Skip to main content

Webinars for 10,000 People, the Email Marketing Channel & Chat Multi-Select

· 21 min read
Copera Team
Copera Team
Product Team

Webinar channels arrive: run a live event or a premiere for up to 10,000 people from a studio that lives inside Copera, then read who came, what they asked and how long they stayed. A new Email Marketing channel brings campaigns, audiences, journeys, forms, site tracking and lead scoring in one place. Chat now lets you pick several messages at once and forward, copy, file, thread, ask AI about or delete them together. Ask Copera transcribes audio and video files you send it, and AI agents can answer your customers directly inside Omni.

New Features​

Webinars — Live Events and Premieres for Up to 10,000 People​

A webinar channel is a place to run an event: you present from a studio, your audience watches, asks and votes, and everything that happened is kept afterwards (#3928, #3977).

  • Live or premiere — go live from your camera and screen, or schedule a premiere of a recording you upload. A scheduled premiere can be switched back to a live event before it starts.
  • A studio inside Copera — the studio fills the channel instead of opening a separate page: the program monitor, the stage, the backstage tray, a meeting-style toolbar and one side panel at a time for Q&A, chat, people and polls. Leaving asks first, and a device check runs before you go on air.
  • A stage you arrange — put up to ten people on stage, switch layouts, spotlight someone, share a screen or play a video. Stage changes show instantly and roll back if the server refuses them.
  • Backstage that stays backstage — speakers waiting backstage are only seen by the team, never by the audience, and bringing someone backstage can be undone from a toast.
  • Audience up to 10,000 — the first seats are fully interactive and larger audiences watch a broadcast stream. If you fill up mid-session you can upgrade the size without ending the event, and the new seats open straight away.
  • Q&A, chat, polls and a call to action — questions can be asked anonymously, upvoted and pinned, similar questions are grouped for you, polls publish their results to the people who voted, and a call-to-action card can be shown for a window you choose. Chat has a slow mode for busy rooms.
  • Public event pages and registration — a public page with your agenda, speakers and cover, name-only registration while the event is live, confirmation and reminder emails, and a "You're in" page. Registrants get a personal link.
  • Moderators — a moderator opens the studio in moderator mode: Q&A, chat, polls, the call to action and raised hands, with no microphone, camera, go-live or end controls.
  • Away mode — navigate away from a live session and it keeps running in a mini studio with a pill you can return from, a hang-up button and the moderator card. The session also survives a reload, and the sidebar shows a live card with the clock and who is on stage.
  • Session insights — after the event: who came and for how long, a retention chart, the recording and the transcript with an editable AI recap, the attendees list with a person drawer, every question and poll result, and Email answers to reply to the room at once. Attendance can be exported, and a channel-wide list of past sessions can be downloaded.
  • Replay — publish the recording as a replay, count one view per viewer per day, and share a public replay link. A speaker taken off the programme becomes an ordinary attendee for members-only replays.
  • "Your webinar recap is ready" — an inbox item that names who came, what they asked and who received the recap email, with links straight to the insights or to share the replay.
  • Honest about what failed — a recording that failed is its own clearly labelled state with no replay or AI offers, and a skipped AI recap can be retried or added later.

Email Marketing — A Channel for Campaigns, Audiences and Journeys​

A new Email Marketing channel brings campaign sending and everything around it into Copera (#3930).

