# Tired of paying Higgsfield to get my SFW prompts censored, so I rebuilt it on ComfyUI: one-click reels, 25 camera moves, re-shoot from a new angle. All workflows included. — retained record

> Retained reddit record from reddit:comfyui.

Canonical: https://highsignal.app/data/records/a55c13a9bfec0624

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:comfyui)  published 2026-10-23 11:45 ingested 2026-10-24 02:40

[original source ↗](https://www.reddit.com/r/comfyui/comments/1xdbrfi/tired_of_paying_higgsfield_to_get_my_sfw_prompts/)

## retained text



```
Credits gone, "this content violates our guidelines" on a perfectly SFW prompt, try again, more credits gone. I was done.  
  
  
So I built **CreatorOS Studio**: the one-click experience of the hosted video apps, but on **your own GPU,** with **open-weights models only.** No credits, no subscription, no one else's filter on your machine. It's Apache-2.0.  
  
  
What you get, one click each:  
  
  
\- 🎬 **Brief → finished reel.** Write one line. A local LLM plans the shots, each one gets animated **with sound** (LTX-2.5) and the reel is edited for you, with hook, fades and music.  
\- 🎥 **25 camera moves & effects** on any photo: dolly, crash zoom, orbit, bullet time, dolly zoom… plus squish, inflate, cakeify, explode.  
\- 🔁 **Re-shoot**: take a clip and get the same moment **from another camera angle**, original sound kept.  
\- 🗣️ **Talking avatar**: one photo + a voice, lips in sync.  
\- 🧍 **Mix:** drop in any video and your character takes the person's place.  
\- 🖌️ **Brush retouch** on photos *and* videos, a **Cinematic shot card** (lens, aperture, film stock, grade instead of prompt soup), a **timeline editor**, and a **mobile U\***.  
  
  
Under the hood it's plain ComfyUI, and every recipe ships in the repo as a workflow you can open: Krea 2 / FLUX.2 Klein, Wan 2.1/2.2, LTX-2.5, InfiniteTalk, SAM3, MoGe-2, plus Ollama for prompts.  
  
  
For this sub: the app is a front end over **production ComfyUI recipes** (API-format JSON in \`workflows/\`). At job time it patches the prompt, seed, LoRAs and masks into the recipe and queues it. The re-shoot chain is MoGe-2 → CrossViewWarp → LTX-2.5 + the CrossView-Warp IC-LoRA. The jib down is two renders chained: a jib up, then a Wan 2.2 first-last-frame move back down. Happy to go deep on any graph.  
  
  
Honest catch: it wants a beefy NVIDIA card (built on 32 GB).  
  
  
\- ⭐ Repo: [https://github.com/thePloggy/CreatorOS-Studio](https://github.com/thePloggy/CreatorOS-Studio)  
\- ▶️ 2-min demo: [https://youtu.be/axHdAXHv3Kk](https://youtu.be/axHdAXHv3Kk)  
\- 🧪 Try it live on my GPU (SFW, queued, 5 renders/h): [https://forge.taila06702.ts.net:8443](https://forge.taila06702.ts.net:8443)  
  
  
Built for the Comfy #OpenSourceHiggsfield challenge. Tell me what breaks and I'll fix it.
```



document fetched 2026-10-24 10:42 · canonical https://reddit.com/r/comfyui/comments/1xdbrfi/tired_of_paying_higgsfield_to_get_my_sfw_prompts
