Ancora: What Persists. What Returns.

Every object carries two lives. The one it had, and the one it's becoming. Ancora holds both in a single form.

The core concept is the tension between old and new, not blended, not transitioned. A sharp angular fracture where two eras coexist in one object. One side: 1970s Italian craft, aged leather, oxidized brass, raw texture. The other: contemporary minimalism, polished hardware, clean geometry. Two worlds forced together through a deliberate architectural break.The sharp split between old and new is exactly where fashion is right now.

The entire brand identity, product line, and visual world was built from scratch on Melius, first time using the platform. I used the agent throughout to brief directions, iterate on outputs, and refine prompts across multiple node types, from image generation to style references to video.

The process started with research. I searched for references that hit, luxury campaigns, editorial product shots, brand identities that carried real weight. Then I pulled in concept images, product presentations, textures, and mood references that matched the direction I wanted for Ancora. From there I fed everything into Melius, using those references as anchors so the AI could pull the best elements from each, the lighting from one, the texture language from another, the composition from a third. It wasn't about copying any single reference. It was about curating the strongest parts and letting the AI craft them into one cohesive brand identity. Every generation was guided by something real.

The biggest strength was being able to see the full creative process laid out spatially. Each node is a decision, each branch is an iteration. The canvas becomes the proof of the thinking, not just the output. Being able to feed in references and examples, then link them directly to generation nodes meant every product was born from something real. Nothing was random. Every output was inspired, guided, and intentional.

project link:https://app.melius.com/projects/596b09cc-f672-44ca-95cf-c7d8e4a583af/canvas/d71faf89-cabd-479b-9469-76eb8bae0049

instagram post link:https://www.instagram.com/reel/DYaDLkKAhhH/?utm_source=ig_web_copy_link&igsh=MzRlODBiNWFlZA==

Project Name: The Neon Nomad Cyberpunk Vision

Project Link: https://app.melius.com/projects/d9ec5588-2e8a-4758-b758-8473ec3f99c8/canvas/3afa0faa-0f32-4b42-9910-5818006ccdaa

Social Media Post: https://www.linkedin.com/posts/jose-arturo-6b5507332_melius-ugcPost-7461548659147427840-Yv6A?utm_source=share&utm_medium=member_desktop&rcm=ACoAAFO5jykBCSk0VemkPNBaENwE-LE_0HIHznU

Description: A 2K high-fidelity video showcasing a unique visual concept built with Melius agents.

Process Steps:

Formulated the creative concept and briefed the Melius Agent using multiple node types to generate the high-resolution core visual assets on my canvas.

Exported the SOTA model assets from the workspace for advanced post-production.

We generated short 2-second visual bridges between scenes to connect the continuous shot sequence. Due to spending some credits on failed video iterations, I had to use my brother's account to fill those remaining gaps; however, the entire project is still fully viable within the $50 credit allocation.

Secondary Project Link. https://app.melius.com/projects/59016f73-4abf-4991-8336-0d560318aca9/canvas/0e7ba64b-3bcc-4794-acdc-96d9b917c25b

The initial musical concept was created using Melius AI, yielding an excellent draft that perfectly matched the visual theme, and the final audio was later refined in Suno and editing software.

Implemented an external pipeline (Python/MoviePy) to handle native 2K rendering, precise audio synchronization, and smooth fade-outs without losing resolution.

Feedback on Melius:

The canvas node system and agent interaction are incredibly smooth for asset generation. However, adding native timeline support for precise audio editing and high-res export (2K/4K) within the platform would prevent the need for external post-production pipelines.

Another thing I didn't like is that the platform cannot work frame by frame, which makes things very complicated. If it could work that way, you could edit the video surgically. For example, giving a more precise instruction like: "in frame 550, an object appears in the character's hand, and I would like to remove it."