Kling Polish Speech Test: Non-English Dialogue Quality

👁 119 reads

Kling 3.0 handles Polish speech poorly: Polish isn’t on the official language list for either Kling 3.0 or Kling 4.0. Update, October 21, 2026: Kling 4.0 spoke Polish without a single word error in my test (3 of 3 lines in Whisper). We tested it in Polish with the tongue-twister “W Szczebrzeszynie chrząszcz brzmi w trzcinie, a ja płacę trzydzieści dwa złote” (“In Szczebrzeszyn a beetle sounds in the reeds, and I’m paying thirty-two zlotys”) using the latest Kling available via API (3.0 Turbo and 3.0 Pro), added sentences in English and German, and compared the results with MiniMax H3, which readers recommend for Polish dialogue. We didn’t test Kling 4.0 itself because it’s not yet available in the API.

What Languages Does Kling Support

The speech language list changes with each version, but Polish isn’t in any of the official ones. Below we list what the producer stated themselves, without interpretation.

Kling 4.0 also supports Chinese dialects (including Beijing, Taiwanese, Sichuan, Cantonese) and English accents (American, British, Indian). In one paragraph of the note, Hindi appears instead of Indonesian, so the list isn’t entirely consistent. The official multilingual examples in the note are in Spanish and Korean, which we verified by running them through Whisper.

Official Kling AI example → watchStronger Multilingual CapabilitiesKling 4.0 release notes on kling.ai, 15.10.2026

How We Conducted the Test

On September 28, 2026, we checked which Kling was the latest in the API. OpenRouter had 3.0 Pro, 3.0 Standard, and O1; fal.ai additionally had 3.0 Turbo and 3.0 Omni. Kling 4.0 and 4.0 Flash weren’t available anywhere. So we tested three variants via fal.ai:

• Polish only, 5 seconds: Kling 3.0 Turbo Pro, Kling 3.0 Pro, MiniMax H3, and H3 Max. The Turbo, H3, and H3 Max clips came from our test for the video model ranking on the same day, with identical prompts, so we didn’t regenerate them;
• Three languages in one clip, 10 seconds: Kling 3.0 Turbo Pro and MiniMax H3. In Polish the tongue-twister from Szczebrzeszyn, in English “She sells seashells by the seashore for thirty-two dollars,” in German “Fischers Fritz fischt frische Fische für zweiunddreißig Euro.”

The scene was the same everywhere: woman at a kitchen table speaking to camera, static shot, daylight, no subtitles or music. We generated each clip once, without rerolling. We checked the speech two ways. First faster-whisper large-v3 (on CPU), which measures how many words can be correctly recognized (WER, percentage of incorrect words after number normalization). Then we gave just the audio track to Gemini 3.8 Flash with a request for phonetic evaluation: accent, mispronounced words, intelligibility from 1 to 10.

Results: One Sentence in Polish

Two things surprised us. Kling didn’t translate the Polish line to English, even though the version 3.0 guide promises that: it tried to speak Polish, just with foreign phonetics (in the three-language clip Gemini recognized it as American). Whisper in automatic mode didn’t even recognize the language of the Turbo recording (it identified it as Armenian with 0.32 confidence). The second thing: the more expensive Kling 3.0 Pro performed worse than the cheaper Turbo. With one attempt per model this could be chance, but neither of the two Klings came close to H3.

Results: Polish, English, and German in One Clip

This is the most interesting result of the entire test. It appears both models have one “native” voice: Kling English, H3 Polish. Changing language mid-clip doesn’t change the accent, only the words. This is our interpretation of two clips. Whisper won’t catch this, because it recognized H3’s English sentence flawlessly despite the Polish pronunciation. Only the phonetic evaluation showed the difference: in H3 “she” sounds like “si,” and “the” like “de.” The German sentence was difficult for both, neither model did it well.

Gemini’s evaluation is a second opinion, not an oracle: the model received just the audio and the text that was supposed to be spoken. Its conclusions agree with the Whisper numbers in every case where Whisper could measure something.

Can Kling 4.0 Handle Polish Dialogue?

Yes. Update, October 21, 2026: we generated Kling 4.0 clips in kling.ai, including the same tongue twister. Whisper large-v3 transcribed it with zero word errors, and two more Polish lines in other clips were also error-free (3 of 3). The Kling 4.0 note still doesn’t list Polish (the list ends with “and more”), so treat this as a tested result on a small sample, not an official promise. What’s known about the new version, we’ve gathered in the guide to Kling 4.0.

How to Create Polish Speech in Video Today

• Polish dialogue directly in the clip: MiniMax H3 or H3 Max. In our test H3 Max was best and cheapest (0.04 USD per second in 768p). More about the model in the article Hailuo AI and MiniMax H3.
• Image from Kling, voice separately: generate a clip without speech, record or synthesize a Polish line, and match lip movement. Kling has a separate lip-sync mode from audio for this (in fal.ai as kling-video/lipsync/audio-to-video). We didn’t test this mode.
• English narration, subtitles: if the material can be in English, Kling speaks it cleanly, as seen in both our test and the producer’s examples.

Official Kling AI example → watch„A continuous 7-second shot set in the UK. Two friends who have not seen each other for a long time meet…”Kling 4.0 release notes on kling.ai, 15.10.2026

We compare other models on the same prompts in the AI video generator ranking.

Frequently Asked Questions

Does Kling Speak Polish?

Not officially. Polish isn’t on the language list for either Kling 3.0 or Kling 4.0. In our test Kling 3.0 pronounced the Polish sentence with foreign pronunciation, and Whisper misrecognized 33 to 83 percent of the words.

What Languages Does Kling 4.0 Support?

According to the note from September 28, 2026: Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French, and Indonesian “and more,” plus Chinese dialects and English accents.

Which Video Generator Speaks Polish Best?

In our test from September 28, 2025, MiniMax H3 Max performed best: Polish tongue-twister with no incorrect words, for 0.20 USD for 5 seconds. Regular H3 got one word wrong.

Does Kling Translate Polish Text to English?

The Kling 3.0 guide promises this for unsupported languages, but in our test it didn’t do it. It tried to speak Polish, but with foreign phonetics.