After a bit of deep-delving, it’s the voice clone features that have me hooked right now. ElevenLabs can make a digital version of your voice from just 30 seconds of training speech. And it’s fast. I expected a bit of a wait for audio processing the first time I used it. But no – after reading in a couple of passages of sample text, my digital TTS voice was ready to use within seconds.
For a quick ‘n’ easy tool, it does a brilliant job of picking up general accent. It identified mine as British English, captured most of my Midlands features (it struggled with my really low u in bus, though – maybe more training would help), and it got my tone bang on. Scarily so… I can understand why cybersecurity pundits are slightly nervous about tech like this.
Your Voice, Another Language
The most marvellous thing, though, was using my voice to read foreign language texts. Although not 100% native-sounding – the voice was trained on me reading English, of course – it’s uncannily accurate. Listening to digital me reading German text, I’d say it sounds like a native-ish speaker. Perhaps someone who’s lived in Germany for a decade, and retains a bit of non-native in their speech.
But as far as models go, that’s a pretty high standard for any language learner.
ElevenLabs’ TTS interface with the custom voice ‘Richard’ selected, ready to read some German.
The crux of it is that you can have your voice reading practice passages for memory training (think: island technique). There’s an amazing sense of personal connect that comes from that – that’s what you will sound like, when you’ve mastered this.
It also opens up the idea for tailoring digital resources with sound files read by ‘you’. Imagine a set of interactive language games for students, where the voice is their teacher’s. Incredible stuff.
The catch? Much of the brand new tech isn’t available to the average user-on-the-street yet.
That’s why I was thrilled to happen across TTS service ElevenLabs recently. ElevenLabs’ stunning selection of voices powers a number of eLearning and audiobook sites already, and it’s no hype to say that they sound as close to human as you can get right now.
Even better, you can sign up for a free account that gives you 10,000 characters of text-to-speech conversion each month. For $5 a month you can up that to 30,000 characters too, as well as access voice-cloning features. Just imagine the hours of fun if you want to hear ‘yourself’ speak any number of languages!
Using ElevenLabs in Your Own Learning
There’s plenty to do for free, though. For instance, if you enjoy the island technique in your learning, you can get ElevenLabs to record your passages for audio practice / rote memorising. I make this an AI double-whammy, using ChatGPT to help prepare my topical ‘islands’ before pasting them into ElevenLabs.
The ChatGPT > ElevenLabs workflow is also brilliant for dialogue modelling. On my recent Sweden trip, I knew that a big conversational contact point would be ordering at coffee shops. This is the prompt I used to get a cover-all-bases model coffee-shop convo:
Create a comprehensive model dialogue in Swedish to help me learn and practise for the situation “ordering coffee in a Malmö coffee shop”.
Try to include the language for every eventuality / question I might be asked by the coffee shop employee. Ensure that the language is colloquial and informal, and not stilted.
The output will be pasted into a text-to-speech generator, so don’t add speaker names to the dialogue lines – just a dash will suffice to indicate a change of speaker.
I then ran off the audio file with ElevenLabs, and hey presto! Custom real-world social prep. You can’t specify different voices in the same file, of course. But you could run off the MP3 twice, in different voices, then splice it up manually in an audio editor like Audacity for the full dialogue effect. Needless to say, it’s also a great way for teachers to make custom listening activities.
The ElevenLabs voices are truly impressive – it’s worth setting up a free account just to play with the options and come up with your own creative use cases. TTS is set to only get better in the coming months – we’re excited to see where it leads!
The addition of voice chat mode to ChatGPT – soon available even to free users in an impressive, all-new format – opens up tons of possibilities for AI speaking practice. When faced with it for the first time, however, learners can find that it’s all a bit undirected and woolly. To make the most of it for targeted speaking practice, it needs some nudging with prompts.
Since AI crashed into the language learning world, the prompt bank has filled with ways to prime your chatbot for more effective speaking practice and prep. But there’s one activity I’ve been using lately that offers both structure, tailored to your level and topic, and a lot of fun. I call it So Interview Me!, and it involves you playing an esteemed expert on a topic of your choice, with ChatGPT as the prime-time TV interviewer.
So Interview Me!
Here’s an example you can paste into ChatGPT Plus straight away (as text first, then switching to voice mode after the initial response):
Let’s role-play so I can practise my Swedish with you. You play the role of a TV interviewer on a news programme. I play an esteemed expert on the topic of ‘the history of Eurovision’. Conversational turn by turn, interview me in the target language all about the topic. Don’t add any translations or other directions – you play the interviewer and no other role. Wind up the interview after about 15 turns. Keep the language quite simple, around level B1 on the CEFR scale. Are you ready? Start off by introducing me and asking the first question!
