SShortSingh.
Back to feed

Solo creator documents AI filmmaking process behind sci-fi short sequel

0
·2 views

A solo content creator has shared a detailed production diary outlining the step-by-step process of making a sequel to their AI-generated short film 'I'm Not Going to Wake Them.' The original one-minute film, inspired by 90s dark sci-fi anime and the game Blame!, cost $16 in asset generation and took three days to complete. For the sequel, the creator is collaborating with an AI assistant named Simona for brainstorming plot ideas, generating visuals, and crafting scene breakdowns. The project uses a range of cost-saving techniques, including HTML manipulation via Chrome's debug port to create animated text effects locally at no cost. The series follows a ship AI called 'the Goddess' and aims to explore a dystopian narrative with an eventual hopeful resolution.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Developer Builds AI Tool That Converts Room Photos Into Before-and-After Videos

A developer has created RoomFacelift, an AI-powered interior design tool that goes beyond static image generation by producing short before-and-after transformation videos from a single room photo. Users can select their room type, preferred interior style, and how much of the space they want changed before the tool generates a redesigned version. A key challenge during development was preventing the AI from altering unintended elements such as camera angles, windows, doors, and room proportions, which the developer addressed by introducing two distinct redesign modes offering different levels of transformation. The platform supports over a dozen interior styles — including Scandinavian, Japandi, and Art Deco — each backed by a structured design recipe to improve consistency. RoomFacelift is built using Next.js, Supabase, fal.ai, and FFmpeg, and is currently live at roomfacelift.com.

0
ProgrammingDEV Community ·

How Using CloudFront as an API Gateway Exposed Security and Governance Gaps on EKS

An engineering team building a microservices-based identity platform on Amazon EKS initially routed both frontend and REST API traffic through Amazon CloudFront, believing it a natural fit since the CDN was already in use. While CloudFront successfully forwarded all required HTTP methods to the Application Load Balancer and Kubernetes Ingress, it was never designed to manage API lifecycle concerns such as consumers, stages, quotas, and authorizers. This architectural misstep meant the API origin remained publicly addressable, security policies were fragmented across multiple layers, and differentiated rate limiting for machine-to-machine clients required custom logic outside CloudFront. The revised design separates responsibilities by keeping CloudFront in front of the frontend while routing public API traffic through Amazon API Gateway, which connects privately to the EKS cluster via VPC Link. The key lesson was distinguishing between edge content distribution, which CloudFront handles well, and API governance, which requires a dedicated management layer.

0
ProgrammingDEV Community ·

Developer Builds AI Bridge to Run Agents Safely in Linux VMs on Apple Silicon

A macOS developer has designed an architecture called AI Bridge to address a core conflict in running AI coding agents on Apple Silicon: executing LLM-generated code safely in an isolated Linux VM while retaining full GPU acceleration. Most popular agent frameworks run unverified code directly on the host machine, exposing the filesystem, network, and credentials to potential misuse. Apple's Virtualization framework does not support GPU passthrough to Linux guests, making model inference inside a VM significantly slower than on the host. The AI Bridge solution splits the workload — the agent runs inside a disposable Linux VM for isolation, while the language model runs natively on the host via Ollama or MLX with Metal GPU access. The two components communicate through a vsock channel, allowing the in-VM agent to interact with the host-side model as if it were a local server.

0
ProgrammingDEV Community ·

Non-coder builds drawing-prompt web app using vibe coding after friend's suggestion

A self-described non-coder with only basic HTML experience from high school built a web app called Daily Doodle after a friend encouraged him to try vibe coding. The app is a drawing-prompt spinner featuring nine toggleable categories such as Animal, Occupation, and Vehicle, with separate Beginner and Advanced modes. Additional features include retro sound effects built using the Web Audio API, a Firebase-backed global spin counter, a repeat-avoidance option, and a recent history list. The project is hosted for free on GitHub Pages and consists of a single vanilla HTML file, keeping the technical stack minimal. The creator was inspired by a desire to help people overcome the blank-page paralysis of not knowing what to draw, echoing a philosophy attributed to Disney Imagineering leader Marty Sklar.