As global content consumption increases and demand for non-English content exceeds demand for English-language movies and shows, IMAX leverages AI to expand localization of original content.
The entertainment and media industry grew 5% to $2.8 trillion in 2023, according to a PwC report. The industry is expected to continue expanding, but the compound annual growth rate is expected to remain at nearly 4%, or $3.4 trillion, over the next five years. Non-English content is also rapidly growing in English-language markets such as the US, UK, Australia, and Canada. Last year, Netflix reported that viewership of non-English content in the UK had increased by 90% over the past three years.
Considering all this, IMAX is considering localization using AI to attract more attention.
The Canadian production company, known for its large-scale theaters and immersive movie experiences, on Monday partnered with Dubai-based startup Camb.ai to use AI voice models to translate original content, including documentaries. announced.
Camb.ai has already deployed AI dubbing and voice translation for live sporting events and leagues such as the Australian Open, Eurovision Sport, and Major League Soccer, using the Boli model for voice-to-text translation and voice emulation for voice-to-text translation. offers Mars. The models are available through the startup's DubStudio platform, which supports 140 languages, including a variety of low-resource languages for which no significant data exists on the internet.
“Companies like OpenAI and Anthropic have a different vision for society,” Akshat Prakash, co-founder and CTO of Camb.ai, said in an exclusive interview. “They're trying to build models that are very horizontal and can cover a wide range of tasks… We don't have to do that at all. Some of our models have less than 100 million parameters. , some are very specific.”
Prakash is a former Apple engineer who helped develop Siri's AI and ML models and co-founded Camb.ai with his father, Avneesh Prakash, last year.
“We are generations apart and grew up in India. After 30 years, we faced the same language challenges,” the CTO told TechCrunch.
He said Camb.ai pre-trained 70% of its models using commercially available, academically licensed datasets. The remaining 30% includes fine-tuning data obtained through early partners deploying models for AI-based dubbing and translation.
“What we are not doing, and what we have scrupulously avoided completely, is potentially trashing the internet,” Prakash argued. “Some companies think they can get away with it because they're building consumer-facing apps and tools, so they think it's OK to scrape something like 10 petabytes of the internet.”
Camb.ai uses a “three-layer” approach to provide AI-based translation. It consists of a foundation layer of Boli and Mars models, an infrastructure layer that hosts these AI models, and a front-end DubStudio platform.
Unlike other AI-based models, Camb.ai's Boli takes input speech tokens and produces output text tokens in the translated language, preserving nuance, Prakash claims. Once Boli generates text, Mars uses the same audio input signal to translate that text into speech, creating a real-world audio performance that includes ambient sounds, such as a background score of cheering spectators at a sporting event. Capture.
Prakash told TechCrunch that Camb.ai's technology enables voice translation in up to 10 languages simultaneously with a 20-30 second delay, which can be covered by a 30-40 second delay for streaming or broadcasting.
IMAX will roll out AI translation in stages, starting with high-resource languages. Implementation follows internal testing of Camb.ai's technology against original content.
“While we are in the early stages of our partnership, we continue to work together to further explore its potential and how it can best move us forward,” said Mark Welton, president of IMAX Global Theatres. said.
Although he did not provide details, Welton suggested that the introduction of AI would lead to savings in translation costs.
Camb.ai currently has a team of 50 people. In February, the company raised a $4 million seed round led by Courtside Ventures. Prakash told TechCrunch that the company is closing on a larger pre-Series A round to expand its reach and number of employees.