Deepdub

Deepdub is an Israeli company that develops an AI voice layer used for media localization and for voice-enabled AI applications. Its core offering is an end-to-end AI dubbing and voiceover platform that plugs into the post-production workflow of content owners, combining generative speech models with human-in-the-loop work by native-speaking adapters and an in-house production team of project managers, linguists, dubbing directors, sound engineers and casting directors. The platform ingests video, uses automatic speech recognition to produce transcripts in more than 130 languages, applies natural-language segmentation and speaker identification, and generates dubbed audio using emotional text-to-speech (eTTS), speech-to-speech conversion, accent control and voice referencing, in which an existing recording guides generation of a new performance in another language.

The company sells through two delivery models: a managed service in which Deepdub takes ownership of a customer's localization needs, and Deepdub Go, a SaaS/virtual-studio product that gives post-production teams direct control over dubbing and voiceover, including accent fine-tuning and emotional voice selection. A licensed voice bank with full commercial and broadcast rights lets clients choose speakers and speaking styles, and the company has a beta royalty program intended to compensate voice actors whose voices are used. More recently Deepdub has extended the same speech stack to conversational and agentic AI, offering a voice API for AI agents positioned around long-form stability and multilingual output, with published use cases in customer service and sales and vertical framing for property management and debt collection.

Deepdub was co-founded in 2019 by brothers Ofir Krakowski (CEO) and Nir Krakowski (CTO) and is headquartered in Tel Aviv, with staff also in the United States. It maintains a public GitHub organization with open-source and integration projects, including an audio manipulation library, a Python API client, an agent plugin, and forks of voice-agent frameworks such as LiveKit Agents, TEN Framework and Pipecat.

Founding story

Ofir Krakowski, drawing on roughly three decades of computer science and machine learning experience including leading machine learning work in the Israeli Air Force, concluded that content is created in a single language and that translation and dubbing are convoluted, costly and slow, leaving most content unavailable across languages. He and his brother Nir, both entertainment enthusiasts, founded Deepdub in 2019 to apply AI to audio-visual localization, deliberately targeting the hardest case — major theatrical releases — on the premise that solving it would allow the company to handle projects of any size.

Business model

Deepdub sells AI localization to enterprises through two models: AI-driven, human-assisted managed dubbing services delivered by its in-house production organization, and self-serve/SaaS access via Deepdub Go and a voice API for AI agents, which offers a free trial tier alongside sales-led enterprise engagement. Voices in its voice bank are licensed with full commercial rights for broadcast use, and a beta royalty plan is intended to pay voice talent whose voices are used.

Revenue comes from AI dubbing and voiceover services sold to content owners and distributors, subscription/self-serve access to the Deepdub Go platform and voice tools, and API access for voice-enabled AI agents.

Traction

Deepdub reports dubbing thousands of hours of content into languages including English, Latin American Spanish, Portuguese, German and Hindi, and support for over 130 languages (30 as of mid-2023). Named work includes the Portuguese series "Vanda" for Legendary on Hulu, dubbing of entire channels on Pluto TV, the film "Every Time I Die" on Netflix, the Hebrew series "Hatufim" dubbed into English, theatrical releases, top-tier series, FAST channels and library content, plus a multi-series partnership with Topic.com. The company said it was the first to dub an entire feature-length film into Latin American Spanish using AI voices and works with multiple Hollywood studios. Headcount was over 30 in May 2023.

Latest developments

Announcements listed by the company include a partnership with Jeen Talk to bring its voice AI to enterprise contact centers (June 2026), the introduction of what it describes as the first agentic dubbing co-worker (April 2026), and support for Channel Oishii's North American TV launch (April 2026). The website's primary positioning has shifted toward a production-grade voice API for agentic AI alongside media dubbing, and it promotes its Phantom X 3.2 model's expressivity benchmark result.