# AXMORF Studio Taking agent-led creation through to verifiable local video delivery. A local video production workspace built on Remotion. A coding agent organizes scenes, narration and covers, while a defined production process and validation deliver a video, two covers and a publishing manifest. Valid artifacts can be reused, and revisions are developed in isolation. [Website source](https://zzzxc.com/projects/axmorf-studio) ## Starting with a video brief Making a video means bringing a topic, narrative, scenes, narration, subtitles, sound and covers into one work. An agent can contribute to each part, but aligning them in timing, visuals and delivered files still requires a clear production process. AXMORF Studio builds that process around a local workspace owned by the user. The user gives a creative brief to their coding agent; the agent follows the project rules to organize content, write visual scenes and run production. Source code, media, configuration, intermediate artifacts and final files remain in the workspace, where they can be inspected and revised. ## Giving creative tasks explicit inputs and results Each story corresponds to a Remotion Composition, with its content divided into individual scenes. Scenes, the overall visual direction and covers have distinct responsibilities. Narration and subtitles follow a shared timing structure, keeping multiple creative tasks from taking control of the same content layer. Before production, the system checks whether inputs are complete, which artifacts can be reused, which tasks must run again and what is blocked. Preparation that calls media services has its own entry point. The agent then creates within the assigned task scope, while fixed programs handle validation, assembly and rendering. The workflow is: creative brief → project definition → production checks → media preparation → scene and cover creation → validation and rendering → local delivery. ## Letting sound establish the timing The actual duration of narration affects subtitles, scene transitions and visual rhythm. AXMORF derives timing from the sample count of the generated audio, then maps it to video frames. Validated audio is fixed so that later production uses the same sound and timing information. Scenes and subtitles can therefore follow the narration that will actually be heard. Narration, background music and sound effects also have distinct roles: narration carries the story, background music covers the prescribed content interval, and scene sound effects contribute as separate audio. ## Keeping work that is already complete Creation usually involves repeated adjustments. AXMORF identifies reusable results from task inputs and artifact contents. When inputs are unchanged and files still pass validation, those results can be retained. Editing one scene propagates reruns through its actual dependencies, preserving other scenes, covers or narration that remain valid. When revising an existing work, the new version is produced in an isolated candidate area first. It replaces the current project and delivery only after passing checks. If candidate production fails, the previous work remains available. This gives experimentation a clear boundary and reduces the risk of damaging an existing result halfway through a revision. ## Completion means four inspectable files A delivery contains video.mp4, a landscape cover-4x3.png, a portrait cover-3x4.png and publish.json. The publishing manifest holds publication copy and delivery information; the user handles the actual upload separately. Delivery checks cover the file set, checksums, media formats, dimensions, frame rate, frame count and whether the media decodes completely. The workspace also provides a local web control center for configuration, diagnostics, progress and the current delivery, alongside Remotion Studio for visual previews. ## The engineering focus AXMORF leaves open-ended creative judgment to the agent and assigns input rules, task boundaries, result validation and delivery replacement to programs. The agent decides how a passage should be expressed; the system checks whether it produced a result that can become part of the work. This division makes failures easier to locate. Environment preparation, media services, individual scenes, rendering and final delivery each have their own status and checks. Valid results can also survive into later runs, giving the creative process a traceable foundation for revision. ## Current stage and scope The project is in public beta, with an npm workspace distribution and recorded production runs using Codex and Hermes. It requires a working Node.js environment, a rendering browser and configured media services. The workspace is local, while selected external services may still involve network requests and costs. Automated checks establish file validity and some production constraints. Visual quality, storytelling and a complete listening review still require human judgment. The project does not currently manage platform accounts, upload videos or schedule cloud production. Verification scope across hosts and execution modes is documented in the acceptance records for each version. ## From brief to local delivery ### Define the project Turn the creative brief into a story and scene structure. ### Check production Identify missing inputs, reusable artifacts, reruns and blockers. ### Prepare media Prepare required media and establish timing from validated narration. ### Create scenes & covers The agent works within explicit scene, visual and cover responsibilities. ### Validate & render Assemble the work and check its files, media properties and decoding. ### Deliver locally Keep a video, two covers and a publishing manifest in the workspace. ## Explore the project - [Source & project documentation](https://github.com/AXMORF/axmorf-studio) [Contents](https://zzzxc.com/agent/en/index.md)