{"id":267254,"date":"2026-10-14T09:13:42","date_gmt":"2026-10-14T16:13:42","guid":{"rendered":"https:\/\/picsart.com\/blog\/?p=267254"},"modified":"2026-10-14T09:13:42","modified_gmt":"2026-10-14T16:13:42","slug":"ai-video-translation-in-picsart-flow","status":"publish","type":"post","link":"https:\/\/picsart.com\/blog\/ai-video-translation-in-picsart-flow\/","title":{"rendered":"Picsart Flow now has an AI video translator with lip-sync"},"content":{"rendered":"

The latest Picsart Flow release is live, and it puts an AI video translator right on the canvas: the new Translate node.<\/strong> 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.<\/p>\n

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<\/a>.<\/p>\n

The Translate node dubs a video into any language<\/span><\/h2>\n

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:<\/p>\n

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

    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.<\/p>\n

    190 target languages, regional variants included<\/span><\/h2>\n

    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:<\/p>\n

      \n
    • Spanish (Mexico)<\/li>\n
    • English (UK)<\/li>\n
    • Chinese (Cantonese, Traditional)<\/li>\n
    • Arabic (Egypt)<\/li>\n<\/ul>\n

      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.<\/p>\n

      No transcript, no prompt, no source language to set<\/span><\/h2>\n

      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.<\/p>\n

      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.<\/p>\n

      Two modes for AI video dubbing: Speed and Precision<\/span><\/h2>\n

      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.<\/p>\n

      \n\n\n\n\n\n\n
      Mode<\/th>\nLip-sync<\/th>\nWhat else it gives you<\/th>\n<\/tr>\n<\/thead>\n
      Speed<\/th>\nStandard lip-sync<\/td>\nFast turnaround<\/td>\n<\/tr>\n
      Precision<\/th>\nHigh-quality lip-sync with re-rendered mouths<\/td>\nContext- and gender-aware translation<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/figure>\n

      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.<\/p>\n

      One clip, as many languages as you need<\/span><\/h2>\n

      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.<\/p>\n

      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.<\/p>\n

      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.<\/p>\n

      The cost shows before you run<\/span><\/h2>\n

      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.<\/p>\n

      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.<\/p>\n

      How to translate a video with AI in the new node<\/span><\/h2>\n

      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.<\/p>\n

      \n \n
      \n
      \n
      \n

      \n 1.<\/span>\n Connect a clip with speech <\/p>\n

      Connect an MP4, MOV or WebM video of at least 2 seconds, with clearly audible speech, to the Translate node.<\/p>\n <\/div>\n <\/div>\n

      \n
      \n

      \n 2.<\/span>\n Pick the language <\/p>\n

      Choose one of 190 target languages in the Language picker. The source language is detected automatically, and no prompt or transcript is needed.<\/p>\n <\/div>\n <\/div>\n

      \n
      \n

      \n 3.<\/span>\n Choose a mode <\/p>\n

      Speed gives fast turnaround with standard lip-sync. Precision gives re-rendered mouths, high-quality lip-sync and context- and gender-aware translation.<\/p>\n <\/div>\n <\/div>\n

      \n
      \n

      \n 4.<\/span>\n Check the estimate and run <\/p>\n

      The cost estimate shows on the node before you run. The translated video appears on its own node next to the original.<\/p>\n <\/div>\n <\/div>\n <\/div>\n <\/section>\n\n