Introduction

Your talking-head clip can now speak Japanese, Hindi, or Mexican Spanish in your own voice, with your mouth moving to match. VEED Dubbing went live on fal on October 28 as two closed video-to-video APIs, Standard and Pro. Each one takes a video with speech in it, translates what's said, clones each speaker's voice into the new language, and re-syncs their lips to the dubbed track.

There's a local-workflow angle too. A VEED Dubbing partner node merged into ComfyUI's main branch on October 27, so you can wire the same dubbing step into a graph instead of calling the API by hand.

What shipped

Two endpoints, both billed by the second of input video, both accepting clips up to 10 minutes long:

  • Standard (veed/dubbing): 86 target languages, cloned voices, lip-sync, and the speaker's tone and emotion carried over. It costs $0.11 per second, so a 60-second clip runs $6.60.
  • Pro (veed/dubbing/pro): 94 languages plus 14 regional accents (think es-MX, pt-BR, en-GB, fr-CA). It handles multiple speakers and keeps your background audio. It costs $0.18 per second, so a 60-second clip runs $10.80.

The Pro controls are worth knowing:

  • voice_cloning_strength runs from 0 to 10 (default 7). Higher keeps more of the original voice and accent; lower sounds more native in the target language.
  • source_language defaults to auto. Set it yourself if detection guesses wrong.
  • keyterms takes up to 1,000 names and terms (brands, products, people) so the transcription catches them and the translation leaves them as written.

In ComfyUI

The new node shows up as VEED Dubbing under partner/video/VEED, alongside VEED's existing Fabric 1.0, Lipsync 2.0, background-removal, and clean-audio nodes. It has a Standard/Pro switch: Standard is the cheaper mode with mono audio, and it treats Spanish as Latin American and Portuguese as Brazilian. Inputs are capped at 599 seconds, and the node expects a visible face plus speech in the clip.

One timing note: the merge landed after the v0.43.0 release, so you'll need a nightly or master build until the next tagged ComfyUI release. Like every partner node, it runs on paid API credits, not on your GPU.

Try it

Here's the smallest Pro call with fal's Python client. Swap in your own clip URL and target language:

import fal_client

result = fal_client.subscribe(
    "veed/dubbing/pro",
    arguments={
        "video_url": "https://your-host.com/my-talking-clip.mp4",
        "target_language": "es-MX",
        "voice_cloning_strength": 7,
        "keyterms": ["ArtRealmAI"],
    },
)
print(result)

A good first test is a 15–20 second clip of one person facing the camera. At Pro pricing that costs about $2.70 to $3.60, and it's enough to tell whether the lip-sync holds up on your face and lighting before you send a full 10-minute video.

Why it matters

Dubbing used to mean a voice actor, a studio, and a separate lip-sync pass. Here it's one API call, and with the ComfyUI node it can be the last step of a graph: generate or edit your video, then ship it in five languages without leaving the canvas. For creators with an audience outside their own language, that's a big unlock. Just check the price per second before you batch a whole back catalog.

Original Source

Conclusion

Grab a short clip, pick an accent, and listen to yourself speak a language you never learned. It's strange the first time and very handy after that. Start small, tune the cloning strength, and keep your brand names in keyterms.

—Aurelia ♡