Research Report

Application of Artificial Intelligence in the Visual Presentation of Mór Jókai's Works. Research report on the questionnaire survey.

Petőfi Media, Jók.ai Lab, April 2024

Prepared by: Gréta Kocsis, Róbert Lányi, János Nagyillés, Tamás Németh, András Nyírő, István Szakadát

Results

The majority of the respondents who filled out the questionnaire belong to the 19-25 age group, which can be explained by the high response rate to the calls published at the universities of Szeged and Pécs. The high school age group is represented by students from Pannonhalma Benedictine High School and the Alternative Economics High School. Both schools represent a high educational standard and cannot be considered average. The 25-40 age group is the least represented, while the 40-50 age group has a surprisingly high proportion, almost as large as the youngest age group. The participation rate of the over-50 age group is also low, which is expected for an online survey. The participation of different age groups can be important in interpreting the results of the questionnaire, as age can influence technological affinity, reading habits, and attitudes toward literature.

A kitöltők túlnyomó többsége, több mint háromnegyede rendelkezik egyetemi végzettséggel, ez a korosztályos megoszlásból és a meghirdetési körből következik. A középiskolai végzettségűek jelentős kisebbségben vannak. A kérdőívet megmutattuk általános iskolai magyar tanároknak is, de azt mondták, hogy ebben a korcsoportban csak közös foglalkozás keretében lehetne kitöltetni egy ilyen kérdőívet, így ettől eltekintettünk. Mivel a tervezett Jók.ai weboldalt is az elkötelezett érdeklődőknek, és nem a legszélesebb közönségnek szánjuk, ezért ez a torzítás nem jelent problémát az eredmények értékelésében, de tudnunk kell, hogy a magasabb iskolai végzettségűek válaszai befolyásolhatják a felmérés általános eredményeit, például a komplexebb szövegek vagy képek értékelésében.

The majority of respondents, 69.3%, fall in the middle of the scale (3-4 points), indicating that they generally like Jókai's works. The two extreme ends (1 and 5 points) are less common, with fewer respondents giving the highest and lowest ratings. The fewest respondents chose the "do not like" category, which may suggest that Jókai's works are among the most favoured literary pieces among those surveyed.

The majority of respondents have read Jókai's two most famous novels, but it is also interesting that 11.5% said they had not read either. Among the two novels, "The Sons of the Stone-Hearted Man" was read by more respondents, though the difference is not significant.

The correlation analysis of the responses showed that there is no substantial relationship between demographic data and novel preferences, and more surprisingly, there is no significant correlation between these characteristics and the ratings of the images either. This indicates that individual taste and visual culture are more decisive in the evaluation of the images than other factors examined in the questionnaire.

The most votes were received by landscapes created in an authentic historical style, with the number of votes indicated in the bottom right corner of the images. These images are characterized by elements of romantic realism, which manifest in the idealized yet detailed portrayal of the beauty of nature. The compositions follow a classic arrangement, emphasizing perspective and depth in the style of 19th-century Hungarian landscape painters. The colour palette is rich in warm hues, with golden and autumnal tones, creating a dramatic effect. The play of twilight lights and shadows adds a deep, atmospheric quality and a sense of space to the images. The technique is delicate, precise, and realistic, with elaborated details.

A képek elkészítéséhez használt ChatGPT utasítások részletesek voltak, és szorosan kapcsolódtak a regény szövegéhez. Például a jobb alsó képet a következő promptra készítette el a ChatGPT: “Jókai Mór ‘Az arany ember’ című regényében Brazovics Athanáz házát így írja le: egy ódon úri butorzatú szobáról van szó, amely jobb oldalról állótükörrel és toilette-asztallal, balról himző-asztallal és kanapéval rendelkezik. A falakon 1816-os divat szerinti olajfestmény arcképek díszlenek. A szobában található egy szárnyajtó hátul, valamint jobbról és balról oldalajtók. Egy üvegcsillár csüng le a padmalyról, amely további eleganciát kölcsönöz a térnek. Alkoss képet erről a házról, egy jellegzetes Osztrák-Magyar Monarchia-beli város főutcáján ábrázolva a házat a 19. századi első felében.”

Similarly, portraits that reflect the spirit of the era received high ratings. The four portraits shown as examples are detailed and create a photorealistic effect. The colour palette is based on warm and natural tones, ranging from beige to brown, blue, and grey, suggesting a sense of timelessness. Their style is classic, reminiscent of 19th-century portrait painting. In the compositions, the human figure is the central element, with a strong emphasis on body language. The themes of the four images reflect respect for the past and traditions, further reinforced by classic attire and period settings, conveying timelessness and dignity.

The prompts for image creation in these cases also followed the character descriptions from the novels. For example, for the third image, we used this prompt: “Describe the character of Euthym Trikalisz from the novel and his appearance. Then create a portrait of him in the manner characteristic of the 19th century, considering that Euthym is 60-70 years old and of Turkish descent.”

