Sign In

Sora is able to generate text in multiple languages

8

Sora is able to generate text in multiple languages

Not much to say past the title, but as I was experimenting with the amount of text that Sora is able to generate a prompt, I got curious if it could do other languages.

A photorealistic black-and-white image styled as a 1946 historical courtroom photograph. In the dock of the Nuremberg Trials sits a lone white goose, surrounded by towering wooden walls and flanked by two unsmiling American MPs. The goose is perched on a custom stool, secured by a leather strap around its midsection. Behind it, a row of grim-faced international justices look on. The lighting is high contrast, casting hard shadows on the courtroom woodwork. In the foreground, a microphone is positioned in front of the goose as if it’s about to testify. The audience gallery is filled with military officials and reporters. Despite the official setting, the goose sits with perfect stillness and defiant posture — its beady eyes scanning the room. One of the guards nearby stares forward, expression unreadable. Typed archival caption below, in official U.S. Army Signal Corps style: “TRIAL OF FORMER OCCUPATION OPERATIVE, CASE 1147 – GOOSE DENIES CHARGES. DEFENSE ARGUES NO EVIDENCE LINKS DEFENDANT TO JÄGERFELSEN EVENTS.” NÜRNBERG, MARCH 1946 – U.S. ARMY SIGNAL CORPS ARCHIVE 1147-B


I was trying to recreate one of the iconic "Roof Korean" images (with upgrades), and my first attempt I thought to try to actually get the text right in the sign.

Riots? Where are the roof koreans

I don't understand Korean, and I gave GPT the image and told it to include the text from the sign within the prompt, looking at it now, it looks like it missed on the last character, but Sora was able to generate the text correctly according to the prompt:

assets_task_01jr2t4832eykrpgqnvbqz1evx_img_1.webp

A photorealistic recreation of the iconic 1992 "Roof Koreans" rooftop photo during the LA riots. The scene shows a smiling Korean-American man in a red polo shirt holding a modern M249 light machine gun slung over his right shoulder in the exact same relaxed pose from the reference image. Behind him, several Korean-American men are posed identically to the original photograph, including the man in a yellow cardigan seated cross-legged on the roof’s edge. Prominently in the background is the red Korean sign that reads 가주마켓, with its unique white Korean characters on red. The setting shows the same street corner with 90s-era signage like Carl’s Jr., power lines, streetlamps, and palm trees under hazy Los Angeles daylight. The style is 1990s photojournalism, with light lens grain and slight overexposure to match the analog photography of the era. The only difference from the original is that all weapons have been updated to modern military hardware — M249 SAWs, RPG-7 launchers, and tactical body gear — giving the image a surreal, militarized twist while preserving the exact body language and rooftop layout.


I don't know Russian either, but I know the Cyrillic alphabet, and I know what Cyrillic cursive looks like, the differences between their print and cursive is far more than that of English's, so I was curious if it would be able to create it. My first attempt, which produced it in print.

Generated image

A photorealistic image of a decaying, abandoned urban interior—cracked concrete walls, exposed rebar, peeling paint, and scattered rubble. The lighting is dim and natural, filtered faintly through broken windows. On one large interior wall, the phrase "они уже здесь" is scrawled in bright red cursive Cyrillic script, with thick paint drips running downward like blood, stark against the bleak surroundings. The rest of the image is bathed in a dreary color palette of washed-out greys, muted browns, and cold shadows, evoking a sense of desolation and foreboding. Dust particles hang in the air, and faint light catches on them. The composition feels like a still from a found footage horror film or a post-apocalyptic thriller.


I then prompted it with more specificity:

A photorealistic image of a crumbling, abandoned concrete building. The atmosphere is desaturated and bleak, with a muted palette of greys and browns. Dust and rubble litter the floor. Dim natural light filters through a broken window on the right side, casting faint, cold illumination across the cracked walls. On the largest interior wall, written in bright red Russian cursive script (рукописный шрифт), are the words "они уже здесь", dripping like fresh blood — the paint runs downward in long, uneven streaks, contrasting starkly with the surrounding decay. The cursive handwriting is unmistakable, fluid, and eerily human, amplifying the unsettling tone. The scene evokes a chilling sense of dread, as if something unspeakable has already begun.


Then I went to make a cover image for this article with a few languages off the top of my head that didn't use the Latin alphabet as well, for this I'll show you a generation with the same prompt that isn't the cover image, it is able to produce this consistently:

Generated image

A photorealistic black-and-white photograph of an open journal lying flat on a wooden surface. The journal’s pages are unlined, textured slightly, and lit with soft, moody overhead lighting. Centered on the left page is a vertical list of language names, each handwritten in its own script in realistic ink-based cursive, descending line by line down the page.

At the top is the word “English”, written in flowing English cursive script.

Below it, written in Russian cursive (рукописный шрифт):

русский” — clean, stylized Cyrillic handwriting.

Next, the word “中文” in elegant, Traditional Chinese calligraphic script.

Below that, “한국어” written in Korean Hangul, flowing in a graceful, semi-cursive handwritten style.

Then, “日本語”, handwritten in Traditional Japanese Kanji, fluid and natural.

Finally, near the bottom of the page, “العربية”, written in flowing Arabic cursive script, with its elegant sweeping lines.

Resting diagonally across the bottom right corner of the journal is a classic black fountain pen, its tip hovering just above the page with a single drop of ink suspended at the nib — casting a sharp, dramatic shadow on the page beneath.

The entire image is styled as a high-contrast monochrome photograph, emphasizing textures in the paper, shadows in the ink, and the timeless quality of handwritten language.


I thought this was really interesting, almost certainly not the first person to come across this. Anyways, right now I'm doing a lot of experimentation with Sora, I've been adding Sora generations on Civitai that have the prompt included in a collection so that I can see what Sora is able to do and how it responds to different prompting.

If you find the collection interesting in terms of the images or the fact they have the prompts, following it is appreciated!

I'm interested in what other people are able to get out of it as well, so if you upload any sora generations with the prompt, please feel free to @schmede in the comments and I will (most likely) add it to the collection (probably not a good idea to fully commit to that, sight unseen and unconditionally).

Thanks for reading!

8