  • Campaigns in an email studio — design an email, save versions as you go, see a live preview and comment on it with your team. Version history is browsable and an edit made minutes after the last one folds into it instead of cluttering the list.
  • Schedule, reschedule, unschedule — the studio header says exactly when a campaign goes out, in the timezone it was scheduled in, warns that further edits will go out with it, and lets you send it back to draft.
  • Audiences and segments — build lists, create a segment from a filter, and describe an audience in plain words to have one drafted for you. Estimates are shown per audience and only ever count the contacts you are allowed to see.
  • Results that read at a glance — five headline numbers plus delivery details, a recipients table, recent activity and a CSV export, all scoped to the contacts you may list.
  • Journeys, forms and site tracking — publish a sign-up form with double opt-in, track activity on your own site, and let both feed journeys that send follow-up emails and credit conversions.
  • Lead scoring — score contacts from what they do, with each signal counted exactly once even if it arrives twice or late. Scoring rules live in Settings → Email sending.
  • Deliverability you can watch — sending reputation, warm-up limits and daily counts per sending domain, a do-not-send list, and a preference centre that refuses to resubscribe a suppressed address.
  • Developer tools — webhooks for email events with per-endpoint retries and a breaker that stops hammering an endpoint that is down, plus transactional sending that is never charged or delivered twice, even if a request is retried.
  • Readable in every theme — status pills and email status chips meet accessibility contrast everywhere, the studio works fully from the keyboard, respects reduced motion, and German copy uses the formal register throughout.

Chat — Select Several Messages and Do Something With Them​

Pick messages from the message menu or by right-clicking, in both the list and bubble layouts, and the composer is replaced by a selection bar (#4001).

  • Forward to several places at once — a redesigned forward dialog with chips, recents and people, a preview, an optional note and a plain reason next to any target you may not post in. Partial results are reported with a Retry, and a forward is never sent twice.
  • Copy, or file as tasks — copy the selection as text, or create one board task per message from a single dialog, with a warning when a message's file cannot be attached.
  • Start a thread from a selection — move your own messages into a thread, or copy them in as quotes. Authors are notified that their message moved, and the parent list stays live.
  • Ask AI about them — open the AI chat drawer with the selected messages as context.
  • Delete together — a confirmation above the bar, all-or-nothing.
  • Forwarded messages keep their credit — the original author is shown in bubbles too, and a forward of a forward credits the message it actually copied.

Ask Copera — Send It Audio and Video​

Ask Copera now accepts audio and video files and transcribes them for you (#3996, #4000).

  • Up to 1 GB and three hours — drop a file into the composer, or pick one from Drive, and watch the upload progress on the chip.
  • A transcript card in the conversation — the transcript appears as a card you can read, retry if it fails and reopen later from your history.
  • Name the speakers — tell Copera who is who and the transcript is relabelled; suggestions are offered but never applied for you.
  • It picks up the language itself, and a file with no speech says so plainly instead of failing.

Transcriptions — Upload a Recording, or Save a Chat Transcript​

  • Upload a recording into All transcriptions (#3999) — add an existing audio or video file to your transcriptions hub, optionally telling it how many speakers to expect, and get its own page with the transcript and summary. An upload that is interrupted is never charged twice, and closing the dialog mid-upload no longer leaves a half-made entry behind.
  • Save an Ask Copera transcript to Transcriptions — one action on the transcript card files it in the hub, carrying over the speaker names you confirmed.

Omni — AI Agents That Answer Your Customers​

Agents can now work inside Omni channels as a first responder, with a person always one step away (#3980, #3986, #3987, #3991).

  • An AI agents tab on the channel — see which agent answers the channel, pick one, create a new one from there, and read its settings without leaving.
  • First responder — the agent answers inbound customers, reads voice notes and photos, files a lead or a record without asking for approval first, and books a call on the calendar when that is what the conversation needs.
  • Hand over to a named person — a hand-off now reaches the person the customer or the agent named, even on an inbox with nobody set as a default, instead of leaving the thread unassigned.
  • Customers are told it is an AI — the channel's disclosure line is now part of the agent's first reply on a conversation, sent once, so it can never arrive without the reply it belongs to.
  • Set up from plain words — tell the agent in your own words which inbox to answer and who to hand conversations to, and it configures itself. It reports back if it could not finish.
  • A closing "thanks" stays with the agent (#3933) — a customer saying "perfect, thank you!" no longer triggers a hand-over to a person just because the agent had nothing to look up.
  • Nothing is answered twice — a reply held for approval is dropped if the customer has written since, so nobody is asked for their email again after they already gave it, and the agent answers from the conversation as it stands now.
  • Teammates named on an inbox can reply — someone added when the inbox was created can now send from it, instead of having their reply silently fail.

