Admin sign-in
Enter the admin password to continue.
Application User · Practice
Voice Catalogue
Enable voices for applications and review which applications use them.
- MarinOpenAI · Voice ID: marin
- CedarOpenAI · Voice ID: cedar
- AlloyOpenAI · Voice ID: alloy
- AshOpenAI · Voice ID: ash
- BalladOpenAI · Voice ID: ballad
- CoralOpenAI · Voice ID: coral
- EchoOpenAI · Voice ID: echo
- SageOpenAI · Voice ID: sage
- ShimmerOpenAI · Voice ID: shimmer
- VerseOpenAI · Voice ID: verse
- TizianaProvider: HeyGen · Speech engine: ElevenLabs · Model: eleven_multilingual_v2Provider voice ID: ebbc053d046742c582f273f1334c58d8Approved for local Italian Learner rehearsal
Existing Fluently Italian Learner recording. Rehearsal approval does not enable GPT Live.
Direct ElevenLabs voices
LOCAL VOICE STUDIO · V0.3
Voice Studio
Shape a character. Start a conversation.
Application permissions are server-owned and read-only. Scene, character and language preferences save locally in this browser; publishing and application-wide persistence are not implemented.
Effective application configuration
Practice stages are independent of conversation modes. My Language uses L1; Conversation uses L2. Live Rehearsal is entered from a matching prepared script with explicit paid authorization. Live transcription is enabled for local testing. Ingle and Fluently support explicit Rehearsal translation preparation. Persistent recording, stored playback and Shared Help remain unavailable. Scripted Rehearsal uses temporary browser recordings only. Live-provider verification remains manual.
Application selection is a local administrator test context, not authenticated application identity.
Application Language Policy
Local draft only. This does not change server permissions or publish configuration. User activities always use the server-owned effective policy.
Settings marked planned are declarations only. Publishing, database storage, recordings, playback and recorded attempts are not implemented.
Characters
Create a Host, choose their voice, then use them in your scenes.
Create your first Character to choose a Host for your scenes.
My Scenes
Scenes and practice context saved on this device.
No saved scenes yet.
Scenes for your users
Shape a scene with your Host, review it, then make it available.
Describe your scene
Your microphone stays off until you choose I’m ready. End saves your conversation for review.
Review your scene
Make it yours. These directions guide the Host across the three conversation perspectives.
This scene is available to your users.
My Words
Words from the Essentials you hear and the scenes you practise. Understand them one at a time.
Your words will grow from the Essentials you hear and the scenes you practise.
Keep exploring
Ask about any part. Your host will explain at your pace.
Your microphone stays off until you choose I’m ready.
Your scene
Save keeps your scene on this device and translates it for practice. Translation uses AI; audio starts only when you choose Hear.
Application User Preview — no provider calls until Start Live Test. GPT Live usage incurs cost.
History
Ready when you are
Conversation
Live transcript
In-memory only. ASR can make recognition errors. Host text is generated speech, not proof that every word was heard. Input transcription adds provider usage charges.
Actual L2 speech — separate from prepared material
Hiding does not delete. Ended transcripts remain until refresh or the next session. Timestamps indicate client observation; language is the configured language, not detected language.
Translate
Review original L1 conversation
End a My Language conversation and review its finalized transcript first. Only finalized, non-failed turns are included; provisional fragments are excluded.
Review Conversation
Only the rehearsal version changes. Join only fragments of the same response. Omitting a Host line is your explicit decision, not an automatic judgement.
Conversation for translation
Prepare Translation sends the finalized conversation to OpenAI and incurs a small AI charge. Nothing is translated automatically. Cached results are reused when available.
Scripts stay in page memory only. The server cache is bounded and temporary; restart or eviction can require paid preparation again. No database or browser storage.
L1 is the unchanged source transcript. L2 is translated practice material, not actual learner speech. Actual L2 speech appears separately in the live transcript.
Practice
DEMONSTRATION — prepared sample, not your conversation
Listen
Practise
Rehearsal Complete
Play My Conversation combines Julie’s Marin audio with accepted recordings. Play Pro Conversation combines Julie’s Marin audio with Tiziana’s prepared Italian Learner speech and does not require Learner recordings. Missing Tiziana lines are generated only after you select Hear Pro or Play Pro Conversation; these explicit actions may incur HeyGen charges. Pro audio is available only when the local development integration is enabled.
Full prepared script — optional reference
Live Rehearsal — paid AI session
Live Rehearsal and Repeat incur Realtime usage charges, plus input transcription charges where enabled. No new translation is generated.
Your suggested response
What you actually said (L2)
Guide position and recognition diagnostics
Only safe diagnostic categories are shown; no raw provider error messages or secrets are logged.Complete transcript and original prepared script
Actual L2 speech — separate from prepared material
Original L1 / prepared L2 guide
During Live, future prepared answers remain hidden until requested or explicitly confirmed. End restores the complete script.
Speed applies to the next Host line. Use Hear again to replay it at your chosen speed.
Take your time. Your microphone stays off until you choose I’m ready. Choose I’m done when you finish speaking.
Rehearsal assistance
Rehearsal assistance
Position uses exact Host-text matching or your explicit confirmation. Paraphrases and detours can leave it unknown. Learner word counts never advance it. Future prepared responses stay hidden during Live.