A year later, Openai still did not leave the sound cloning vehicle

At the end of last March, Openai declared the EU service “a small-scale preview”, Sound engineHe claimed that the company could clon the voice of a person with a speech for 15 seconds. About a year later, the instrument remains prejuded and Openai did not give a sign that he could start work or start everything.

The company may not want to disseminate the service widely, use abuse fears, but can make an effort to invite the regulatory study. Openai has historically Defended Prioritize “Bright Products” to the security account and Rush Releases beat the rival firms to the market.

In the statement, an open spokesman told Techcrunch that the company continues to test the vote engine with a restricted set of “reliable partners”.

“We will be able to learn how (we) use technology, we will be able to learn the model of the model,” said Press Secretary. “We were pleased to see the different ways to use spoken therapy, language learning, customer support, video game characters, to the AI ​​avatars.”

Retreated

Openai’s Text Speech API, as well as the sound engine that strengthens the sounds of Chatgpt Sound modeCreates natural sound speech closely similar to the original speaker. The tool converts written symbols into speech, only limited by certain guards according to the content. However, it was the subject of changing the delay or not released windows.

As a statement in June 2024 Blog PostTaking into account the sound engine model, different sounds, emphasis and speaking styles, it learns to predict the most likely sounds that a dynamic will develop a certain text transcript. After that, the model can only create “talk words”, which reflects the text of the text and the types of speakers loudly.

Openai was first intended to present a sound engine first to call API on March 7, 2024, in the framework of a blogging blogging by Techcrunch. The plan “Giving a group before a valid debut, the Devs, which gave a group of” social benefits “, was a priority for construction applications or” innovative and responsible “use of technology. Openai even had brand And $ 15 million for millions of $ 15 per million for “Standard” sounds and a million dollars for millions of dollars for “HD quality”.

Then, the company postponed the announcement in the eleventh hour. Openai ended the sound engine without the entrance option after a few weeks later. Openaii would be limited to a cohort of 10 more devins, where the tool started working in late 2023.

“We hope that the deployment of synthetic sounds and how society can adapt to these new opportunities” Openai Sound engine announcement wrote in a blog post At the end of March 2024. “According to these conversations and these small-scale tests, we will make a more informative decision on the scale and how to place this technology.”

Long in the work

According to Openai, the voice engine since 2022 has been in business since 2022. Company claims In order to demonstrate their potential and risks, 2023 did the “global politicians of global politicians” in the summer of 2023.

Several partners today have access to a sound engine, including the starting facility, which allows people with disabilities to communicate more naturally. CEO Carlos Pereira, TechCrunch-A Talking TechCrunch, Livox, an online request of the mediator (many of the many do not have Internet), technology found “really impressive”.

“The possibility of sound quality and sounds to speak in different languages, especially for our customers,” Pereira “creates the most effective and easy to use.”

Pereira has launched a possible sound engine and says it does not see any signs planned to start charging for the service. So far, Livox did not have to pay for use.

In the post of 2024, Openai, one of the dismissal of the sound engine, the United States said that the US election has the potential. By discussions with discussions, watching the sound engine, including several reduced security measures, including a number of voting.

Before using the sound engine, the creators must obtain “open consent” from the original speaker and must “open disclosures” to the sounds of the EU. The company did not say how these policies were applied. Thus, it can prove that it is very difficult for a company with OpenIA resources.

In the blog posts, Openai thought that he had hoped to check the “voice identification experience” and “voice identification experience” that prevented the creation of sounds that prevented prominent sounds. Both are a technologically ambitious project and reflect the poor reflection of a company that is often charged with them Sidelling Security Initiatives.

Effective filtering and identity check, responsible sound cloning converts fast basic requirements for technological releases. AI sound cloning was the third fastest growing scam of 2024, According to one source. Led fraud and Bank security inspections Fight to fight to pass as privacy and copyright laws. Harmful actors used the sound cloning to create mischievous depths Celebrities and PoliticiansAnd these depths are spread as wildflies on social media.

Openai can leave the sound engine next week – or never. The company repeatedly said that the service should be small. But one thing is clear: the causes of optics, the cause of safety or both, the limited preview of the sound engine, turned into a longest time in Openai history.