The fun of it is that you are the star of the show. You can completely throw yourself into it, interacting with your interviewer with all the gusto and gumption of a true expert. Or you can have some fun with it, throwing it off with silly answers and bending the scenario to your will (maybe you turn out not to be the expert!).
Either way, it’s a brilliant one to wind up and set going before you start the washing up!
Just a couple of weeks after the excitement around Hume.ai, OpenAI has joined the emotive conversational bandwagon with a stunning new release of its GPT-4o model.
GPT-4o is a big deal for language learners because it is multimodal in much more powerful ways than previous models. It interacts with the world more naturally across text, audio and vision in ways that mimic our own interactions with language speakers. Demos have included the model reacting to the speaker’s appearance and expression, opening a path to more realistic digital conversation practice than ever.
As with Hume, its voice capabilities have been updated with natural-sounding emotion and intonation, along with a deeper understanding of the speaker’s tone. It even does a better job at sarcasm and irony, long the exclusive domain of human speakers. Heck, it can even sing now. Vocal, emotional nuance – at least simulated – does seem to be the latest big leap forward in AI, transforming the often rather staid conversations into something uncannily humanlike. And as with many of these developments, it almost feels like it was made with us linguists in mind.
As for the multimodal capabilities, we’ll have to wait a little longer, unfortunately – chat updates are being propagated more gradually, although you may already the next time you open chat mode, you may already get the message that big changes are coming. Definitely a case of watch this space – and I don’t know about you, but I’m already impatiently refreshing my ChatGPT app with increasing frequency!
If you’ve kept up with my copious social postings, you’ll know that I’ve spent the last week in beautiful Malmö, following my Eurovision language dreams. Perhaps not the calmest of years to choose – the contest itself was mired in controversies that just seemed to be compounded by poor decision after poor decision. At times, the atmosphere felt incredibly on edge. Needless to say, the joy that was Switzerland’s Nemo winning was the tonic we all needed.
As for my language goals, though, it’s been a blast.
I put all that to the test this week. And I think I can finally say, without piquing my impostor syndrome to breaking point, that I speak Swedish. Ja, äntligen pratar jag svenska! Granted, coffee shop counters have been the main playground for my newfound skills, but with each interaction I’ve felt more and more confident using it.
Avoiding (Un)Helpful hands
One obstacle I was very wary of at first was the helpful English-speaker. You know the type if you’ve been to a country with really strong, widespread anglophone knowledge. You try out your target language, only to get English back at you by default. It’s often enough to scare you back into your shy language learner box and accept defeat.
In Malmö, however, it didn’t happen once. That’s perhaps more to do with my obsessive fascination with mimicry, rather than Malmoans’ inherent desire to help learners of Swedish. I’ve spent a lot of time listening to Swedish podcasts and watching Swedish series to train my ear. Then, in my spare time, I’ve rehearsed speaking phrases out loud, laying it on thick with the accent and paying particular attention to the Swedish tones. I’d clown around with it, role-playing an authentic Swede. Melodifestivalen introductions were particularly fruitful ground for this – låt nummer ett : Carola! I’d pronounce in the shower, in my finest continuity announcer svenska.
It may all sound completely bonkers, but it worked a treat. I ended up sounding decent enough for Swedes to assume I had a better grip of the language than I probably (certainly) do, but it stopped the dreaded automatic-switch-to-English, and gave me more precious time practising with real people. Once my level became apparent and the deception was revealed, I could hop in with a jag lär mig svenska (I’m learning Swedish), which resulted in some nice compliments and occasionally, a new word or two explained by the other party. My favourite was vispgrädde, whipped cream, explained by a very patient and lovely Espresso House barista!
So, I’ve come out of my Swedish adventure with a refreshed appreciation of accent-training as an indispensable part of any language learning regime. Podcast-shadowing, talking to yourself, singing in the shower – however daft it feels, it just works. Give it a go if you’re sceptical – I bet you’ll be surprised.
The only thing I have to do now is relearn how to speak Norwegian again without sounding Swedish…
If the socials are anything to go by, so many of us language learners are already using AI platforms for conversation practice – whether text-typed, or spoken with speech-enabled platforms like ChatGPT.
Conversational interaction is something that LLMs – large language models – were created for. In fact, language learning and teaching seem like an uncannily good fit for AI. It’s almost like it was made for us.
