# Prime Video is finally tackling the dubbing uncanny valley with lip-syncing AI > Source: > Published: 2026-09-09 18:00:22+00:00 # Prime Video is finally tackling the dubbing uncanny valley with lip-syncing AI Watching a foreign show with a dub usually means ignoring the fact that the actor's mouth is moving in a completely different rhythm than the audio. Prime Video is trying to fix this by using AI to re-animate the lip movements to match the translated audio. Right now, this is only live for the English dub of the German series Maxton Hall. If you've seen the show, the shift is noticeable—it's not a perfect 1:1 match yet, but it removes that jarring disconnect where a character finishes speaking and the audio keeps going for another two seconds. ## How this differs from standard auto-dubbing Most AI dubbing tools, like the ones Meta and YouTube have been rolling out for creators, focus on the audio translation and synthetic voice cloning. They get the voice right, but they leave the video untouched. Prime Video is layering a visual effects pass on top of the audio. Instead of just swapping the sound, they are modifying the actual pixels around the mouth to align with the phonemes of the English dub. From a technical standpoint, this is much heavier than a simple audio overlay. It requires frame-by-frame manipulation of the actor's face. While they aren't releasing the specific model version or the exact latency of the render, the result in Maxton Hall suggests they are using a generative fill or warping technique similar to what you see in high-end deepfake software, but constrained to the mouth area to avoid "melting" the rest of the face. ## Where the tech still fails Despite the polish, this isn't a magic bullet. I noticed a few specific issues while testing the output: - **Micro-expressions:** When an actor is speaking while laughing or crying, the AI occasionally flattens the emotional nuance of the mouth to force the lip-sync. You lose the "acting" in favor of the "matching." - **Lighting inconsistencies:** In some scenes with harsh side-lighting, the AI-generated mouth movements create a slight shimmer or a blur that doesn't match the rest of the skin texture. - **Processing cost:** This isn't something that can be done in real-time for a live stream yet. It's a post-production process. The cost of rendering an entire series like Maxton Hall with this tech is significantly higher than traditional dubbing. ## Comparison with other platforms - **YouTube/Meta:** Fast, automated, but purely audio-based. Great for 10-minute vlogs, but looks amateur for cinema. - **Prime Video:** Slow, VFX-integrated, and curated. It's designed for high-fidelity prestige TV where the visual quality cannot drop. [Next Managing MLflow models across multiple AWS accounts is the only →](/en/threads/9048/) [a practical ChatGPT prompt guide](https://tanyan888.com/), with plenty of directly applicable cases. ## All Replies （3） So glad this is happening. I usually struggle with subtitles on my 65-inch TV, but maybe this works with Flawless AI? Finally! I can't stand how distracting the mouth mismatch is during K-dramas. Does this actually work with the 4K streams? I want to try this tonight. Does the processing happen server-side or is it some local plugin like Wav2Lip?