AI Voice Style Transfer Engineer

Ulara AI

Hybrid · $160k – $200k / year · Full-time · Posted 4 months ago

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

About the role

Voice style transfer sits between voice cloning and text-to-speech: given a target speaker's characteristics — their timbre, pace, prosodic patterns, breathiness, resonance — synthesize speech that carries those stylistic traits without requiring hours of training audio. A brand should be able to define a 'voice identity' from a 3-minute sample and have that identity applied consistently across all their generated content. Our voice platform processes 40 million minutes of synthetic speech per month for media, advertising, and enterprise clients. Style transfer is the next major capability: brands want consistent voice identity across product demos, marketing content, and customer interactions without requiring dedicated voice talent for every piece of content. You'll lead the voice style transfer research and implementation: designing disentangled representation models that separate style from content, training transfer models on our 1,200-hour voice corpus, and implementing the style conditioning interface that lets clients define and apply voice identities. The evaluation challenge is real — perceptual quality assessment requires listening panels, not just objective metrics, and you'll need to build both.

Responsibilities

Design disentangled voice representation models separating speaker style from linguistic content
Train style transfer models enabling target speaker characteristics from 3-minute reference samples
Build style conditioning interface for brand voice identity definition and application
Implement perceptual quality evaluation combining listening panel studies and objective metrics
Develop style consistency monitoring ensuring voice identity preservation across content types
Optimize inference latency for real-time style transfer at production scale

Requirements

6+ years in speech synthesis, voice ML, or audio deep learning research
Deep expertise in neural TTS, voice conversion, or speaker representation learning
Strong Python and PyTorch skills with audio signal processing experience
Familiarity with disentangled representation learning approaches
Research publication record in voice AI or speech synthesis preferred