Could I push one identity reference across wildly different aesthetics—cyberpunk, classical painting meets sci-fi, horror cinematics, video game environments—while maintaining emotional arc?

The scenes are intentionally dense—layered environments, symbolic imagery, deliberate pacing—because I wanted to see if my creative instincts could translate through generative tools.

I pushed Veo 3.1 with complex VFX transitions, aggressive morphing sequences, multi-axis camera movements, dense scene compositions, and rapid environmental shifts.

Not to see what the system does well, but to find where it strains—and what that reveals about the road ahead.

Using my own face was intentional—I am the character. Across multiple scenes with heavy morphing and frame-to-frame generation, identity drift still appeared. Even with restructured prompts, explicit negative constraints, and reinforcement, the system would introduce a different Asian male face.

This isn't a bug—it's a structural challenge. Existing approaches—fine-tuning (LoRA, DreamBooth), embedding-based methods (IP-Adapter, InstantID)—each trade off consistency against flexibility. For narrative, advertising, or branded content, solving this is non-negotiable.

What would actually unlock this:

— Project-level identity references that persist across scenes
— Scene-level reference stacking (not constant re-uploads)
— Explicit reference carryover for frame-to-frame generation

Solving identity persistence doesn't just improve quality—it expands the category of stories creators can tell. This is where AI video shifts from demo to production tool.

In one scene featuring Guanyin, the intended motion was simple: water pours from a vase held in her left hand. Despite explicit prompts, masking, and negative constraints, the model repeatedly switched which hand performed the action.

This isn't a prompting failure—it's architectural. Current models don't maintain persistent skeletal tracking across frames. "Left hand" and "right hand" aren't stable internal concepts—the system optimizes for gesture realism frame-by-frame, not anatomical continuity over time.

I tried multiple approaches: anchoring the water origin spatially, keeping hands static, removing the pouring gesture initiation entirely. None reliably solved it.

The reality:

Until models maintain object-anchored reasoning across time, this remains a design-around constraint—not a prompting problem with a prompting solution.

Flow's Scene Builder is intuitive and fast. Extend seamlessly continues from the last frame; Jump maintains identity across cuts—both smart solutions to the 8-second generation limit.

But these solve continuity within a scene. The real unlock is transitions between scenes.

Most AI videos rely on jump cuts because that's what the tools make easy. Cinematic storytelling lives in the in-between moments: the match cut, the morph, the breath between scenes. Right now, those require manual work outside the tool.

What I'd build:

A lightweight transition layer—cut, match, morph, dissolve—where the AI generates the bridge between two different scenes. Select clip A, select clip B, choose transition type, generate. This reduces post-production fragmentation, encourages intentional pacing, and moves the tool from clip assembly toward actual scene composition.

Sound design and music work well at the 8-second scene level. In another Flow project, I explored Latin-inspired scoring, and the tonal quality held up.

The challenge is continuity. Each scene behaves as an isolated fragment—no throughline, no arc. And audio has harder unsolved problems: dialogue and lip-sync, voice consistency across scenes, music beats and rhythms that align across cuts, sound effects that match generated environments.

The unlock:

A final assembly layer for global music, transitions, voiceover, and audio continuity. That's where creators shift from assembling clips to shaping narrative. That's where AI video stops feeling experimental and starts feeling intentional.

Flow offers three export options: 270p animated GIF (not practical for production), 720p original (soft but coherent), and 1080p upscaled (sharper but artifact-prone). The 1080p neural upscaling often over-synthesizes—hallucinating texture detail that wasn't in the original across the entire image. On faces, the effect is especially damaging: the model adds what looks like wrinkles and skin imperfections, making faces look unnaturally aged or degraded.

The 720p original feels more visually coherent—just too soft for final delivery. Creators shouldn't have to choose between soft but natural and sharp but hallucinated.

This compounds in Scene Builder, where exports appear limited to 720p with no option for controlled upscaling at the scene level. That forces extra post-processing steps and inconsistent workflows—exactly the kind of pipeline fragmentation that pulls creators out of the tool.

What would help:

— Multiple upscaling profiles (neutral / filmic / sharp) with different synthesis aggressiveness
— 1080p+ exports directly from Scene Builder
— Face-aware upscaling that preserves rather than invents texture
— Quality parity with tools like Topaz

Early prompts often reference named artists—Dalí, Escher, Ocampo—to steer tone and visual language. But once initial images are generated, those AI outputs become the references for subsequent iterations. The lineage collapses into synthetic intermediates.

Downstream, I encountered something unexpected: three different scene images, each with what looked like a signature in the lower-right corner—similar in style but not identical. I hadn't prompted for authorship or attribution.

Generative models synthesize from learned patterns rather than retrieving stored works. But signatures appearing across multiple generations blur that distinction—raising questions about how stylistic influence propagates and re-materializes in ways neither creator nor platform can explain.

This is provenance drift: as creators iterate through AI-generated references, visibility into where influence originated rapidly degrades. For personal work, acceptable. For commercial contexts, ambiguity around attribution and compensation becomes harder to ignore.

This matters beyond ethics. The industry is splitting: AI-native production agencies are emerging rapidly, while traditional players—illustrators, VFX houses, unionized talent, brand legal teams—remain skeptical. On social media, the criticism is loud: AI "steals" artists' work, trained on their images without consent or compensation. Whether you agree or not, the perception is real and it's blocking adoption. Without provenance clarity, this divide deepens. The tools that build trust infrastructure—not just capability—will be the ones that bridge it.

What the industry needs:

— Influence tracking across generation chains
— Opt-in artist participation framework
— Mechanisms to acknowledge and compensate sources of inspiration

As generative tools mature, ethics isn't separate from UX—it's a question of observability. Building trust in AI-native creative systems may depend on it.

AI video is at an inflection point. We're moving from generation (make me a clip) to systems (help me build a film). The tools that win will be the ones that solve:

1. Identity — persistent characters across scenes and sessions

2. Continuity — transitions, pacing, narrative throughline

3. Control — project-level management, cost visibility, exportable quality

4. Integration — audio, voiceover, and post as first-class citizens

Google Flow has strong foundations. The UX is thoughtful, the creative ceiling is high, and the Scene Builder points toward the right abstraction. What comes next is the shift from creative toy to creative infrastructure.

Building Amor Fati meant developing a working methodology for constrained AI filmmaking:

— Single identity anchor: One reference image, used consistently

— Prompt iteration as exploration: Not optimizing for output, but mapping system behavior

— Deliberate breaking: Pushing morph intensity, scene complexity, and temporal jumps to find failure modes

— Assembly as authorship: The edit—pacing, juxtaposition, sound—is where the film actually gets made

If there's interest, I'm happy to share the full prompt structures, constraint frameworks, and assembly decisions that went into this project.