Start with the job you want Violet Avatar to do
Violet Avatar is now organized around AI workers. Choose a job, then add capabilities, knowledge, workflows, and an Avatar. You can try live translation and public services immediately, or sign in to build reusable, shareable, and trackable AI workers.
Use-case first
The main paths cover call translation and interpreting, interviews, learning and teaching, customer support, and custom workers that extend to tours, companionship, and marketing.
AI worker and Avatar are separate
The AI worker defines the job, knowledge, and workflow. The Avatar is the 2D, 3D, or voice-only interface people interact with. The same worker can use different appearances.
Live interaction and background work
The Live workspace handles voice, captions, translation, and Avatar interaction. Workflows let an agent produce videos, narration, captions, and reviewable outputs in the background.
Explore, create, run, improve
Start with public use cases and templates, create your version, then use worker details and activity records to keep improving it.
Public discovery and selected instant experiences do not require sign-in. Creating, saving, managing, scheduling, and viewing private activity require an account.
Three ways to begin
Pick the entry point that matches what you need now. You do not need to understand models or low-level settings first.
Try without setup
Onsite interpreting listens and translates on one device. AI-assisted call on the home page creates an ad-hoc room. Explore opens public AI workers.
Use an invite code or share link
Enter a service invite code from the home page, or open a meeting, workspace, or public-service link directly.
Sign in to create your own
Open the App, choose a use case and template, describe the outcome, then create an AI worker and launch it from Live.
Chinese and English UI
Switch between Traditional Chinese and English from the header. The choice applies across public pages, the App, and this guide.
Your first session
Choose an entry
Use onsite interpreting face-to-face, AI-assisted call for remote conversation, or the App for a persistent service.
Allow the required device access
Voice features need a microphone. Video, exhibit recognition, recording, or presenting may also need camera and screen-sharing access.
Complete one real interaction
Speak, join a call, or launch a public worker and verify that captions, audio, and device inputs work.
Sign in to save and manage
Create your own AI worker when you need to reuse, share, review activity, or change capabilities.
Use a current Chrome or Safari release. If audio does not start, check browser microphone permission and the operating system input device.
Manage every AI worker in one App
The signed-in App organizes every service by outcome. The home view shows recent Live activity, use-case entries, and your AI workers. The sidebar links to Create, My AI Workers, Live, Workflows, and Tools.
Home
See worker counts by use case, recent Live sessions, and direct paths back to the same worker or a new version.
My AI Workers
Your AI workers and meeting secretaries are collected in one place and grouped by the jobs they can complete.
Live workspace
Launch configured translation, interview, learning, and support workers and return directly to the live experience you need.
Workflows and Tools
Background agent work, review, schedules, and history live in Workflows. External tools and MCP servers appear under Tools when connected.
The top search finds AI, tools, or work. Account and language controls currently live in the header; the full personal settings page is still being completed.
Create from a use case and template
The creation wizard reduces technical setup to four steps. Use it for the first version, then open the advanced editor when you need detailed control.
Nine named templates
AI meeting secretary, onsite interpreter, AI interviewer, AI teacher, AI support agent, AI tour guide, AI companion, AI marketing assistant, and custom AI worker.
Outcome-based capabilities
Choose whether the worker should listen, speak, translate, teach, analyze, search, send updates, or summarize. The system maps outcomes to modules.
Advanced editor
Configure basics, real-time capabilities, work automation, plugins, and settings, including STT, dialogue, TTS, Avatar, memory, vision, and summaries.
Visibility and entry points
Where supported, choose private, public, or invite-only access and configure captions, reminders, tasks, and session presets.
The four wizard steps
Pick a use case
Start with translation, interview, learning, support, or a custom job.
Choose a template
Select the role closest to the desired outcome. You can adjust it after creation.
Decide what it should do
Name the worker and choose capabilities. Some templates also request job, course, brand-source, or support-knowledge details.
Preview and create
Confirm capabilities and entry points, then open the worker backstage or launch it in Live.
Open advanced settings from worker details after creation. Capabilities that require an external service, role data, or device permission show the required setup in the interface.
Configure once, launch repeatedly, keep improving
Worker details store role capabilities and runtime settings. Live launches real interactions, while activity records help you understand what actually happened.
Worker details and backstage
Review category, runtime settings, knowledge, usage, and activity. Edit courses, interview criteria, FAQs, memory, exhibits, or brand data according to role type.
Launch from Live
Start translation, interview, learning, or support from one page. Configured workers are shown first so you can begin interacting immediately.
Sharing and invitations
Depending on service type, share an invite code, QR code, service link, or meeting link with external participants.
Activity and sessions
Supported services can store recent Live activity, transcripts, recordings, summaries, and usage. Available data depends on service configuration and consent.
Recording, transcripts, notifications, and cross-session memory should be disclosed and managed according to your organization's retention policy.
