#OpenSourceHiggsfield runs to Nov 3, winners Nov 10: four LTX builds to know

What changed. On Oct 30 at 6:10 PM MDT, @ComfyUI said it was extending the challenge to Nov 3 at 9 AM PT, pointing to the response from builders ("If we win, we all win"). A follow-up said entries already in can be resubmitted until then. The timeline on the challenge page now reads: build window Oct 15 to Nov 3, Build Nights in SF (Oct 28), LA (Oct 29) and New York (Nov 2), and winners on Nov 10. As of Monday morning (Nov 2, MST) we couldn't find a public reply from Higgsfield to the cease and desist that Comfy made public on Oct 29.

Seam, by @jdnichollsc. Give it two shots (images, 1-6 s clips or a .glb 3D model you orbit in the browser), choose an effect, and LTX-2.5 renders the transition between them on your own GPU through ComfyUI. Per the README, ten effects passed the author's own quality bar (smoke, water bend and whip pan among them). It can chain a whole sequence of hand-offs and re-render only the one that failed. Every take saves as a ComfyUI workflow you can reopen. It ships with a CLI and a Claude Code plugin (a skill plus an MCP server) so an agent can plan, render and review transitions. MIT.

Chimera, by @_aCalle. You give it a 3D model and a short action prompt. LTX-2.5 makes a motion video from a render of the model, then Chimera tracks points through that video and fits the model's own skeleton to it. The result exports as GLB, FBX or BVH, with root-motion and in-place clips. Models without a rig can be auto-rigged with SkinTokens. The author built it on an RTX 5080 with 16 GB, and the README says short moves with a clear silhouette work best. MIT.

VFX 4 All, by @nomaaistudio. It has 45 effect presets, such as walking through fire, turning to liquid chrome or a clone army. Qwen Image Edit 2511 builds a film scene around your photo, then LTX-2.5 animates the effect with sound. It can also pick up a clip you filmed at a chosen frame and splice the effect in. You can run it on your own ComfyUI with a Comfy API key, or try presets as Comfy Cloud apps billed to your own account. Open weights only, MIT.

Earthfall, by @jamie_jk. Pick a landmark and add a portrait, and it makes an 8-10 s dive from low orbit through the clouds to that person at the location, an open take on Higgsfield's Earth Zoom. Per the README, the flight is generated in reverse and played forwards. It uses FLUX 3 Klein for stills, LTX 2.5 for video and Qwen3.5 4B as an automatic checker, and needs an NVIDIA GPU with 24 GB (or a remote ComfyUI). The README notes that everything beyond the landmark itself is invented. GPL-3.0.

Our take: there's one more day to enter, and the judging date has moved, so if you're waiting on results, plan for Nov 10. Of the four, Seam's agent plugin and Chimera's rig-fitting are the most useful outside the contest. Turning generated motion into editable animation on your own skeleton is a real gap in open tools today.