The latest Picsart Flow release is live, and it puts an AI video translator right on the canvas: the new Translate node. Hand it a video of someone speaking, and it hands the video back in another language, in the same voice, with the lips moving to the new words.

That last part is the leap. A dub usually gives itself away: a stranger’s voice, or a mouth that keeps moving after the sentence has ended. The Translate node goes after both. Flow detects the spoken language, translates the speech, clones each speaker’s voice and re-syncs the lips, so the result looks and sounds like the original was recorded in the new language. It is live now in Picsart Flow.

The Translate node dubs a video into any language

AI video translation now lives inside the canvas, as one node. Feed it a clip with clearly audible speech, and a single run carries the whole dub from start to finish:

  • Detects the spoken language. The node works out on its own what language the clip is in.
  • Translates the speech into the language you picked.
  • Clones each speaker’s voice. Every person on screen keeps the voice the audience already knows, now speaking the new language.
  • Re-syncs the lips so every mouth moves with the translated words, not the original ones.

One video goes in and one translated video comes out. The dub lands on a new node right next to the original, so the source and the translated version sit together on the canvas, ready to play back to back.

190 target languages, regional variants included

Open the Language picker and 190 target languages roll out. Regional variants get their own entries instead of hiding behind one parent language, so the video translator can aim at the exact version of a language an audience speaks. A few from the list:

  • Spanish (Mexico)
  • English (UK)
  • Chinese (Cantonese, Traditional)
  • Arabic (Egypt)

Every one of the 190 gets the full treatment. The speech is translated into that language, carried in the cloned voice of each speaker, and matched by re-synced lips.

No transcript, no prompt, no source language to set

Video translation usually starts with homework: a transcript to clean up, a source language to declare, instructions to spell out. The Translate node skips all of it. Flow detects the spoken language by itself, so there is not even a field for it.

There is no transcript to upload and no prompt to write. The Language picker is the only required choice on the node. Pick the language, pick a mode, and the setup is done.

Two modes for AI video dubbing: Speed and Precision

The Translate node ships with two modes, each tuned for a different priority. Speed is for fast turnaround with standard lip-sync. Precision re-renders the mouths for high-quality lip-sync and makes the translation context- and gender-aware.

Mode Lip-sync What else it gives you
Speed Standard lip-sync Fast turnaround
Precision High-quality lip-sync with re-rendered mouths Context- and gender-aware translation

Reach for Speed when turnaround is what counts and standard lip-sync does the job. Switch to Precision for clips where the speaker’s mouth is front and center, and where the translation has to follow context and gender to land right. Either way, each speaker’s voice is cloned and every run returns one translated video.

One clip, as many languages as you need

Each run produces one language, and the canvas turns that into a launch plan. To ship the same video in five markets, branch five Translate nodes from the same clip and set a different language on each.

Every branch is a separate run, and each one drops its translated video on a new node next to the original. The original never leaves the canvas, so a sixth language next month is just one more branch from the same clip.

Branches do not have to match, either. One language version can run in Speed while another runs in Precision, because the mode is set on each node.

The cost shows before you run

Cost scales with two things: the length of the source video and the mode. A longer clip costs more than a shorter one in the same mode, and switching modes moves the price too.

No surprises at the end, though. The estimate shows on the node before you run, so the number is right there before you commit. Each Translate node is its own run with its own estimate, so a clip branched into five languages means five runs, each priced on its own node.

How to translate a video with AI in the new node

The Translate node takes MP4, MOV or WebM files of at least 2 seconds, with clearly audible speech. A quick spoken clip qualifies as well as a long one. From there, it is four steps from the original to the dub.

1. Connect a clip with speech

Connect an MP4, MOV or WebM video of at least 2 seconds, with clearly audible speech, to the Translate node.

2. Pick the language

Choose one of 190 target languages in the Language picker. The source language is detected automatically, and no prompt or transcript is needed.

3. Choose a mode

Speed gives fast turnaround with standard lip-sync. Precision gives re-rendered mouths, high-quality lip-sync and context- and gender-aware translation.

4. Check the estimate and run

The cost estimate shows on the node before you run. The translated video appears on its own node next to the original.

Also new in this release

Two canvas upgrades ship alongside the Translate node, and both make it faster to reach.

  • The right-click menu opens in two columns. What to create sits on the left, and what to do with it sits on the right.
  • Translate and Captions are offered wherever a video can connect. That means on the canvas, on a video node or on a video lane.

Tips for AI video translation in Picsart Flow

Start with clear speech

The node works from the speech in the clip, so pick a clip with clearly audible speech.

Leave the transcript out

The source language is detected automatically, and the Language picker is the one required choice.

Use Precision when the mouth matters most

Precision re-renders mouths for high-quality lip-sync and adds context- and gender-aware translation.

Read the estimate first

Cost scales with the length of the source video and the mode, and the estimate shows on the node before you run.

Branch for every language

One run makes one language, so give each market its own Translate node from the same clip.

Translate your first video in Picsart Flow today

The Translate node is live now. Connect a video with clear speech to it in Picsart Flow, pick a language and run. The dub appears beside the original, in the same voice, with the lips moving to the new words.

Get answers to common questions

It is an AI video translator on the Flow canvas. Connect a clip with speech, pick a language and run, and the video comes back in that language with the same voice and lips that match the new words.