Three ways to communicate across languages
Choose onsite interpreting, AI-assisted call, or an AI meeting secretary based on location, whether a link is needed, and whether you need a summary afterward.
Onsite interpreting
Use one device face-to-face. Pick the source and target languages, swap direction when needed, and start listening. The screen keeps transcript and translation history.
AI-assisted call
Create an ad-hoc remote room from the home page and share it. The call supports caption translation, voice translation, microphone, camera, and screen-sharing controls.
AI meeting secretary
Designed for recurring and multi-person meetings. Invite an AI assistant or interpreter for transcription, translation, summaries, and action-item organization as configured.
Languages and voice
Call captions can be untranslated or translated to Chinese, English, Japanese, or Korean. Voice translation can be toggled separately; actual voices depend on worker settings.
Start translating
Choose the situation
Use onsite interpreting in one location, AI-assisted call remotely, or create a secretary for recurring use and follow-up.
Set languages
Choose different source and target languages and enable voice translation if required.
Start listening or invite people
Allow microphone access and speak. For remote calls, share the room link with participants.
End and review
Return to the relevant service to review any supported transcripts, translations, summaries, tasks, or session records.
Onsite interpreting does not require inviting another person. AI-assisted call requires sharing a room link. Saving audio, summaries, and recordings depends on service settings.
AI teachers, course generation, and teaching tools
Learning and teaching cover public courses, AI voice interaction, classes, analytics, and integrated course generation from YouTube or slide decks.
Public learning and AI teachers
Choose a course and study by unit and segment. The AI teacher can explain, ask questions, listen, translate, and provide learning feedback as configured.
YouTube to interactive course
Enter a YouTube URL or video ID. The system obtains captions, runs speech-to-text when captions are missing, and analyzes segments and knowledge points.
PPTX / PDF to course
Upload a .pptx or .pdf. A background job parses slides, writes per-slide narration and knowledge points, then opens the result in the editor.
Teaching and analytics
Teachers can create courses, manage students and invitations, inspect progress and sessions, and open learner analytics where supported.
Create an AI course
Create an AI teacher
Choose the AI teacher template and configure name, language, voice, Avatar, and interaction capabilities.
Choose a source
Paste a YouTube video or upload a PPTX / PDF slide deck.
Let the background job run
Progress updates as segments are created. You can leave or switch devices and resume from the job or draft list later.
Edit, publish, and track
Review segments, narration, knowledge points, character, and visibility, then use course and teaching dashboards to track activity.
Long videos and slide decks can take several minutes or longer. Do not cancel just because you leave the page; return later to edit the completed background job.
From practice interviews to formal recruiting
Use the AI interviewer for individual practice or create a formal interview service with campaigns and candidate management. Depending on settings, sessions can retain transcripts, recordings, and evaluation data.
Structured interviews and follow-ups
Ask questions by job profile, segment, question set, personality label, and scoring dimension, with follow-ups based on the answer.
Recording, transcripts, and consent
Enable video recording, turn-by-turn transcripts, and pre-join consent independently when creating the service.
Campaigns and invitations
Create a campaign, share the candidate entry point, manage progress, and inspect sessions and results on supported management pages.
Evaluation and review
Generate configured scoring and feedback, then review available transcript or recording data. Fields depend on the template and service configuration.
Create an interview service
Choose the AI interviewer template
Start from the interview use case, name the service, and set the interviewer style.
Define the job and scoring
Add job information, interview segments, questions, and scoring rules.
Choose data retention
Enable recording, transcripts, and consent prompts according to the real purpose.
Invite and review results
Share the candidate entry point, then return to management for sessions, scores, and available records.
For formal interviews, clearly disclose the purpose, retention period, and audience whenever personal data or video is collected.
Put FAQs, knowledge, and escalation in one workflow
The AI support agent answers common questions from supplied knowledge, then routes situations that require a person according to escalation rules.
FAQs and knowledge base
Maintain FAQs, knowledge sources, and articles. Supported editors can add file or URL sources.
Reply and escalation policy
Define tone, response limits, and escalation rules so the worker knows when to answer, clarify, or hand off.
Runtime preview
Where available, inspect the settings that will actually reach the dialogue engine before publishing.
Activity and usage
Use supported session, transcript, usage, and service-state views to strengthen knowledge from real customer questions.
Create a support worker
Choose the AI support template
Create from the support use case and set the name, role, and visibility.
Add approved knowledge
Organize FAQs, articles, and supported file or URL sources.
Define escalation
Specify when the AI must not answer, needs confirmation, sends a notification, or creates a case.
Test before publishing
Verify answer sources and tone, then share the service and update knowledge from session evidence.
AI support should answer only from approved information. Update the knowledge source whenever prices, policies, or high-risk information change.
Delegate background work, review outputs, and schedule runs
Workflows split long-running agent work into Inbox, Pipelines, and History. Built-in Movie Editor flows process video, generate narration, or operate a website to record a tutorial.
