Audio Drama Voice Studio

Challenges in Multi-Role Voiceover

When producing audio dramas, audiobooks, or advertisements with multiple characters, manually assigning unique voices and recording lines for each role is inefficient and error-prone. Traditional methods involve tedious adjustments for voice matching and sequencing, increasing post-production complexity.

How the Skill Works

This skill automates multi-role voiceover via a toolchain. Core capabilities include character table management, voice assignment, and line synthesis, with key steps as follows:

Core Capabilities and Key Steps

1. Character Table Construction: Read the script, identify all speaking characters and narrator, build a character table with age, personality, etc., ensuring each role has a distinct record.
2. Voice Assignment: Assign a unique voice ID to each character, maintaining consistency throughout. For cloning, use beatra.voices.clone and require explicit authorization confirmation.
3. Line Splitting and Synthesis: Split the script into individual lines, each synthesized with the assigned character's voice via beatra.speech.synthesize. Synthesis is billed per weighted characters: Chinese characters count as two, others as one, with independent calls per line.
4. Delivery and Review: Deliver numbered audio files in script order, annotated with characters, and report duration, size, MIME type, and URL. Review ensures voice, pronunciation, and sequence are correct.

Key Process Details

• Character setting and script splitting are free; only synthesis is paid, with pricing from beatra.models.list real-time cards.
• Each line has a unique client_request_id; changes only require re-recording that line, preserving delivered lines.
• For fault recovery, use beatra.tasks.get to poll task status and recover only failed lines.

Scope and Limitations

Applicable Scenarios

• Audio dramas and audiobooks production
• Novel adaptations for narration
• Dual or roundtable dialogue programs
• Dialogue-driven courses or training
• Two-character script ads
• Interactive stories

Inapplicable Scenarios

• Single-person narration: Use ai-audiobook-narration.
• Single-person podcast: Use ai-podcast-voiceover.
• Short-form voiceover: Use short-form-voiceover-audio.
• Single-voice work: Use beatra-ai-voice-studio.

Considerations

• Only handles voiceover recording; does not mix, crossfade, master, or assemble timelines.
• Voice cloning requires separate authorization confirmation and one-time payment.
• Use the bundled scripts/mcp_client.py for operations; do not configure host Beatra Connectors.
• First use requires registration via beatra.installations.register, with automatic updates that fail open without affecting execution.