AI Agents — They Start With Your Access and Trim It Themselves​

Creating an agent used to mean answering a stack of cards before it could do anything (#3974, #3978, #3979).

  • No setup cards — a new agent reaches whatever you can open and share, the first time it needs it, on your own turn. Cards you were asked before this change are retired in the channel's history.
  • Never more than you — the agent's level on anything is capped by yours, read at the moment of every action: lose access or get demoted, and the agent loses it with you, in the same breath.
  • It says what it opened — when your access opens something a teammate in the channel could not open themselves, the agent posts one short line in the channel naming the item and you, in your language, once.
  • Then it trims — after about two weeks of real work the agent gives back the access it never used, keeping only what its job actually needs.
  • Outbound conversations work from day one — agents created in the app can start customer conversations without hunting for a switch nobody mentioned, and the message explaining it names who can allow it and where.
  • Voice notes get written answers — send a two-minute voice note explaining a job and the reply comes back as text you can skim, copy and search, unless you asked for audio.
  • A clearer roster — the agent list shows tasks done, who is waiting, and whose agent it is.
  • The store lists free presets only, with no prices, ratings or creators.

Boards — To-Do Roll-Up and Repeat Are Now Real Automations​

Roll-up and repeat used to be invisible machinery: tasks moved and nobody could tell why (#3993).

  • Both appear in the Automation center — turning roll-up on, or repeating a to-do, now creates a named board automation you can see, edit, pause or delete like any other.
  • Preview before you commit — the roll-up dialog shows which to-dos would move, and the repeat dialog shows when the next copies land, rejecting an end date before the first one.
  • Clear indicators — a to-do moved by roll-up says so, upcoming copies appear as placeholders, and details popovers explain the rule behind them.
  • Permissions respected — without the automations permission the options are disabled with the reason, and existing automations are shown read-only.
  • Schedules to the minute — the schedule editor now accepts minutes, so roll-up can run at 00:05 in your own timezone.

AI — Ultra-Fast Mode​

  • A faster lane you choose (#3972) — turn ultra-fast on from the composer's + menu and chat, agents, Designs and Code Review each switch to their own fast model line-up. Turning it on asks once (with a "don't ask me again"), turning it off is a single click, and a chip in the composer always shows which mode you are in.

Boards — AI That Builds the Board With You​

  • Native board authoring (#3969) — ask for the board you want and Copera creates the tables, columns, views and forms itself, keeping hidden forms and existing field settings intact instead of overwriting them.

Automations — The AI Input Comes First​

  • Describe it, don't build it (#3990) — the automation center now leads with the AI input wherever it is available, with a quiet Create manually under it for the blank builder. Previously you could only describe an automation on the empty state, so anyone with one automation already lost the option.
  • Pinned views and pages get their own section (#3982) — right under All transcriptions, with four slots and an "N+" tile for the rest. The grey shelf below now holds channels only, and its + opens a channel picker that adds and opens the channel in one step.

Computers — A Desktop That Looks Like a Desktop​

  • A refined, macOS-like look (#3992) — a frosted top bar with the Copera menu and clock, a dock, new theme, icons and cursors, proper fonts including colour emoji, and a Copera wallpaper. Logging out always asks first, and the Super key opens the Copera menu.
  • Status at a glance (#3998) — the block above the channel header is gone. A status icon before the name is grey, orange or red and lists every notice on hover (a click pins them open), and a chip beside the settings shows how many agent slots are used, linking to each agent's chat.
  • Brave by default — Brave is the default browser and sits in the dock, with rewards, wallet, VPN and AI features switched off. Firefox stays installed as a fallback.
  • Agents live in settings — a machine's agents moved into an Agents tab of its settings, and a machine that runs a single agent no longer claims to have ten slots.
  • Dusk wallpaper, and an agent's session no longer starts on a logo wallpaper.

