cd /news/artificial-intelligence/woobladub-ai-dubbing-that-preserves-… · home topics artificial-intelligence article
[ARTICLE · art-105011] src=woobladub.com ↗ pub= topic=artificial-intelligence verified=true sentiment=· neutral

WooblaDub: AI dubbing that preserves the original picture, music and ambience

WooblaDub, a new AI dubbing workspace, preserves the original picture, music, effects, and ambience while allowing directors to rebuild only the dialogue, with the company stating it never holds or trains on voice models. The platform, currently in evaluation with launch partners and with no language route yet passing locked-media review, offers a six-pass workflow from importing the cut to delivering stems, script, cue sheet, and QA record, and positions itself against automated translators and visual dubbing by keeping the picture untouched and checksummed, preserving audio elements, and letting users connect their own voice provider accounts.

read6 min views1 publishedAug 20, 2026
WooblaDub: AI dubbing that preserves the original picture, music and ambience
Image: source

Picture locked · Sound in progress

WooblaDub is a dubbing workspace. You rebuild the dialogue and nothing else — the picture is never re-rendered, and the music, effects and ambience of the original scene stay underneath it.

Nooooo!

Neeein!

Directionsustained panic, collapsing at the end

All 6 lines #

Thirteen seconds of Tears of Steel, English into German. Six lines, two characters, one argument. Music, effects and room tone are the original recording in both versions; only the dialogue was rebuilt.

A dub is not a translation. It is a second performance, and someone has to direct it. #

This is a working environment, not a button. You cast the voices, shape the reading and adjust the line against picture. The machine does the labour; the judgement stays with you. We are in evaluation with launch partners, and no language route has passed a locked-media review yet — the Languages page says which ones are close.

A multilingual dubbing production workspace. Dialogue is rendered through a voice provider account you connect — a licensed stock voice, or one you hold the rights to. The adaptation, casting, direction and approval stay with you. Picture-immutable by default.

An instant video translator. A face-regeneration product. A voice-model owner. An unattended one-click output.

Locked cut to delivery, in six passes. #

§ 02 — The workflow### Bring in the cut

Video, screenplay, SRT or transcript. The timeline is built from what you already have.

Adapt the dialogue

Translate, then rewrite for mouth and rhythm. Source, gloss and adaptation on one screen.

Cast the voices

One voice per character across a whole title. Casts bind to characters, so they survive a season.

Direct the take

Quieter. Further back in the room. Slower. You write the note a director would give.

Review and approve

A-B the original against the dub anywhere on the timeline. Every decision keeps its history.

Deliver the record

Final video, stems, cue sheets, subtitles and the QA record a distributor asks for.

Three things get called dubbing. #

§ 03 — The comparison| Automated translator | Visual dubbing | WooblaDub | | |---|---|---|---| | The picture | Untouched | Re-rendered to fit the new dialogue | Untouched, and checksummed | | Music, effects, room tone | Usually flattened into one track | Preserved | Preserved, provenance recorded | | Changing one line | The whole output regenerates | The whole output regenerates | That cue reruns | | How you direct it | A prompt | A prompt | A note, cue by cue | | Whose voice models | The vendor's | The vendor's | Yours, in your own account | | What you receive | A finished file | A finished file | Stems, script, cue sheet, QA record |

The performer decides. The editor is accountable. #

§ 04 — Consent & rightsThere is no mechanism in WooblaDub that would let us own, resell or clone a voice, because we never hold one.

We do not own voice models

You connect your own provider account. Consent stays inside the agreement you already hold.

We do not train on your material

Your material produces your dub. It does not improve output for anyone else.

No performer is synthesised without agreement

Where a provider holds a performer’s consent, that provider stays the record of it.

An editor of record signs the work

Every approval carries a name and a date. Since 2 August 2026 the EU AI Act has expected someone to be identifiable as responsible for a synthesised performance. That is the same person.

Your project leaves with its production files. #

No black box. Everything the dub was made of leaves with you, in the formats a mixer and a QC house already use.

A different scene, English into Spanish. Every element you can hear other than the dialogue shipped in the package as its own stem.

Launch focus · film, episodic and vertical drama

Launch focus · European film and television

Launch focus · feature and episodic

Available after language and voice review

Priced per project, from one scene to a full title. #

§ 06 — Access### Per project

One scene, one episode or a full title. No subscription.

  • Editor and revision history
  • Production files included
  • Storage billed separately

Priority production

Faster queues, series continuity, deeper delivery controls.

  • Shared casts and glossaries

  • Team review and approvals

  • Priority processing

  • DAW-ready interchange

The questions that come up first. #

Access opens in stages. Create an account, tell us what you are dubbing, and we will set the room up around it.

Whose voices are these? #

Yours to choose. Dialogue is rendered through a speech provider account you connect, so the voice is either a licensed one from that provider's pool or one you hold the rights to and bring yourself. WooblaDub never owns or holds a voice model. What is human in every case is the adaptation, the casting, the direction of each reading and the approval — each carrying the name of the person who made it.

Do you change the actors' faces, and how does lip sync work without that? #

No, and by writing to the mouth rather than changing it. Adaptation is timed against the source speaking window and the visible articulation, which is the constraint a dubbing adapter has always worked inside. Source and output video stream hashes ship with every delivery, so that the picture was untouched is checkable rather than promised. Products that reanimate the performer's face solve it the other way, and on a tight close-up they solve it better — that is a different product with different consent questions.

Whose voice is being cloned? #

None, unless you bring one and the rights to it. The default is a stock voice from your provider's licensed pool. We do not own or sell voice models, and we will not synthesise a performer who has not agreed to it — the digital-replica terms your production is bound by are the terms this workflow runs inside.

What happens when I change one line? #

Only the affected cue reruns, along with whatever depends on it. The revision graph shows what changed and what stayed.

Is the M&E reconstructed? #

If you supply one, it passes through untouched and is recorded as supplied. If you don't, we separate one and label it a reconstruction, and the reconstruction has to prove itself: the separated dialogue and background must sum back to the original mix, channel by channel, above a published fidelity threshold. That is not the same thing as a filled M&E: where footsteps and cloth are buried in the dialogue, recovering them is a foley session, not a processing step.

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @woobladub 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/woobladub-ai-dubbing…] indexed:0 read:6min 2026-08-20 ·