OPENAI announces its next generation of audio models that will give you a voice

2025 will be the year of AI and agents OPENAI He has moved his tokens in recent weeks to announce the GPT-4.5 or he himself yesterday his model of IA at the highest with O1-Pro. Today, Openai has announced new audio text and text audio models away in the API that will open the doors to developers to create AI agents with a more expressive, natural and more customizable voice.

The objective is to provide an audio model that can be used by AI agents and that developers can Personalize your voice or accent. If the advanced model of Chatgpt voice has different voices, what OpenAi has done is now to provide the same experience so that any developer can create a unique voice.

Two models come into play here, GPT-4O y GPT-4O Mini For transcription and which stands out to provide important performance in this type of task in order to minimize words of words, language recognition or efficiency compared to others from Openai.

For Openai, it was not easy to reach this step, because the two models are part of the new audio Exhaustive learning techniques Intermediary and with diversified and high quality audio data sets.

There are several very important shades in these new audio models such as their ability to Understand accents in the voiceReduce possible failures when recognizing voice and improvements in transcription capacity even when audio input includes noise, special accents and speed variety in speaking.

OPENAI also has announced the price Among these new audio models for AI and the GPT-4O transcription model, it has a cost of $ 6 for a million audio entry, $ 2.50 per million text entrance tokens and 10 dollars per million text tokens.

He GPT-4O Mini model It has a cost of $ 3 per million input audio chips, $ 1.25 per million text entrance tokens and $ 5 per million audio entry. The costs per transcription minute are 0.6 cents for GPT-4O, 0.3 cents for GPT-4O Mini and 1.5 cents for GPT-40 Mini TTS.

To better understand What new audio models mean From Openai, we collect the words from the same team of its announcement: “Faced with the future, our objective is to continue to invest in improving the intelligence and precision of our audio models and to explore the means so that the developers can incorporate their own personalized voices to create even more personalized experiences aligned with our safety standards.”