Computers — Oversight, Take-Over and a Movable Call Card​

  • Owners and admins can oversee a private computer (#3958) — a computer an agent works on with its creator's access used to read as a guest limitation to the workspace owner. Now the owner and administrators see its status, whose it is and which agents work on it, with a card that explains why the console belongs to the creator.
  • Take it over — where you are allowed to, Transfer to me moves the computer to you. An agent still working on it can be disconnected first, and the machine only changes hands once that removal is confirmed, so a console is never handed over with someone else's credentials still live.
  • The Watching drawer, redesigned — each watch is one quiet row with the person, the channel, the mode and the last activity, a menu on hover or keyboard focus, an empty state and a loading skeleton. Why watches are not being served is said once at the top instead of on every row, and History now says why each watch ended.
  • The meeting pop-up stays where you drop it — the floating call card no longer snaps to a corner. Drag it with the pointer, touch or arrow keys, and it is remembered per device, pulled in when the window shrinks and allowed back out when it grows.
  • Fifteen terminal tabs — the shared console limit went from ten to fifteen.

Bug Fixes​

  • Creating a workspace could get stuck halfway — workspace creation is now resumable, a second attempt picks up where the first stopped instead of starting over or racing it, and the tailored setup it generates is normalised so a half-made workspace is never left behind (#4008).
  • Ask Copera transcription replies turned into "Message unavailable" — reloading a conversation quarantined every transcription reply the moment its run finished. They now reload intact (#4000).
  • AI images failing, losing their edits or disappearing — generated images keep their identity across a conversation's history, edits reference the right picture, a failed attempt can be retried instead of ending the turn, paid images survive a failed import, and a refusal is stated plainly rather than looking like a crash (#4003).
  • "Download video" on a transcription never downloaded — the browser tried to buffer the whole recording in the tab, so a large file simply spun forever. Downloads now go through your browser's own download manager, and a failed one tells you instead of saving an error (#3976).
  • "Reply in thread" stopped working on a message that already had one — a thread left empty by a failed send made the option refuse for good. Starting a thread now opens the existing one, and copying messages into a new thread never leaves an empty thread behind (#4002, #4004).
  • Messages appearing under the wrong author in chat — overlapping pages of history are deduplicated before authors and dates are grouped, so a repeated message no longer breaks the run of names (#3994).
  • Row status change emails were never sent — people who asked to be emailed when a row's status changes now are, in their own language, with a link to the row, and a failure to send no longer holds up the rest of the row's work (#3994).
  • The CSV import's People step was cut off in other languages — the field and button row is laid out as a grid, so placeholders are readable in Portuguese, Spanish, German and Italian, and everything stacks on narrow screens (#4007).
  • The chat selection bar overflowed in other languages — it now measures what actually fits and moves actions into the overflow menu one by one, instead of using English-sized widths (#4005).
  • Long Portuguese text breaking the interface — the AI pill, the webinar screens, the email studio, Tempo, the board dashboard and the Spark player no longer clip or wrap badly: fixed widths became minimums, truncated labels gained tooltips with the full text, and actions collapse to icons when they do not fit (#3985).
  • Wording and grammar in Portuguese and Spanish — row activity history lost the space between the name and the action, export descriptions used the wrong regional word, the board permission reason read as broken grammar, "My Disk" and the drive's own name disagreed, column type chips showed raw internal codes in every language, and inbox counters now use proper plurals (#3985).
  • A promised date like 30/10/2026 read differently per reader — a date in a message is now read in the day-and-month order of the message it appears in, whatever the locale of whoever looks at it (#3963).
  • An invoice amount repeated in a subject line counted as a second, separate promise (#3964).
  • Settings → Computers would not load — the list asked for every machine in the workspace and was refused for not naming one, so the page failed on every open (#3956).
  • Computer agents went offline after a deploy — a legacy agent host was refused on its first call, treated it as having been revoked, and waited silently. Those hosts stay online, and machines whose records were wrongly flagged during the rollout are repaired automatically (#3954, #3958).
  • Compute consoles lost their session when you navigated away — consoles and their terminal order survive navigation again, and a console that is genuinely unavailable now says why, covering setup delays and session access, instead of showing a blank state (#3967).
  • Long WhatsApp messages were rejected — a message over the provider's limit is split into ordered parts on a paragraph, line or word boundary, with media and the quote on the first part. A rate limit pauses only the remaining parts, and replies and receipts are matched to the right part so a split message is never duplicated or counted twice (#3975).
  • WhatsApp photos and voice notes failing during a provider outage — a timeout or an unreachable media host is retried quietly and left as failed afterwards, instead of being reported as a product error (#3981).
  • Tempo's date picker crashed on the Tempo theme (#3970).
  • A chat message with a broken date crashed the message list — the bubble now simply shows no time (#3981).
  • Editing a row could blank its other cells — an update on a row whose cells were not loaded yet now waits for the server instead of guessing (#3981).
  • The inbox thread list could lock up while scrolling (#3988).
  • A CSV with two columns of the same name was rejected — repeated headers are now kept apart automatically, in board, to-do and contact imports, and the preview matches what the import actually reads (#3961).
  • My Work could not find a row by its board — the search now matches source board names, accents and case included (#3966).
  • Links in contact timeline notes were not clickable, and long notes overflowed their card (#3966).
  • Moving an all-day event in a repeating series produced a zero-length day (#3966).
  • Other people's read markers in a channel were visible in channel responses; they are now limited to the person asking (#3966).
  • Agents booking a calendar invite for the person asking failed with a mismatch when no member was named, which it is not meant to require (#3933).
  • "Don't message anyone else" also silenced the people just named — a prohibition now only holds the people it is actually about (#3941).
  • Installed agents could not send email — their writes outside the workspace are audited as the person who authorised them, carrying that request's own identity rather than a stand-in, so a legitimate send is no longer refused (#3946, #3952).
  • Duplicate AI plan subscriptions could be created when a plan's local link was missing; the live plan is now identified by its price and a change is refused rather than guessed.
  • A dialog or editor going stale after a release — code editors disposed mid-render, and a page still loading an old version's files, now recover with a single reload instead of reporting an error (#3962).
  • A read-only document asking to force-save is now logged quietly rather than surfacing as a failure (#3962).

