Repository navigation
TaleMondo Studio: a film-production workspace around local MiniMax H3 and ComfyUI #17629
Klimentios
started this conversation in
Show and tell
Replies: 1 comment
|
I'd want the approved opening frame linked to the final take. That way I could see what was signed off before the shot went into the GPU queue. |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
I built TaleMondo Studio to keep an AI film project together beyond a single generation graph. The documented rendering path sends MiniMax H3 jobs to ComfyUI on your own NVIDIA GPU. Studio also lets you import a custom ComfyUI graph as a reusable form, local REST endpoint, and MCP tool, then reopen that workflow in ComfyUI for graph editing.
The production side connects role-tagged character and environment references, franchise and scene boards, shot direction, a GPU queue, take review, and an edit timeline. A coding agent can inspect project state, prepare shots, and monitor a queued job through the MCP tools; the creator still approves GPU work and selects the final takes.
What is documented today: local H3 through ComfyUI; a Mac interface; and an H3 server setup that calls for 96 GB VRAM and 256 GB system RAM. Imported workflows for other models depend on their graphs and installed nodes and have not been verified for this release. The code also contains a separate MiniMax API route. The short tour is edited and does not show a live GPU render.
Source and setup · Product walkthrough and seven video examples
MiniMax AI created H3; Comfy-Org provides the native H3 nodes in ComfyUI; Kijai maintains tools used in optional paths. TaleMondo is my independent, noncommercial application.
I would especially value feedback from people managing multiple H3 shots: which project decisions or render receipts are hardest to keep traceable between ComfyUI and the final edit?
All reactions