Can OpenAI speech models preserve voice identity while reducing a foreign accent in prerecorded videos?

Hello everyone,

I’m a software engineer building an AI-powered application, and I’m trying to understand whether this workflow is technically possible using OpenAI’s current or upcoming speech models.

Our goal is not voice cloning, dubbing, translation, or text-to-speech.

Instead, we want to preserve the speaker completely while improving only the pronunciation.

The intended workflow looks like this:

  1. A user uploads a prerecorded video.
  2. Our application extracts the audio.
  3. We can also provide an accurate transcript of the spoken content if needed.
  4. The AI processes the speech while preserving:
    • the speaker’s original voice identity,
    • emotions,
    • speaking style,
    • pacing,
    • pauses,
    • timing,
    • overall delivery.
  5. The only desired change is improving pronunciation and reducing the foreign accent.
  6. The processed audio is merged back into the original video.

The goal is that the person still sounds like themselves, with the same emotions and timing, but with a much more natural accent.

I’ve already contacted several companies working in speech AI (Krisp, Sanas, Deepgram, Speechmatics, Rime, and others), but I haven’t received a clear technical answer about whether this exact workflow is currently supported.

My questions are:

  1. Is this technically achievable with the current OpenAI speech models or APIs?
  2. Would this require combining multiple models, or is there a speech-to-speech workflow that already supports it?
  3. Are there any recommended APIs or models that preserve voice identity while improving pronunciation?
  4. Is this something OpenAI is actively exploring or planning to support in the future?

I would really appreciate any technical guidance or recommendations.

Thank you!