Skip to content
trending

Benchmarks

VEED Dubbing on fal.ai: Polish Voice Cloning Test

I tested VEED Dubbing on fal.ai with Polish and English clips. Pro delivered cleaner Polish; Standard left English words in its output.

By 3 min read
VEED Dubbing on fal.ai: Polish Voice Cloning Test

👁 117 reads

VEED Dubbing, released on fal.ai on October 28, 2026, translates the speech in a video, clones the speaker’s voice and re-syncs the lips, for $0.11 per second of input (Standard) or $0.18 (Pro). In my test, English to Polish and Polish to English, Pro got the meaning right in both directions, while Standard left English words in its Polish output. I picked Polish because it is a hard, heavily inflected target that cheap models often fumble. Three runs on two 5-second clips cost about $2.40.

Standard covers 86 languages, Pro 94 plus 14 accents

VEED Dubbing Standard VEED Dubbing Pro
Target languages 86 94 and 14 regional accents (for example, pt-BR, es-MX and en-GB)
fal.ai price $0.11 per second of input video ($6.60 per minute) $0.18 per second ($10.80 per minute)
Video length Up to 10 minutes Up to 10 minutes
Extras A list of proper names and terms to preserve The same, plus voice-cloning strength (0-10), accent selection, multiple speakers and background-audio preservation

The model can detect the source language automatically, or you can select it yourself. A list of up to 1,000 terms can help preserve brand names and people’s names during translation. Voice-cloning strength works as fal describes: higher settings retain more of the original voice’s timbre and accent, while lower settings make the new language sound more natural.

My test: English to Polish and back

I used two five-second clips generated earlier by AI. In the first, a barista says, “Your flat white is ready. Enjoy it.” In the second, a woman at a table says a Polish tongue twister: “W Szczebrzeszynie chrząszcz brzmi w trzcinie, a ja płacę 32 złote.” (“In Szczebrzeszyn a beetle buzzes in the reeds, and I’m paying 32 zlotys.”) I transcribed every result with Whisper speech recognition instead of judging by ear.

Dealat Kinetyra
Typeless Pro - 12 Months
$49 instead of $144 -66%
See the deal →
Run Output Assessment
Pro, English to Polish “Twoje flat white jest gotowe. Smacznego!”
(“Your flat white is ready. Enjoy your meal!”)
Correct, with the natural Polish “Smacznego” instead of a literal “enjoy it”
Standard, English to Polish “for your flat white is gotova. Smacznego.”
(English words left in, then “Enjoy your meal.”)
A mix of languages; English words remained in the Polish version
Pro, Polish to English “In Scebzheshen a beetle buzzes in the reeds, and I’m paying 32 zlotys.” The meaning was good, but the town’s name was rendered phonetically

Pro processed each clip in 48-62 seconds; Standard took 75 seconds. The frames show that the model altered the mouth movements while leaving the face, background and lighting unchanged. A five-second clip isn’t enough to assess lip sync in longer speech.

Before-and-after frames from VEED Dubbing Pro: English to Polish and Polish to English
Left: original; right: VEED Dubbing Pro. Top row: English to Polish; bottom row: Polish to English. The mouth movements changed, while the rest of the frame stayed the same.

Listen for yourself

Original in English:

VEED Dubbing Pro, Polish:

VEED Dubbing Standard, Polish:

Original in Polish, followed by VEED Dubbing Pro in English:

Five things to know before you dub

  • Choose Pro for harder languages. Standard left English words in my Polish test. That was one run on a short clip, but I’d consider the cost of that kind of mistake in an ad greater than the price difference.
  • Your video needs an audio track. The example file in fal’s documentation has none, and the API rejects it.
  • You pay per second of input. A 10-minute video costs $108 with Pro. Test a short clip first.
  • Add proper names to the term list. I didn’t add “Szczebrzeszyn” in my test, so the model rendered the name phonetically.
  • Consider whose voice you’re cloning. The model clones the voice of the person in the recording. Only use recordings you have rights to and get consent from the people involved.

Other ways to dub a video, including apps that need no API, are in my guide AI dubbing: how to translate a video while keeping the original voice. Per-second prices of video models are compared in AI video cost per second.

// read next

More on this from QUANTSLANT

Piotr Olszewski

Editor and publisher · Quantslant

I edit and publish Quantslant. AI tools help with research, drafts and translation; choosing topics, running the tests, checking facts and the decision to publish are mine, and so is the responsibility for what appears here.

read nextguidesAI Agent Decision Models: Choosing Tools and Asking Follow-UpsDesign an AI agent to choose between answering, calling a tool, asking a follow-up, refusing, or handing off - and validate each step.

Most read

  1. 1Best AI Video Generator Apps for Your Phone (2026)
  2. 2Claude Opus 5.5 Motion Graphics: 60s Showreel from One Prompt
  3. 3Wan 2.2 on RTX 4090: timings, VRAM and four test clips
// site map

This page in other languages5

Quantslant in numbers 1,423posts in archive22news from the last 7 days18video models in the observatory
🎬 AI Video Benchmarks: Higgsfield DoP vs OpenAI Sora • Higgsfield vs Runway Gen-3 • AI Video Credit Calculator
× ‹ enlarge ›