But there’s one thing that’s been missing up to now – emotional awareness. In everyday conversation with other humans, we use a range of cues to gauge our speaking partner’s attitude, intentions and general mood. AI – even when using speech recognition and text-to-speech – is flat by comparison. It can only simulate true conversational interplay.
A new LLM is set to change all that. Hume.ai has empathy built-in. It uses vocal cues to determine the probable mindset of the speaker for each utterance. For each input, it selects a set of human emotions, and weights them. For instance, it might decide that what you said was 60% curious, 40% anxious and 20% proud. Then, mirroring that, it replies with an appropriate intonation and flex.
The platform already supports over 50 languages. You can try out a demo in English here, and prepare to be impressed – its guesses can be mind-bogglingly spot-on. Although it’s chiefly for developer access right now, the potential usefulness to language learning is so clear that we should hopefully see the engine popping up in language platforms in the near future!
If you didn’t already know, I’ve been spending a year Swedifying my Norwegian. The goal? Eurovision fun days in Malmö, of course. And as the final test of my newfound svenska draws close, it seems like a good time to take stock of what – and how – I’ve been learning.
I’m a big fan of finding content that speaks to you, be that books or television and film. Personally, I’m much more likely to keep coming back to learning content if I find it fun. With that in mind, and through much trial and error, I’ve found some things that I love in Swedish – things that have made my Swedish journey so much more effective.
Here are some of the biggies from my past twelve months!
Crime Fiction
It bears repeating: extensive reading is one of the most sure-fire ways to solidify your familiarity and ease with a foreign language. It’s the vastness of the input – as you soak up a story, things are bound to stick. And this particular genre was a bit of a no-brainer, as I also loved crime fiction in my original scandilang, Norwegian (check our Jørn Lier Horst – the author of the first novel I read completely in Norwegian, Blindgang).
Thanks to a nice little Swedish section at the London branch of Foyles, I found one I liked the look of – Benvittring by Johan Theorin – and have been tiptoeing through that for the last few moths. It has all the dark, moody suspense we’ve come to love with Scandi noir, and fits neatly into a series – the Ölandssviten books – if I want more when I come to the end (and I will – it’s great). Maybe it’s because the genre is so formulaic, but it all seems so familiar – Horst and Theorin could be writing cousins.
Although a few of Theorin’s works have been translated in English, unfortunately Benvittring hasn’t, yet. That does give me the sense that I’m one of the first outside Sweden to read it, which feels very special and exclusive. The downside, of course, is that I can’t recommend it to friends who don’t speak the language.
Now, I love some light listening in the morning when I’m going about my bits and pieces. Something chatty and informal, that you can have on in the background and selectively drop into, much like picking at a smörgåsbord (which, incidentally, means sandwich table in Swedish). Swedish Radio‘s first station, P1, fits the bill perfectly – lots of opinion-piece phone-ins, interesting documentaries, the odd overacted melodrama, and hourly bulletins to satisfy the news junkie in me.
Better still, Swedish Radio is available as a third-party skill on Alexa, and I do love to recruit my digital assistants as language learning buddies. So it’s as easy as putting my morning coffee on and exclaiming Alexa, play Swedish Radio to get some listening practice on. I let my attention dip in and out of it as I go about my other business, and I haven’t done any structured or focused listening with it. But it’s been fundamental in re-tuning my ear to the shape of Swedish – vital as someone hopping over from a very closely related language.
Young Royals
Now I know I’m not alone here, and I’m in much larger company than other Swedish learners. The Netflix coming-of-age drama Young Royals depicts the blossoming romance between Crown Prince Wilhelm and classmate Simon, and has been a bit of a breakout sensation. There’s even an official Spotify playlist, which has introduced me to new music much more with it than I can admit to being. Think The Crown, but cooler (and probably no less made up).
The show’s popularity has led many to the language, too. There are whole Reddits about Young Royals sparking a Duolingo obsession. Despite that, the next-best thing – after it being in Swedish, of course – is that there’s a dubbed English version too, so I can recommend it to non-linguaphile friends and family.
What’s more, once you watch, and rate, a Swedish-language show on Netflix, you’ll have more recommended to you. Thanks to my Young Royals binge, I’ve discovered a whole lot more Swedish content on the platform since.
