How do you make a Hollywood star sound like Diana, Princess of Wales?

How do you set about transforming Kristen Stewart, the gruff-voiced, frequently mumbly, famously introverted Californian star best-known to the world as Bella Swan from Twilight, into Diana, Princess of Wales? The makers of Spencer have all the usual answers, in terms of costuming, comportment, hair and make-up, and Stewart has bucketloads of acting talent to tie all that together.

But they also have the voice to worry about – and it’s here that the dialect coach William Conacher comes in. It was Conacher who helped Emma Corrin to perfect Diana’s voice for The Crown (and she spoke glowingly of him before winning the Golden Globe for Best Actress earlier this year). That was perhaps an easier task given Corrin’s background - born in Royal Tunbridge Wells, boarding school-educated - yet it was Conacher, too, who coached Naomi Watts for the ill-fated 2013 Diana film. Watts, who was born in England but raised in Wales and Australia, must have been a sterner test.

“I have a lot of Diana in my history,” Conacher admits, when I speak to him by phone, even if his list of credits over the past decade-and-a-half extends far beyond her, to the likes of Cillian Murphy’s period Brummie in Peaky Blinders, or Rami Malek’s Oscar-winning Freddie Mercury in Bohemian Rhapsody, as well as his villain Lyutsifer Safin in No Time to Die.

When we talk, though, he’s en route to Elstree Studios, for the latest day’s work on Season 5 of The Crown, where he’s assisting Imelda Staunton, who’s taking over as the Queen, and Elizabeth Debicki, who’s replacing Corrin in the role of her most famous daughter-in-law.

Conacher works exclusively with actors: he trained as one himself for three years, before realising he didn’t have the reactive gene you need to be any good. After stints in advertising, he re-entered the thespian world and found himself teaching at RADA for eight years, where he worked with such soon-to-be-famous names as Ben Whishaw, Tom Hiddleston and Andrea Riseborough.

Spencer fell into Conacher’s lap through a long-standing connection with its screenwriter Steven Knight (who also created Peaky Blinders), but the process of getting hired was, as he puts it, a comedy of errors.

“There was a mix-up with my agents,” he cheerily explains. They told him that the Argentinian Pablo Trapero was directing it. (In fact, Spencer is the baby of Chilean auteur and critical darling Pablo Larraín.) “I didn’t know Trapero’s work, so I got hold of this movie of his The Clan, watched it, loved it, and then the call was set up with Pablo the next day”.

The wrong Pablo, unfortunately – who politely had to correct him when he started rhapsodising about The Clan. “For the first five minutes of the conversation I was mortified, but also trying to work out who I was actually talking to! Luckily, I had seen [Larraín’s previous biopics] Jackie and Neruda. And he couldn’t have been nicer about it.”

The next thing he knew, Kristen Stewart was phoning him out of the blue, when he was off-guard one evening after a boozy dinner, and their work began.

“We were all very much on the same page about the voice, as it related to Steven’s script and what it represented in the characterisation. I think I said ‘hyperbolic’ about the dialogue and Pablo said ‘melodrama’ – it was heightened and stylised in a way that felt very appealing to me.

“We started about three and a half months before the shoot. And it was quite an interesting process, because we didn’t touch the dialogue for a long time.”

Conacher’s method when he’s analysing a person’s voice, which he would have applied on his first ever Diana assignment, is to start off with video footage. “First I watch it without listening. I turn the volume down so I’m looking at physically what they’re doing. Then I try to start speaking in what I imagine that voice would be like. And then I’ll do the opposite – I’ll cull the audio and just listen to it on MP3, listen again, and see how it makes me move. And then I think, OK, what physical changes do I have to make to sound like that.

“Our politicians, for instance, have quite polarising voices. Sometimes it’s literally about the placement of the larynx. So Gordon Brown, for example,” – he adopts an imitative growl – “has that kind of pushed down larynx voice, whereas Harold Wilson’s is vairy haigh. You’re looking for the thing that will distill it for the actor – not the cheat, exactly, but the most obvious way in.”

To guide Stewart over those months of rehearsal, Conacher supplied her with what he calls a “drill sheet”, with phrases to help her master specific vowels. Here’s a sample: “The farmer transported the plants from France in a cart across the grass. Then he had a bath.” Imagining the different ways that might be pronounced by an American born in Los Angeles, then a Brit, then a posh Brit, and then specifically Princess Diana gives you some idea of the sculpting process involved.

“We did that for really quite a long time, and we also worked on transcripts of speeches. By the time we started working on actual dialogue in December, she’d cracked all the things she had found difficult.”

Stewart found the process especially helpful, he says, because she got the accent nailed before turning to the lines, and never had to speak any of those words without saying them in Diana’s voice. Other actors – famously, Christian Bale and Daniel Day-Lewis – stay in character so religiously they hang on to their accent between takes.

“Actors quite often ask me if they should do that. There’s a risk to it – if you’re making mistakes, you’re only making those mistakes get deeper into your muscle memory. So you have to be 100 per cent certain that you’re not doing it wrong.”

Stewart, he says, wasn’t overfamiliar with dialect coaching, even though she’d learned a British accent for Snow White and the Huntsman (2012), which Conacher describes as “pretty good.”

“I have to say she had a pretty solid grasp of RP before we started – it was just the Diana of it all we got specific about.”

What are the tricky aspects of Diana’s voice, then? “She’s posh, but not royal posh. So it’s getting that balance. Like, her O vowel is really specific, because it’s not “eyew”, like the Queen, it’s more like, “doughn’t”.”

When the trailer for Spencer came out, there were only two words spoken by Stewart’s Diana in it – the words “they don’t” at the end. “In those two syllables, everybody knew it was Diana,” Conacher agrees.

“There’s a great physicality to playing her, too. She does a lot of quirky things with her jaw and her head. And she was performative, as a person. I think we ramped that up for the film.”

Though Conacher is generally credited as “dialect coach”, there are many things to do with jaw, neck and face movement that fall into his purview. “With accents it’s more like learning a dance – it is physical. It’s muscle manipulation, not copying.”

Conacher’s is a behind-the-scenes role that puts him right in the firing line. He points to the Times review of Spencer, which specifically lambasts his work as having “produced nothing but a strange, strangulated whisper”. “To be honest,” he tells me, “strangulated whisper is kind of what we were going for!”

There's a certain irony, though, in receiving the brickbats. “I’ve never heard anybody come out of a cinema going ‘the accents were the best thing about it’," he notes. "When they’re at their best, they’re unnoticeable."

But he’s happy to praise the work of others. He was bowled over by the ensemble work in HBO’s TV drama Mare of Easttown, set in a Philadelphia suburb. “I absolutely believed that they were all from the same place.” Conacher had earlier coached the show’s star Kate Winslet - and Ralph Fiennes - to speak with Germanic accents for The Reader.

Conacher’s other big job lately reteamed him with Rami Malek, who needed to sound unsettlingly foreign as Safin in No Time to Die “without being associated with a particular nationality”, he explains. “We listened to people in the Caucasus region, Azerbaijan and Georgia, but it’s an amalgamation.”

He’d already established a shorthand with Malek, having trained out the American vowel sounds that usually have to be remedied first. “Americans have no distinction between “cot” and “caught” – that short ‘o’ sound that we have doesn’t exist.”

Now, with Spencer complete, and Debicki’s performance in The Crown yet to be unveiled, Conacher is ready to leave the People’s Princess behind. “After Elizabeth, I’m pretty sure I’m not going to do any more,” he says. “That might be the end of the road for me and Diana.”

Spencer is in cinemas from Friday 5 November