Improvements​

  • Fewer phantom errors, again — another large pass across the product made expected situations stop reading as failures: a provider outage, a job that will retry on its own, a worker shutting down mid-task, a concurrent job that already did the work, a brief database blip. Alerts are now raised once per real condition instead of once per affected item, so what surfaces is genuinely wrong (#3955, #3957, #3959, #3960, #3961, #3962, #3965, #3970, #3973, #3975, #3981, #3988).
  • Classification and spam checks are never paid for twice — a retried inbox classification gets its own ledger entry, so a retry is no longer refused as a conflict, and an email's fallback is still charged for at most one provider call (#3973, #3975).
  • An agent's spent week is announced once — a turn stopped by the weekly budget and the budget alert used to post the same sentence twice in the same second, and a review was queued for every run that ended over budget. Both now happen once (#3987).
  • Faster email marketing at scale — campaign counters, audience estimates, thumbnails and version lists no longer rewrite or re-read the whole campaign on every event, so busy campaigns stop pushing quieter ones out of the way and lists stay quick (#3930).
  • Maintenance recovery for purchased computers — a machine that cannot be reached can be recovered through restricted maintenance access, without opening a console or enabling password logins.
  • A clearer console status — the messages explaining why a computer console is unavailable were rewritten in all six languages to cover setup delays and session access.
  • Agents are named properly in notices — a creator with no display name is still named, a calendar is said as its owner's, and every notice about what an agent opened exists in all six languages with its plural forms pinned by tests (#3991).