Drag Race Sweden
Staying with the queer theme, here, I credit the fabulous Drag Race Sweden with one very useful power-up: colloquialisms. Not the odd idiom here and there, but the whole gamut of real, everyday, lived Swedish spoken between friends. The language used between the competing queens is so informal that it’s an antidote to the staid dialogues of standard text books. It’s thanks to that – along with the accompanying Swedish subtitles – that I’ve learnt vocab like taggad (psyched) and peppad (stoked) and so much other emotive language that is totally transferrable to the Eurovision context. Yes, in Malmö I’ll be sharing my colloquialised opinions left, right and centre, and it’s all thanks to Robert Fux. That’s not to mention the catchphrases… Må besta quinna vinna! A sentiment that fits Eurovision like a glove.
Getting into a foreign language TV show opens up a web of connected socials, too, and Drag Race Sweden has provided some very entertaining accounts. In particular, if you’re interested in the accents of Skåne – the Swedish region where Malmö is situated – then competing queen Elecktra’s TikTok is worth a follow. There’s even another Eurovision link-up there, as she was one of the contestants at this year’s Melodifestivalen. And of course it was in Skåne dialect, which she had form for after performing the hilarious Unna daj (Treat Yourself) in her season. Banna maj was every bit as wonderfully camp. #ElecktraWasRobbed, indeed.
By the way, for learners of Swedish and other languages, Wow Presents, which hosts most of the worldwide Drag Race content, is well worth the £4-ish a month it costs to subscribe. Fabulous, binge-worthy fun that’ll have you laughing and learning.
Courses and Traditional Content
Of course, I also invested my time and dosh in a couple of courses at the beginning of this journey. How could I not, being the book fiend that I am. As a Norwegian speaker, though, not many courses are geared up to the false – or rather, transferring – beginners, and I found it a slog to get past those early chapters where it seemed as though I was treading the same ground all over again. My Swedish side-step, piggybacking on media content created for the Swedish market was, by comparison, much more dynamic, interest-holding and effective as a strategy.
That’s not to say that some traditional course books haven’t been useful. Teach Yourself Swedish Tutor, for example, is a great dip-in-and-out book with short, snappy chapters, each with a tight grammar focus. Alongside that, old stalwart Duolingo has been predictably very handy for new vocab (and giving it the Swedish treatment has also fostered a much healthier use of the app).
So, armed with my newfound Swedish, off to Malmö I go. Through the fun stuff I used along the way to learning Swedish, I feel I know Sweden itself a lot better, too. And on top of that, it’s been an ace low-stakes, low-pressure, high-entertainment-value way to learn. As such, it’s been one of the most enjoyable, guilt-free dabblings that I’ve had with a foreign language.
It feels like we only just said goodbye to the last one, and another Eurovision Song Content has rolled around again. Once a veritable garden of languages, all competing broadcasters were re-granted a free choice of song language in 1999. Sadly (for linguaphiles) that’s meant English lyrics for the most part.
But linguistic diversity has found a way, too, and not just thanks to those hardy regulars like France, Italy, Portugal and Spain that almost never disappoint with home-language lyrics. The 2023 edition saw the welcome return of tongues long-missed on the Eurovision stage, like Finnish and Russian.
So how does 2024 measure up against that pretty high bar?
The Eurovision Language Contest 2024
Big Firsts
Notably, we have two language debuts at this year’s contest. Azerbaijan, entering since 2008 without a word of Azeri, finally treats us to a few words of this beautiful Turkic language in the entry Özünlə apar (take me with you). And from Australia, a competing member of the family since 2015, we have the uplifting song One Milkali (One Blood) featuring lyrics in Yankunytjatjara, a Pama-Nyungan language from Western Australia. Azeri and Yankunytjatjara may not feature as their full entry texts, but it is a beautiful thing to celebrate new languages on the Eurovision stage!
As an aside, as one commenting fan dubbed it, it’s that moment when Yankunytjatjara makes it to Eurovision before Scottish Gaelic and Welsh. We UK fans live in hope…
There’s a first for Armenian, too. While we’ve heard the language in previous entries, 2024 is the first time it will be the sole language of an Armenian entry. Jako has a world music fusion vibe, and a simple message of be yourself, which is a noble sentiment in any language.
Many Happy Returns
The it’s been TOO long! prize must go to Norway this year. Norway has sent a song with Swahili lyrics (2010) more recently than it has one på norsk (2006). The latter, Christine Guldbrandsen’s Alvedansen, didn’t even do particularly badly, so heaven knows what put them off.
