CineTTS
Integrated into Cinetwork, this feature generates high-quality neural audio for scripts to aid actor rehearsals.
Overview
A production feature integrated directly into the Cinetwork platform, converting film and TV scripts into multi-character audio streams using advanced neural text-to-speech.
Why it matters
Cinetwork supports major productions (TF1, Netflix, France TV). Actors on these sets need realistic rehearsal tools. Unlike robotic TTS placeholders, CineTTS provides natural, emotion-capable voices that help actors memorize lines and grasp the rhythm of a scene before stepping onto the set.
Key capabilities
- Neural Voice Quality: Uses actor-like neural voices (via ElevenLabs) to create a pleasant listening experience, far superior to standard OS accessibility voices.
- Consistent Casting: Allows manual assignment of specific voices to characters, maintaining audio consistency across episodes.
- Rehearsal Mode: Features “smart mute” capabilities, allowing an actor to silence their own character’s lines to practice against the audio.
Technical approach
- Smart Parsing: Extracts dialogue and character metadata from PDF/Final Draft inputs to assign correct voice IDs dynamically.
- Streaming Delivery: Delivers mixed audio directly within the secure web application to prevent file leaks while ensuring instant playback.
- Dialogue Pacing: Leverages external model logic to handle natural pauses and timing between cues.
Role
Lead Developer for the feature integration. Managed the pipeline from script text extraction to audio generation and player UI.
Outcomes
- User Validation: Qualitative feedback from dozens of actors confirmed the “natural” quality significantly improved their rehearsal process.
- Rapid Delivery: Prototyped and deployed as a beta feature ahead of schedule.
Tech stack
TypeScript, Next.js, ElevenLabs SDK