AI short film workflow with an agent
AI video production with an agent splits the work. CoreSpeed keeps the production's decisions in shared memory, so every session starts from the same cast, rooms and rules, and runs the generation jobs for performances, dialogue, lip sync and music. Rehearsal, screen recordings, compositing and the final edit happen in your own tools.
The job. Run a short-film production where the agent keeps every decision in memory and runs the generation jobs, with a person reviewing before each hand-off.
Paste this into Claude Code, Codex or Cursor
set up https://corespeed.io/SKILL.md (read it raw, e.g. with curl), then save our film's cast, locations and style rules to shared memory, and generate one five-second test shot of scene 1 with a video model that costs under 300 credits.One prompt.Three steps.
Set up first. Nothing on CoreSpeed. Rehearsal (the film in the post used Blender), screen recordings, compositing, the mix and the final edit happen in your own tools; CoreSpeed does the memory and the generation jobs.
Write the rules down once
Cast, rooms, wardrobe and style go to shared memory. Every session reads them before it plans.
Pilot the hard shots
Generate one test shot and review it at normal speed before you run the batch.
Finish in your editor
Dialogue, lip sync and music come back as jobs. Compositing, the mix and the cut happen in your own tools.
The tools it calls
memory__remember· Save the cast, rooms and style rulesmemory__search· Start each session from themmedia__list_models· Find a video model and its ratemedia__get_model· Read its inputs and price before runningmedia__generate· Run the test shotmedia__get_result· Collect the job, from any session
Every tool and its price: Video generation, Audio generation, Memory