Three video pipelines
Process an uploaded video, turn a script into narration, or create a website tutorial. Describe the request in conversation or select a pipeline directly.
Preflight and browser dry run
Website tutorials check the URL, narration, operations, login data, and demo assets. The dry run must pass before the background queue starts.
Outputs and review
Jobs can create video, audio, subtitles, and translated variants. Inbox shows running and review states with preview, download, publish, and discard actions.
Schedules and history
Pipelines supports daily or weekly routines, run now, enable, and disable. History keeps recent completed-work filters and records.
Delegate a Movie Editor job
Describe the job or choose a pipeline
Explain the video you want, or select uploaded video, script narration, or website tutorial.
Provide assets and output settings
Add video, URL, steps, narration, voice, intro/outro, aspect ratios, and subtitle languages as required.
Queue the background job
Website recording first runs a dry run. After submission, close the dialog and follow queued, running, and completed states in Inbox.
Preview, publish, or discard
Review video, audio, captions, and AI highlights, then choose an internal link, download, or a connected publishing platform.
YouTube, Facebook, and Instagram publishing works only when backend OAuth is configured and the user's account is connected. Otherwise use internal sharing or download.
Share files, messages, and calls
Workspace combines files, messages, and multi-person calls behind one shared entry point for team collaboration, file exchange, and meetings.
Shared spaces and passwords
Create a named share, copy a guest link, and set, change, or remove password protection.
Files and messages
Upload, download, and manage files while keeping text and image messages in the same collaborative context.
Multi-person calls
Start audio or video calls from supported workspaces and use microphone, camera, camera-flip, and screen-sharing controls.
Connect an AI secretary
Link a supported meeting secretary to provide transcripts, translation, summaries, and task organization in calls.
Create a shared workspace
Create a shared space
Name the workspace and decide whether its contents require a password.
Add content or an AI
Upload files, leave messages, or connect an existing AI secretary.
Share the guest link
Send the link and any password to participants. No application installation is required.
Collaborate and review
Exchange files, message, and call from one entry, then review any supported transcripts and summaries.
Anyone with the share link may be able to access workspace content. Use a password for sensitive material and review members and files regularly.
Tours, companionship, marketing, and Avatars
Beyond translation, interviews, learning, and support, the template marketplace offers AI workers for tours, companionship, and marketing. Each worker can use a 2D, 3D, or voice-only presentation where supported.
AI tour guide
Configure exhibit knowledge, guide persona, greeting, speaking style, and vision for museums, exhibitions, and destinations.
AI companion
Configure persona, topics, reminders, proactive outreach, care contacts, and memory policy for a tailored companion experience.
AI marketing assistant
Configure brand voice, products, CTAs, output formats, and knowledge sources to present products, prepare assets, and support marketing work.
Avatar presentation
Choose a 2D photo, 3D model, or voice-only presentation for the service and device. The Avatar handles appearance while the AI worker handles jobs, knowledge, and workflows.
Choose the closest worker in the template marketplace. If no template matches exactly, start with a custom AI worker.
What works immediately and what needs setup
Account, device, and external-service requirements vary by feature. This section explains the actual conditions for availability.
Available without sign-in
Public explore, selected public AI workers, onsite interpreting, and ad-hoc AI-assisted calls created from the home page.
Sign-in required
Creating and managing workers, saved activity, workflows, schedules, private courses, interview management, and workspace management.
Tool catalog is conditional
The Tools page has MCP/plugin entry points. Installed tools appear only after a real registry or worker plugin configuration is connected.
External publishing is conditional
YouTube, Facebook, and Instagram require platform OAuth configuration plus user authorization. The UI does not present them as connected when they are not.
The complete personal settings page and public tool registry are still being completed. Account and language controls are currently available from the header.
Real-time engine, skills, and data boundaries
Violet Avatar composes real-time voice, structured skills, background agent work, and Avatar presentation so use cases can share core technology while keeping distinct workflows.
Real-time voice pipeline
Microphone audio becomes text through speech recognition. The dialogue engine follows intent and flow, then TTS, captions, and the Avatar render the response.
Agent Skills and guardrails
Interview, learning, support, tour, and other roles use domain-specific skill fields and execution rules with input, output, and tool-use restrictions.
Memory, knowledge, and runtime
Role settings, knowledge sources, cross-session memory, and session content are managed separately. Supported services provide runtime previews before publishing.
Activity, usage, and access
Identity, visibility, invitations, consent prompts, and session settings determine who can use a service and what data is retained.
The real-time interaction path
Listen and recognize
Receive audio from a microphone or call and convert it into text.
Understand and act
Use the worker's skills, knowledge, workflow, and safety rules to choose a reply or tool action.
Generate and present
Produce text, captions, and speech through a 2D, 3D, or voice-only interface.
Record and improve
Subject to consent and settings, save activity, summaries, outputs, or usage for ongoing improvement.
Not every use case enables every capability. Follow the worker detail page, Live controls, and feature-status messages for the current configuration.