This year, though, Norwegian folk metallists Gåte were the surprise vanquishers of fan favourites Keiino, pipping them to the Norwegian ticket with the song Ulveham and breaking the Norwegian drought. Its beautifully haunting arrangement builds on traditional Kulning calls from the mountain herds of Norway, featuring lyrics drawn from Telemark dialect.
https://www.youtube.com/watch?v=UipzszlJwRQ
While the return of Finnish was last year’s joy, its loss this year is tempered by the return of its close cousin, Estonian. The collaboration between 5miinust and Puuluup will present (Nendest) narkootikumidest ei tea me (küll) midagi (the crazily-titled We (sure) know nothing about (these) drugs), the first time Estonia has presented its national language since back-to-back eesti keel in 2012 and 2013. Incidentally, it wasn’t all English for Estonia in the interim – they achieved a solid top ten in 2018 with a song in Italian, of all tongues.
Going Dutch, Again
Dutch had fared similarly poorly in the anglophone takeover too – until recently. After one of many mid-noughties semifinal failures, the Netherlands ditched its national language following the 2010 contest. It took until 2022 for Dutch to pop up again, with considerable success – De diepte ended up of the left side of the scoreboard in the Torino contest. Two years later, Dutch is back again, this time with Joost Klein and Europapa.
Lithuania has also shied away from using its home tongue on the Eurovision stage. It took 21 years for the language to be heard again after a mediocre result in English and Lithuanian in 2001. But that return made the 2022 final, with Monika Liu scoring a solid result just outside the top ten. This year, Silvester Belt is aiming to do even better with the catchy Luktelk (Wait).
Greece will be looking to mirror that national language return to success, too. Greece’s last two attempts with full or partial Greek lyrics ended in very rare semifinal failure for the country, in 2016 and 2018. Marina Satti aims to be the first Greek-singing finalist since 2013, with a self-ironising, catchy, ethnopop banger.
Doubling Up
French and Spanish fans have an extra bite at the language cherry this year, and from perhaps surprising sources. Thanks to the return of Luxembourg to the contest – after an incredible 31 years away – we have a song with mixed French and English lyrics in the tally. As for Spanish, we can thank the Sammarinese win of Spanish rockers Megara for the fact that this year’s entry from the microstate will be in Spanish, not Italian or English.
Mixed Bag from the Balkans
We can always count on the Balkans for some non-anglophone fun at Eurovision. This year, we have, interestingly, two proper-name songs in Serbian Ramonda and Slovene Veronika. Only Albanian and Croatian lose out to English entries (although Croatia is doing very well for that as a pre-contest bookies’ favourite!).
The Hardy Annuals
And of course, we have our stalwarts, our indefatigable linguistic champions – France, Italy, Portugal and Spain. They’ve kept the national language flags flying almost without fail throughout the modern free-language era, and we should celebrate each of them for that. Italy in particular is a veritable feast of lyrics, with the hugely talented Angelina Mango firing them out in a fast-paced three minutes. Little wonder that she is also one of this year’s hot favourites for the top.
We might almost add Ukraine to this list, having not only sent, but won in Ukrainian in recent years. Ukraine opts for a cool mix this year with the duo Alyona Alyona and Jerry Heil.
And for the Germanists…
No consolation for the Germanists, this year – again. 2012 was the last time German – or at least a dialect of it – formed part of a Eurovision song lyric. That honour goes to Austria’s Woki mit dem Popo (pretty much shake your bumbum in Upper Austrian dialect), which failed to make the final that year.
Can you believe it’s been that long? Me neither. But there’s small consolation in the fact that Germany had a stonker of a song in their national final this year. Galant’s Katze (cat) may have fallen at the final hurdle, but it has all the makings of a cult classic.
Which are your favourite non-English entries this year? And which language do you yearn to hear again on the Eurovision stage? Let us know in the comments!
I’ve been crash-learning Swedish (well, side-stepping into it from Norwegian) more and more intensively of late. And one of the most pleasant linguistic detours I’ve made has been through the lush valleys of Swedish proverbs.
Proverbs and sayings have always been a favourite way in of mine when working on a language, and for several good reasons. Firstly, they’re short, and usually easier to remember by design so people could easily memorise and recite them. Secondly, they’re very often built around high-frequency structures (think X is like Y, better X than Y) that serve as effective language models.
Bättre en fågel i handen än tio i skogen (Better one bird in the hand than ten in the forest)
But there’s another big pay-off to learning through proverbs that is more than the sum of their words. They pack a lot of meaning into a short space – drop them in and you’re calling to the conversation all the nuance they carry. Think of the grass is always greener… You don’t even need to mention the second, missing part of that English proverb, and it already calls to mind countless shared parables of misplaced dissatisfaction. And since they’re based on those parables and folk histories that ‘grew up’ alongside your target language, proverbs can grant us some fascinating cultural insights, too.