A kérdőív kitöltői alacsonyabb pontszámokat adtak azoknak a képeknek, ahol eltértünk a 19. századi vizuális világtól. Ezeken a képeken keverednek a fikciós, szürreális és a realisztikus, történelmi elemek. A színpaletta élénk, kontrasztos és gazdag, a kékek, vörösek és zöldek dominanciájával, ami dinamikussá teszi a képeket. Mindegyik kép erőteljes vizuális narratívával rendelkezik. A történelmi és modern elemek egyaránt jelen vannak és kölcsönhatásban vannak egymással. Az ábrázolások a valóság és a képzelet határait feszegetik. A promptokban elszakadtunk Jókai regényeinek világától, például a 4. képhez ezt az utasítást adtuk: “Készítsd el kubista stílusban a Senki szigetének képét.”

The lowest scores were given to images where we allowed the most freedom to ChatGPT. The colour palette in each image is vivid and saturated, highlighting every detail and creating a fabulous, almost magical atmosphere. The depiction style is similar: realistic facial expressions and body language that convey the characters' personalities while placing them in an idealized, often fantastical world. They present reality in an exaggerated, idealized form. These images lead to a fantastic, idealized world where colour, light, and detail are prominent, and where natural beauty is presented in an idealized form. The characters and landscapes at the centre of attention further reinforce this magical and perfected worldview. For example, for the first image, we provided ChatGPT with this description: “Show a beautiful, colourful, realistic natural scene that evokes the mood of this quote: 'There are many troubles in the world that one would like to help if one could. A poor man like me sighs deeply and says: what can I do about it?'” “Mutass egy gyönyörű, színes, realisztikus természeti képet, ami megidézi ennek az idézetnek a hangulatát: ‘Sok baj van a világon, amin az ember ugy szeretne segiteni, ha tudna. Magamforma szegény ember nagyot sóhajt rá, s azt mondja: mit tehetek róla?’”

The questionnaire allowed respondents to provide textual feedback on the images. We categorized the evaluations into five groups, which helped us analyse participants' reactions and opinions about the AI-generated images more deeply. The first category is visual style and artistic execution, where positive comments praised the detailed depiction and use of colour, while negative critiques highlighted the stylistic inconsistencies and improper application of artistic movements. The second category is historical and cultural fidelity, which assessed the historical accuracy and alignment with the era of the images. Here, opinions were divided between faithful and anachronistic representations.

The third category is character portrayal and thematic fidelity, which was particularly important in the context of Mór Jókai's works. Here, participants appreciated images that authentically depicted Jókai's characters and reflected the atmosphere of the novels. However, negative remarks were also made, criticizing deviations from the expected character portrayals. The fourth and fifth categories, which examined technological execution and emotional and aesthetic impacts, also showed significant variations. In these categories, participants positively noted the AI's ability for creative approaches, while their criticisms often related to the formulaic and emotionless nature of some AI-generated images. Here is an example of an image that received a relatively large number of comments.

Does not match our image of Jókai

• Too beautiful, not realistic.
• The drawing is a bit too modern, I think the conditions in 19th-century Hungary were poorer than this.
• This is supposed to be a commercial ship.
• I must not remember the novel well, but I don't think there is such a scene in it, and this drawing doesn't capture the mood of the novel at all for me.
• It doesn't reflect the world of the novel. Timea is a Turkish girl, she wouldn't wear clothes like this.
• If it is intended for an older audience, it’s fine, but for young people, I think there are too many grey and brown tones.

Reservations about AI

• The cartoonish style hides the AI, making it less noticeable.
• Too much like a fairy tale.
• Very American, I would think of Mark Twain's work before Jókai's.
• Kitsch crap!
• Uncreative, inhuman, soulless. Only people who don't care about visual culture create or like this.

Positive Evaluations

• I like the style.
• The artist paid attention to every little detail.
• There's a bit too much grey... otherwise, the illustration is beautiful.
• It's interestingly different from the classic film.
• The colours and perspective are good.
• It fits the style very well.

The full text of the comments on the images is attached to the research report.

Conclusions

The aim of the research was to examine how participants perceive images generated by artificial intelligence based on Mór Jókai's novels. The results indicate that the majority of participants positively evaluated images that remained faithful to the visual style and themes of Jókai's era. Landscapes with the characteristics of classical, romantic realism and period portraits received high scores, authentically reflecting the atmosphere of Jókai's novels and the distinctive features of their characters.

Participants generally rated images that significantly deviated from the 19th-century style lower. Images containing modern or abstract elements, such as illustrations in cubist style, received lower scores because they often did not faithfully reflect the spirit of Jókai's works. Such experimental images were considered more controversial and were found not to fit well with the historical and cultural context created by Jókai.

The research also showed that participants' preferences did not significantly differ based on age, educational background, or familiarity with Jókai's works. This suggests that sensitivity to visual styles and artistic approaches is more general and not tied to age or educational level. There was a general respect for the works among respondents, which positively influenced the reception of the images.

Detailed, closely related images that used specific scenes or character descriptions from the novels generally received better reception. When using AI-generated images, it is important to consider the visual preferences of the target audience and historical accuracy. For images created for Jókai's works, it is particularly important that the style and tone of the images are in harmony with the world of the novels, as this can contribute to a deeper understanding and popularization of the works.