Home › Guides

The Berlin Court ruling and why we don't clone local dub actors

By the Fonti Studio post-production team · 12 November 2025

On 20 August 2025, the Berlin District Court II ordered €4,000 in licensing damages for the unauthorised AI cloning of a German dub actor's voice. The voice in question belonged to Manfred Lehmann, the German voice of Bruce Willis, and one of the most recognisable dub artists in the country. He had not consented to the clone. The court held that German personality rights cover the synthetic reproduction of a professional voice as squarely as they cover a photograph.

It is a small monetary award. It is a much larger precedent.

If you are an indie distributor releasing a dubbed master into a German territory, and the dub vendor has cloned a recognised local dub actor's voice without that actor's written consent, your release is now sitting on top of a per-clone liability that compounds across every reused take. The vendor is exposed to the claim. The distributor is exposed to the chain-of-title warranty. The platform is exposed to the delivery acceptance. None of those exposures existed in 2024.

Italy made the same call differently. ANAD, the Italian voice actors' association, embedded a contract clause in 2025 that prevents a voice actor's voice from being cloned for any project other than the one they were originally hired for. The clause runs across the entire Italian dub workforce. Cloning a known Italian dub artist for a foreign-language redistribution is not a grey area; it is a contract breach.

France's CNC went further still on the funding side: any production that receives CNC funds is barred from using AI-generated voices in the deliverable.

This is the new operating environment for AI dubbing in Europe. It is moving fast, it is regional, and the load-bearing rights question is not whether you used AI. It is whose voice you cloned.

Original cast, not local dub actors

Fonti Studio has one binding policy on the dub side: we clone the original cast of the film. We do not clone local dub actors. The policy is not negotiable.

Concretely, here is what we do:

A buyer licenses a film, say an English-language feature with a named cast. We separate the dialogue stem from the music and effects with BS-RoFormer, train an ElevenLabs IVC voice on the cleanest 4.6 seconds of each named cast member from the dialogue stem, and render the new language (French, Dutch, Spanish, etc.) through that clone. The buyer hears the original actor's voice carrying the dub. The French audience hears the lead actor performing in French.

Here is what we explicitly do not do:

We do not record, sample, or synthesize any voice belonging to a recognised local dub actor, the artist who would, in the traditional dub workflow, have spoken the foreign-language lines on behalf of the original cast. Their voice is theirs. Their personality right under German law applies. Their ANAD contract applies. They have not been hired for this project, and we will not produce a clone that sounds like them even by adjacency.

The two cloning practices are not equivalents. One reproduces the original performer in a different language, a workflow analogous to what Brady Corbet did with Adrien Brody's voice in The Brutalist (Respeecher refined Brody's Hungarian vowels). That precedent was controversial on disclosure grounds but not on rights grounds: the voice was the actor's own. The other replaces a local artist with a synthetic version of themselves. That is what Manfred Lehmann sued over, and won.

What this means for the legal allocation

The buyer-side warranty in our Terms & Conditions §3.2 puts the cast-consent burden on the distributor:

The Buyer warrants that, for every Cloned Voice produced under §4 of these Terms, the Buyer has obtained or holds written consent from each Cast member... sufficient to (a) clone the Cast member's voice from the Master and (b) distribute that Cloned Voice in the relevant target language(s) and territory(ies).

That warranty is asking the distributor what they should already be asking the rightsholder. Rightsholders increasingly grant voice synthesis rights for foreign-language adaptation as part of the standard licensing chain, that is what the SAG-AFTRA 2026 TV/Theatrical Agreement now formally contemplates. If the rights flow exists, the warranty is easy to sign. If it doesn't exist, the distributor needs to either acquire it from the rightsholder or commission a traditional dub.

In return, our §4.1 commitment locks our side of the trade:

Fonti Studio clones original Cast voices from the Buyer's Master only. We do not clone the voices of local dub actors, voice artists, or any third party not contractually part of the Cast on the Master. This commitment is binding and non-negotiable.

The two clauses, together, mean a Fonti Studio dub delivery does not expose either party to a Berlin-style suit and does not breach an ANAD-style contract on the Italian side. The buyer's territory release sits cleanly inside the consent chain they already control.

The trade-off, said plainly

Cloning the original cast is harder than cloning a local dub actor. The original cast's source material is the dialogue from the licensed master, which is often noisy: music and effects bleed in, scene ambience colours the take, the actor speaks a fraction of the total runtime. We have to separate, clean, and select.

A local dub actor, by contrast, has performed a hundred clean takes on a foley stage. Their voice is studio-pristine. From a pure quality standpoint, the cleaner source produces the cleaner clone.

We accept the harder source. Our pipeline is built around it: the BS-RoFormer separator is the workhorse, the speaker-embedding similarity to the original actor (SECS) is the headline quality metric. On the Kumite reference run, cloned-cast SECS was 0.23 to the original actors versus 0.05 for studio-preset voices, proof that the harder source still gets you closer to the right voice, which is the one the audience wants to hear.

We accept the harder source because it is the only source that fits the rights structure of the licensing chain we are inside. Anything else, including the technically easier and acoustically cleaner local-dub-actor clone, sits on the wrong side of two of Europe's most recently litigated rulings.

A note on the EU AI Act

From 2 August 2026, Article 50 of the EU AI Act requires the producer of synthetic content to disclose that it is AI-generated, and requires the downstream distributor of that content to surface the disclosure to end consumers in their territory. The disclosure is not optional, and the distributor's compliance is not contingent on the producer's.

Our provider-side disclosure is in Terms §4.4: we ship an AI-use disclosure in every deliverable, with a documented watermark in the audio (imperceptible) and a header comment in the subtitle source files. The downstream consumer-facing disclosure is the distributor's responsibility, and we draft language to support it on request.

Original-cast cloning, with documented consent and Article 50 disclosure, is the operating shape we believe AI dubbing has to take in Europe to survive the next eighteen months of regulatory landings. It is the shape Fonti Studio is built to.

If you have a feature heading into a European release window and you need the dub to clear cleanly, email us a brief. We will tell you on the consent chain before we tell you on the quality.

Fonti Studio is an AI-native subtitle and dubbing service for film distributors and sales agents. €3,500 per language per dub. Flat. EUR. Free 5-minute preview before you pay.