21 July 2026
🧑🎨 Custom 3D Avatars Have Moved to Avaturn
Custom 3D avatar creation now runs on Avaturn. Our previous avatar provider, ReadyPlayerMe, has been discontinued, so we've moved to a platform we can keep building on — with a better editor and more expressive characters.
What's changed:
- Avatars are created and customised in Avaturn
- More natural expressions and movement in conversation mode
- You can still assign different avatars to different role plays
💡 Avatars you created previously still display as normal. When you next want to edit one, you'll need to recreate it in Avaturn. Built-in avatars are unaffected, and Walter remains the default.
🎙️ Real-Time Voice Role Plays
Role plays can now run as a real-time spoken conversation. Rather than speaking, waiting, then hearing a reply, the learner and the AI character talk to each other in real time — the character responds as the learner speaks, and either can take the next turn naturally.
What's new:
- Real-time conversation — replies come back as the learner talks, not after a processing pause
- Natural turn-taking — a brief "mm-hm" won't cut the character off, and a pause mid-sentence won't end the learner's turn prematurely
- Live transcript — the conversation is transcribed as it happens and saved to session history alongside the audio
- Switch language mid-conversation — without restarting the role play
- The character can bring things in — refer to a document or show an image at the moment it's relevant
- Spoken feedback at the end — the character delivers the review out loud, and the role play is still scored against the same rubric as before
- Survives a dropped connection — if the network blips, the session resumes rather than starting over
⚠️ Enable it per role play. Real-time voice is set on the individual role play in the Admin Portal, not organisation-wide. Open the role play's settings, switch on real-time voice, and pick a voice. It applies to role plays running in agentic mode.
🎓 Add NoodleMe to Your LMS Course
NoodleMe now appears in the LTI Deep Linking activity chooser, alongside the AI Tutor, knowledge groups, role plays and quizzes. Teachers can place NoodleMe directly into a course so students reach it from the LMS rather than navigating separately.
🔒 Finer Control Over Who Sees NoodleMe
Enabling NoodleMe for your organisation no longer means enabling it for everyone.
What's changed:
- A per-agent switch now controls NoodleMe, off by default
- Even with NoodleMe enabled organisation-wide, students only see the NoodleMe button where the agent's switch is on
- Lets you pilot NoodleMe with specific cohorts before opening it up
💡 NoodleMe is not available on Starter and Educator plans.
🚀 A Clearer Way Into a Space
Students told us it wasn't obvious how to begin. A Start Learning button now appears on the Space card and at the end of setup, so the next step is always visible.
🧭 Spot Agentic Mode at a Glance
Agentic Mode now shows as an icon directly on the agent cell, so you can tell which agents are running in agentic mode without opening each one.
📄 Markdown Files in Knowledge
Course material doesn't always arrive as a PDF or Word document. You can now upload Markdown (.md) files anywhere you build knowledge, and they work across the flows that depend on it:
- Contextualised chat
- Learning outcomes generation
- Smart quiz
- Document summarisation
🔧 Fixes & Polish
- Launching an agent in Agentic Mode — selecting an agent now opens the chat interface directly.
- Question dropdown position — multi-line questions now scroll to the beginning of the content instead of the middle.
- Quiz question redirect — jumping to an incorrect question now shows that question rather than a blank screen.
- Dashboard question counts — Role Play chats are no longer counted in Total Questions Asked, so the figure reflects genuine questions.
- Create Space on first sign-in — the Create Space button now appears immediately for brand-new accounts.
- LMS learner view — removed a menu dropdown that didn't belong in the embedded LMS view.
7 July 2026
📓 Introducing NoodleMe — Students Build Their Own AI Tutor
NoodleMe is a mode inside Walter where students build their own AI tutor from their own notes, slides, handouts or lecture recordings — then share it with classmates.
How it works:
- A student creates a Space and adds a few documents they want to learn from
- NoodleMe builds a guided AI tutor from that material, in the same structured, step-by-step format as Walter
- The student can share their Space with classmates, so one person's effort becomes a resource for the whole cohort
Why it helps: building a tutor forces a student to organise and explain what they know — learning by teaching — which is one of the most effective ways to study.
⚠️ Off by default. A superadmin enables NoodleMe for the organisation and decides how it surfaces. Until then, students won't see it.
💡 Bundled with your Walter subscription — each student can keep up to 3 Spaces at no extra cost. Unlimited Spaces are available as an institutional upgrade.
📱 Share a Space with a QR Code
Students can share a Space with classmates using a QR code, so a Space can be passed around in a lecture or tutorial without copying links.
📊 Progress and Insights Inside a Space
The Insights view learners know from Learner View is now available inside a Space.
Learners can see:
- Progress across the Space
- Where they're doing well and where to focus next
📄 Build a Space from Markdown Files or Web Pages
Source material doesn't have to be a PDF or Word document. Students can now add Markdown files or point NoodleMe at a web page URL when setting up a Space.
🧹 Tidier Spaces and Clearer Navigation
A pass over NoodleMe based on how students actually use it.
What's changed:
- Students can delete a Space they created, freeing a slot without waiting for support
- What was labelled Chapter is now Topic, matching the rest of the product
- The Topic Overview card shows sessions and a progress-based status
- Empty sections are hidden rather than shown as blank blocks
⚡ Faster Quizzes Under Load
A round of performance work aimed at the moments when a whole cohort is active at once.
What's changed:
- Quiz submission is now handled asynchronously, so it stays responsive when many learners submit at the same time
- The Admin Portal no longer polls for notifications on a heavy fixed interval
- Frequently requested data is cached, cutting repeated lookups on common screens
✅ No action needed — applies automatically across the platform.
🔗 Larger SharePoint Document Syncs
The SharePoint sync now handles larger files, resolving cases where bigger documents silently failed to come through.
💡 Part of the SharePoint + PowerAutomate integration. If you sync from SharePoint and have hit size limits before, it's worth re-trying those documents.
🔧 Fixes & Polish
- Email sign-in — fixed a loop where the page reloaded constantly when signing in via an email link.
- Google sign-up from an invite link — no longer lands on a blank screen.
- Quiz layout from Question Bank — quizzes created from the Question Bank now keep their "One at a Time" setting instead of rendering with a broken layout in Webchat.
- Quiz start message — the "Submitted quiz" message no longer appears before the first question.
- Add button on the users page — restored the missing + Add button in the Admin Portal.
- Drag-and-drop images — images added by drag-and-drop in the Admin Portal now display correctly in Webchat.
- Knowledge group progress — progress status now displays correctly.
- Adding a URL in Quick Setup — links added while setting up a Space are now recognised reliably, including on retry.
- Logging out of NoodleMe — now returns to the home page instead of an "Unauthorized" screen.
- Attaching documents in chat — the attach picker stays available after sending.
- Under the hood — a round of reliability work on quiz grading and submission, resolving timeouts and message-size failures that could affect large cohorts submitting at once.
30 June 2026
⚡ Faster Quizzes Under Load
A round of performance work aimed at the moments when a whole cohort is active at once.
What's changed: