In March 2024, OpenAI unveiled a “small-scale preview” of Voice Engine, an AI tool capable of cloning a person’s voice using just 15 seconds of speech. Nearly a year later, the tool remains in limited preview, with no confirmed release date—or even certainty that it will launch at all.
The hesitation to roll it out widely may stem from concerns about potential misuse or an effort to sidestep regulatory scrutiny. OpenAI has faced criticism in the past for prioritizing flashy releases over safety, often rushing products to market ahead of competitors.
Testing, But No Release Plans
An OpenAI spokesperson told TechCrunch that the company is still testing Voice Engine with a select group of “trusted partners” and learning from their experiences.
“We’ve been excited to see the different ways it’s being used—from speech therapy and language learning to customer support, video game characters, and AI avatars.”
However, despite its potential, Voice Engine has faced multiple delays and shifting launch plans. Initially scheduled to be integrated into OpenAI’s API on March 7, 2024, the tool was supposed to be offered to 100 selected developers, prioritizing projects with social benefits or responsible innovation. OpenAI had even trademarked and priced it:
- $15 per million characters for standard voices
- $30 per million characters for HD-quality voices
But at the last moment, OpenAI postponed the release, ultimately unveiling the tool without a sign-up option. Only a small group of developers who had been working with OpenAI since late 2023 gained access.
A Project Years in the Making
According to OpenAI, Voice Engine has been in development since 2022. The company even demonstrated the technology to high-level policymakers in 2023, highlighting both its potential benefits and risks.
One of its early testers, the startup Livox, aimed to use it to help people with disabilities communicate. While Livox CEO Carlos Pereira praised the tool’s quality and multilingual capabilities, he noted that its requirement for an internet connection made it unsuitable for many of their users.
“It is the most impressive and easy-to-use tool for voice creation I’ve seen,” Pereira told TechCrunch. “We hope OpenAI develops an offline version soon.”
Despite testing the tool, Pereira hasn’t received any updates about a public launch or potential pricing. For now, Livox hasn’t been charged for its usage.
Safety Concerns and AI Voice Cloning Risks
One key reason OpenAI may be holding back Voice Engine is the risk of misuse, particularly in an era of AI-generated deepfakes. The company acknowledged these concerns in a June 2024 blog post, suggesting that fears of election-related abuse contributed to the delays.
To mitigate risks, OpenAI has introduced several safety measures:
- Watermarking to trace generated audio
- Explicit consent requirements for voice cloning
- Mandatory AI disclosure to audiences
However, enforcing these rules at scale remains a significant challenge. OpenAI has also hinted at plans for “voice authentication” and a “no-go” list preventing the creation of voices resembling public figures—two ambitious security measures that, if poorly executed, could further tarnish its reputation.
The Future of Voice Engine
AI voice cloning has become one of the fastest-growing scams of 2024, with fraudsters bypassing bank security checks and spreading deepfake content of celebrities and politicians across social media. As legal frameworks struggle to keep up, companies like OpenAI must tread carefully.
Whether Voice Engine will ever be fully released remains uncertain. OpenAI has repeatedly stated that it may keep the project small in scope. But one thing is clear: for safety, regulatory, and reputational reasons, this has become one of the longest product previews in OpenAI’s history.
Frequently Asked Questions
Why hasn’t OpenAI released its voice cloning tool yet?
OpenAI has cited concerns about potential misuse, regulatory scrutiny, and the need for further testing with trusted partners before making Voice Engine widely available.
Is OpenAI still developing Voice Engine, or has it been abandoned?
OpenAI continues to test Voice Engine with a small group of developers and has not officially canceled the project. However, there’s no confirmed launch date.
What are the risks associated with AI voice cloning?
AI-generated voices can be used for deepfakes, scams, and fraud, making it difficult to verify authenticity. OpenAI has discussed implementing safety measures like watermarking and speaker verification.
Who currently has access to OpenAI’s voice cloning technology?
Only a limited number of trusted partners and developers have access to Voice Engine, including companies working on speech therapy, language learning, and accessibility tools.
Will OpenAI ever launch Voice Engine to the public?
OpenAI has not confirmed a public release but has hinted that Voice Engine may remain a restricted tool to avoid ethical and security risks.
Conclusion
A year after its initial announcement, OpenAI’s Voice Engine remains in limited preview, with no clear timeline for a public release. While the tool’s advanced voice cloning capabilities have promising applications in accessibility, gaming, and customer support, concerns over misuse, deepfakes, and regulatory scrutiny seem to have stalled its launch. OpenAI continues testing with trusted partners, but whether the tool will ever be widely available remains uncertain. For now, Voice Engine stands as one of OpenAI’s longest-running experimental projects, caught between innovation and ethical responsibility.
