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:
- A user uploads a prerecorded video.
- Our application extracts the audio.
- We can also provide an accurate transcript of the spoken content if needed.
- The AI processes the speech while preserving:
- the speaker’s original voice identity,
- emotions,
- speaking style,
- pacing,
- pauses,
- timing,
- overall delivery.
- The only desired change is improving pronunciation and reducing the foreign accent.
- 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:
- Is this technically achievable with the current OpenAI speech models or APIs?
- Would this require combining multiple models, or is there a speech-to-speech workflow that already supports it?
- Are there any recommended APIs or models that preserve voice identity while improving pronunciation?
- 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!