Help & Voice Studio Guide
Start with a preview
The homepage browser preview requires no account. Voices depend on your browser and installed language packs. It cannot save MP3 or WAV files. For downloadable audio, open AI Studio and sign in.
Account and voice services
Create an account and wait for admin approval if requested. The studio shows whether Azure and ElevenLabs are connected. If no service is connected, ask the administrator to complete server setup; cloning cannot run through browser speech synthesis.
Clone a voice
Record 30 seconds or upload a clean 25–180 second recording. The browser converts supported audio to mono WAV. A single speaker and 60–120 seconds of clean speech generally provide a better input than a short, noisy sample. Confirm permission and create your private clone. A verification or review state requires admin/provider action. Rename the clone and save effect defaults with Edit.
Bangladeshi voices
The catalog includes Mature Male, Mature Female, Young Boy, Young Girl, Teen Age Boy and Teen Age Girl categories. The two adult categories are mapped to Azure’s Bangladesh locale voices, Pradeep and Nabanita; maturity and accent suitability should be checked by listening. Young/Teen categories require approved provider voices before use. The download package does not include six authentic recorded samples.
Long-form generation
Choose a ready voice, enter up to 50,000 Unicode characters, and generate. Text is split at sentence boundaries where possible; very long sentences are divided at word boundaries or the model-safe limit. Completed chunks are retained. Leave the tab open, or schedule the server worker. Use Continue on a queued job after reopening the studio. The finished WAV uses a short, configurable crossfade.
Stories
Add completed generations, assign one of eight tracks, drag a clip horizontally and set its start, trim in, trim out and volume. Overlapping clips mix together. Save your story to your account. Preview and export run in your browser with a 30-minute render limit. Use smaller stories on phones or memory-constrained devices.
Effects
Adjust pitch, speed, reverb, delay, compression and volume while listening to the latest audio. Pitch uses resampling and changes duration. Save presets or store the current effect settings as a private voice’s defaults. Original WAV download keeps the generated audio unchanged; Export with effects and story export render processed WAV files.
Failures and credits
Provider errors may indicate missing permission, unavailable credits, verification or a timeout. A failed request can still consume provider credits. Check usage before retrying. Quota is reserved at submission and again for unfinished text on an explicit retry. No automatic paid retry is performed.
Administration
Admin accounts can approve or suspend users, set clone limits to two or three, set daily character quotas, connect catalog voice IDs and resolve provisioning states. Keys are configured on the server. The deployment package includes the installation guide, CLI installer, worker, maintenance task and tests.