In short, master proverbs and you’ll sound like you really know what you’re talking about in the target language.
Finding Proverbs
For many target languages, you’ll likely be able to source some kind of proverbs compendium in a good bookshop, as they’re as much of interest to native speakers as they are to learners. When you do find a good one, compilations of sayings are the epitome of the dip-in-and-out book. I’ve picked up lots of Gaelic constructions and vocab leafing idly through Alexander Nicolson’s Gaelic Proverbs in my spare moments. It was definitely time for me to try the same with some Swedish.
Without a good Swedish bookshop to hand, though, I turned to the Internet in the meantime. A good place to start is to find out what “[your language] proverbs” is in your target language (it’s svenska ordspråk in Swedish), and see what a good search engine throws up.
Tala är silver, tiga är guld (Talking is silver, silence is gold)
Local cultural institutions in particular can be rich sources of articles on folk wisdom like proverbs. There are some lovely sites and articles that introduce the wise words of svenska in digestible chunks. My handful of Swedish favourites below are each written for a native speaker audience. They all give potted backgrounds on the proverbs in Swedish, making for some great extra reading practice.
This folk-minded article is a wonderful introduction to Swedish proverbs, offering not only examples, but also exploring the characteristics of proverbs and what makes them ‘stick’. There’s a special section on sayings from the Gothenburg area too, which adds a nice local flavour.
This article from the Land publication offers 19 common Swedish proverbs in handy list format. Even more handily, it paraphrases each in order to explain their meaning. Great for working out what some of the more archaic words mean without reaching for the Swedish-English dictionary!
Nordiska Museet offers another well-curated list, with not only paraphrasing, but etymological information on the more difficult or outdated words.
The Proverbial AI
You can also tap the vast training banks of AI platforms for proverbial nuggets. Granted, the knowledge of LLMs like ChatGPT and Claude may not be complete – training data is only a subset of material available online – but AI does offer the advantage of activity creation with the material.
Try this prompt for starters:
Create a Swedish proverbs activity to help me practise my Swedish.
Choose five well-known proverbs, and replace a key word in each with a gap. I must choose the correct word for the gap from four alternatives in each case. Make some of the alternatives humorous! Add an answer key at the end of this quiz along with brief explanations of each proverb.
I managed to get some really fun quizzes out of this. Well worth playing around with for self-learning mini-worksheets!
A Swedish proverbs activity created by ChatGPT-4
AI platforms can also play a role as ‘proverb visualisers’, which is how I generated the images in this article. Proverbs can often employ some quite unusual imagery; letting picture generators loose on those can be a fantastic way to make them more memorable!
However you come across target language sayings and proverbs, you can learn a lot from these little chunks of wisdom. Do you have a favourite saying in any of the languages you’re studying? Let us know in the comments!
The big ChatGPT news in recent weeks is about images, rather than words. The AI frontrunner has added a facility to selectively re-prompt for parts of an image, allowing us to tweak sections that don’t live up to prompt expectations.
In essence, this new facility gives us a second shot at saving otherwise perfect output from minor issues. And for language learning content, like posters and flashcards, the biggest ‘minor’ issue – the poor spellings that crop up in AI image generation – makes the difference between useful and useless material.
Rescuing ChatGPT Posters
Take this example. It’s a simple brief – a stylish, 1950s style travel poster for France. Here’s the prompt I used to generate it:
Create a vibrant, stylish 1950s style travel poster featuring Paris and the slogan “La France”.
I wanted the text “La France” at the top, but, as you can see, we’ve got a rogue M in there instead of an N.
To target that, I tap the image in the ChatGPT app. It calls up the image in edit mode, where I can highlight the areas that need attention:
Then, I press Next, and can re-prompt for that part of the image. I simply restate the slogan instructions:
The slogan should read “La France”.
The result – a correct spelling, this time!
It can take a few goes. Dodgy spelling hasn’t been fixed; we’ve just been given a way to try again without scrapping the entire image. Certain details also won’t be retained between versions, such as the font, in this example. Others may be added, like the highly stylised merging of the L and F in the slogan (a feature, rather than a bug, I think!).
But the overall result is good enough that our lovely 1950s style poster wasn’t a total write-off.
Another case of AI being highly imperfect on its own, but a great tool when enhanced by us human users. It still won’t replace us – just yet!
Image tweaking is currently only available in the ChatGPT app (iOS / Android).
