British accent and pronuciation in gpt-4o-mini-tts

API
1

Does anyone know how to evoke British accent in gpt-4o-mini-tts? Including accent in instructions doesn’t seem to work.

3 Likes
2

“Anyone” does.

Good enough that you know it is not a Californian.

There’s voices a bit more UK, on openai.fm you can see different styles of style, and then you can have an LLM look for words that need an American accent tweaked out of them.

The best is just a different TTS that is simply “EN_UK Female 2”

1 Like
3

This is going to sound facetious. Have you asked ChatGPT?

4

Nothing groundbreaking there:

"The GPT-4o-mini-TTS model doesn’t officially support explicit accent selection or customization, including a British accent. Its default accent is neutral American English.

However, you can slightly influence its pronunciation by carefully crafting the text you input, using British vocabulary, phrasing, or spelling. Still, these techniques won’t reliably produce an authentic British accent, as the voice synthesis remains predominantly American-accented."

BTW Ballad sounds most British in English. Funny how different these voices bahave in different languages. openai.fm

6

That is what a different post on here said too about Ballad.

Here is what I got from chatgpt:

Here you go — a TTS prompt input phrase with a voice clone style prompt for a highly exaggerated London Cockney accent:

Voice Style Prompt (for voice cloning or TTS engine):

Use an exaggerated East London Cockney accent. Drop H’s at the beginning of words (“hello” becomes “’ello”), replace “th” with “f” or “v” (“think” → “fink”, “brother” → “bruvva”), use glottal stops for T’s (“bottle” → “bo’le”), and stretch vowel sounds with dramatic flair (“time” → “toime”, “night” → “naaaaht”). Add cheeky charm and street-smart swagger—like a market trader tryin’ to flog knock-off watches or spillin’ secrets at the pub. Your tone should be bold, fast-talking, mischievous, and confident.

Sample TTS Input Phrase (to test the voice):

“’Ave a butcher’s at this, mate—brand new, top o’ the line, proper tasty stuff, right? You won’t find nothin’ like it down the market, I tell ya! Grab it now or you’ll be gutted, 'cause once it’s gone—it’s gone, innit?”

Let me know if you want a variation (like posh London or gangster grime style).

All that to say that it looks like the phrasing is just as important as noting the accent. Certain word context trigger more pronounced accent. Suppose it makes sense. I’m in Texas trying to use phrase like “Howdy, ya’ll” in a British accent probably isn’t the best example.

I checked the documentation trying to experiment just out of curiosity on openai.fm Given their examples and how they use Tone and Intonation I figured maybe it would be as simple as updating "Accent: Use an strong British accent similar to someone from Scouse.

For intelligent realtime applications, use the gpt-4o-mini-tts model, our newest and most reliable text-to-speech model. You can prompt the model to control aspects of speech, including:

  • Accent
  • Emotional range
  • Intonation
  • Impressions
  • Speed of speech
  • Tone
  • Whispering

Our other text-to-speech models are tts-1 and tts-1-hd. The tts-1 model provides lower latency, but at a lower quality than the tts-1-hd model.

DuckDuckGo said this which may be helpful. Something like Scouse may be less dependent on phrasing where Geordie indicates it uses unique vocabulary.

The strength of a British accent can vary widely depending on the region. Some of the accents often considered to be among the strongest include:

  1. Scouse - The accent from Liverpool, famously associated with The Beatles and known for its distinctive intonation and pronunciation.
  2. Geordie - The accent from Newcastle and surrounding areas, characterized by unique vocabulary and pronunciation.
  3. Yorkshire - The accent from Yorkshire, which can vary significantly within the region but is often noted for its broad vowels and distinct phrases.
  4. Cockney - Traditionally associated with East London, this accent features unique rhyming slang and a distinctive way of pronouncing certain consonants.
  5. Brummie - The accent from Birmingham, which has a unique sound and is often recognized for its vowel pronunciation.

In the end, I wasn’t impressed with any of the results. For me everything I have experimented has come out more Australian sounding then anything. I will keep experimenting though. I enjoy a good challenge.

Is there a particular accent you are going for like from a TV Show or Movie. My immediate thinking was the show Shameless (UK).

1 Like