleprompterScript tracks your speech, holds place on interruptions | Yes — follows your voice, marks your place when interrupted, cues the answer | Not offered |
| Audience Q&A on the lensRoom asks and votes from their own phones, no account | Yes — six-character join code or QR, top-voted question in front of your eyes | Not offered |
| Hands-free hub on the glasses (beta)Run the app from the lens, no phone | Yes — ask a question, read your agenda & to-dos, or command your agent, all on the lens | Not offered |
| Narrow cue columnSlim centered cues — reading looks like a glance | Yes — narrow/medium/wide, long cues widen just enough to fit | Not offered |
| Knowing who's who | ||
| Speaker identificationTells who is speaking | Yes — voice enrollment + wearer detection | Transcript only |
| Name & relabel speakersPut real names to a conversation | Yes — including past sessions, and correcting a name she got wrong teaches the voice | Not offered |
| After the conversation | ||
| AI summaryKey points & action items | Yes | Yes — organized on-device |
| Memory across conversationsRemembers you session to session | Yes — carries people, tasks & context between sessions | Yes — remembers across conversations |
| Chat historyReopen a past chat and carry on | Yes — every Ask conversation kept and listed by day, session chats filed under their session | Not offered |
| Summary templatesShape the recap to the use-case | 20+ templates | Single format |
| To-dos, reminders, briefingsCaptured from what was said, kept across sessions | Yes | Not offered |
| Calendar feedAppointments on the glasses | Yes — read-only .ics | Not offered |
| Ask about your own documentsLoad a contract, handbook or spec and ask about it hands-free | Yes — answers from the matching passage, quoted, with the document and section named; says so when it isn’t in there | No document library |
| Your data | ||
| Audio recordingKeep the audio, not just text | Yes — optional, off by default, stored privately in IRIS’s cloud | By design — transcripts only, no audio kept |
| Export to ObsidianInto your own vault | Yes — auto after each session | Not offered |
| Export to Google DriveA doc per session | Yes — auto after each session | Not offered |
| To-do sync to a task appGoogle Tasks or Todoist, both directions | Yes — edits, dates and deletes travel both ways | Not offered |
| Your car from the lensTesla, through your own Tessie account | Yes — charge and climate by voice; opening it asks first | Not offered |
| Your bookmarks as contextRaindrop, or a self-hosted Karakeep | Yes — Ask searches your own library and cites the link | Not offered |
| Where it's storedWho holds your conversations | Your device and your integrations, plus a cloud backup we hold so you can restore | On device, Even app — EU data protection |
| Make it yours | ||
| Custom modesDifferent behavior per context | Yes — Family, Sales, … with categories | Single experience |
| Bring your own AI keysRun on your own accounts | Claude, OpenAI, Groq, Gemini, Soniox, Deepgram — or managed | Managed only |
| Your own AI agentPipe sessions to a personal agent, or run cues on your own model | Yes — Hermes or OpenClaw with auto-memory, or any OpenAI-compatible model: Grok, OpenRouter, Groq, local Ollama (guide) | Not offered |
| Voice capture with the app closed (beta)"Hey Even" in your own words → to-dos, notes, timed reminders (emailed), calendar events; ask about memories and recent recordings. | Yes — to-dos, notes, reminders & calendar cards, answers from your memory on the lens | QuickList notes only |
| Model & availability | ||
| PriceCost to the wearer | $14.99/mo · $149.99/yr · 3-day trial | Free — included with the glasses |
| DistributionHow you get it | Install from Even Hub | Built into the Even app |
| Data-protection posturePrivacy stance | Per-request relay, nothing used for training. Pro backs up sessions and Ask chats to IRIS’s cloud (on by default, switch off in Settings) | EU standards, no audio stored |
The short version: Conversate is a clean, self-contained tool — free, native, no audio stored. IRIS Pro is for people who want to own what their glasses hear: who spoke, an exportable record, their tools written into, their own agent, their own keys. Different tools for different people.
Compared on 23 August 2026 against Conversate in the Even Realities app v2.2.9, with the memory and live-web rows updated on 14 September 2026 as Conversate gained both, using Even Realities' Conversate page and support docs; the IRIS column was last updated for IRIS 2.1 in September 2026. Even Realities' other first-party apps are not included. Conversate may have changed since; if you spot something out of date, tell us at [email protected].