We Tested SwitchX Live While Seedance Sparked a Deepfake Crisis

The latest Denoised covers three stories that capture the current state of AI filmmaking: a practical tool that might actually fit into existing workflows, a powerful model raising serious copyright questions, and a high-profile director testing whether AI can serve traditional storytelling.

Hosts Joey Daoud and Addy Ghani test Beeble's new SwitchX live, break down Seedance 2.0's viral launch amid deepfake controversy, and debate whether Darren Aronofsky's AI series proves anything about the technology's creative potential.

Can AI background replacement compete with LED volumes? Should it? This episode tests Beeble's SwitchX live, debates whether Seedance 2.0's deepfake capabilities cross a legal line, and questions whether Darren Aronofsky's AI Revolutionary War series proves anything about the technology's creative potential. Three tools, three approaches, one question: what tradeoffs will audiences accept?

Joey and Addy run Beeble's new video-to-video model during the episode, feeding it source footage with alpha masks and reference images. The results? Faces stay intact through environment transformations — the model's core strength.

The LED Volume Question:

Traditional LED volumes cost $50,000+ and require on-set expertise. SwitchX offers a post-production alternative: shoot against a simple background, transform the environment later.

But when does that make sense? For indie filmmakers without LED access, maybe always. For larger productions, it's a calculation: real-time actor feedback vs. post flexibility, interactive lighting vs. iteration cost.

Current Limitations:

Try it yourself: SwitchX is in free beta (cloud-based, no local hardware needed)

ByteDance launched Seedance 2.0 with major upgrades: 15-second multi-shot output (up from 5 seconds), stereo audio, multimodal inputs. It went viral in China — state media compared it to DeepSeek's impact.

Then came the deepfakes.

Users generated videos of "Tom Cruise" fighting "Brad Pitt" on a rooftop. The Motion Picture Association told Deadline: "In a single day, the Chinese AI service Seedance 2.0 has engaged in unauthorised use of US copyrighted works on a massive scale."

Joey and Addy's take:

The quality has crossed a threshold. These aren't obviously fake anymore. The scam potential is real — fake celebrity endorsements, fraud, worse. The technology has outpaced the legal frameworks designed to contain it.

The question: Is this a tool problem or a guardrail problem? And who's responsible when the guardrails don't exist?

Darren Aronofsky's AI studio Primordial Soup released On This Day... 1776, a short-form animated series using Google DeepMind's Veo for visuals and SAG-AFTRA voice actors for performances.

The backlash was harsh. Critics called it "slop" — unnatural camera movements, melting faces, copy-pasted soldiers.

Joey's argument:

AI makes sense for content that wouldn't exist otherwise. Period pieces are expensive: costumes, sets, extras. If the choice is AI historical content vs. no historical content, the calculation changes.

Addy's counter:

But this project has resources. A major director. A TIME distribution deal. If this is the best-case scenario for AI filmmaking, what does it say about the technology's readiness?

The comparison: They contrast the 1776 series with action sequences from The Patriot (2000). Real stunt performers, real pyrotechnics, real camera work create a different kind of impact. The physicality, choreography, and craft — that's what AI currently lacks.

The question: Is "it wouldn't exist otherwise" a sufficient justification when the result doesn't meet audience expectations?

Which approach wins may depend less on the technology itself and more on whether audiences accept the tradeoffs each one requires.

Tools & Platforms:

Projects:

News & Analysis:

WATCH THE FULL EPISODE ON YOUTUBE →