video-engine-routing · diff
v1.0.0 to v1.1.0
10 added, 6 removed. Audit A to A.
What this update changes in the checks
No check changed: every fixed check gives the same answer before and after this update.
A tripped check is a label, not a verdict: read the matched text and the lines below. This week's updates that newly trip a check.
---
name: video-engine-routing
description: Decide WHERE video/image generation runs when cost, volume, or capability forces a choice - subscription credits vs pay-per-use APIs vs cloud ComfyUI. Use when asked "which video engine should I use", "is Higgsfield/fal/RunComfy worth it", "credits ran out", "cheapest way to generate N clips", "do I need ComfyUI", "should I run this locally", or when the video-gen skill escalates a budget or capability ceiling. Contains the pricing table, graduation triggers, and the never-local rule for iGPU machines.
- version: 1.0.0
+ version: 1.1.0
---
# video-engine-routing — when you outgrow your default engine
The default stack (see `video-gen`) assumes a subscription engine. This skill is the
escalation path: what to do when credits run out, volume grows, or the pipeline needs
control the subscription engine cannot give.
## The decision tree
1. **Under ~50 clips/month, mixed styles, no custom pipeline** → stay on the subscription
- engine (Higgsfield or equivalent). Sunk cost, routing across many models, zero ops.
+ engine you already pay for (an all-in-one such as Higgsfield, or an equivalent). Sunk cost,
+ routing across many models, zero ops.
2. **Bursty overflow or budget clips at volume** → pay-per-use API (fal.ai). No subscription,
no credit expiry, cheapest per-clip for budget models.
3. **Custom pipeline: LoRA character consistency at batch scale, ControlNet/IPAdapter
chains, motion transfer, community models** → cloud ComfyUI (RunComfy machines or
Comfy Cloud). You are paying for control, not per-clip price.
4. **Local GPU rendering** → only with a recent NVIDIA card with 16 GB+ VRAM. On an iGPU
machine (Intel Arc/AMD APU, shared memory): never. Modern video models blow through
shared-memory budgets, and quantized variants degrade video coherence. Cloud is
faster and cheaper than wasted local hours.
## Pricing reference (mid-2026, one 5-second 1080p clip)
+ These are the author's working estimates, not vendor quotes, and none is sourced here. Open each
+ vendor's current pricing page before you decide.
+
| Route | Cost per 5s clip | Notes |
|---|---|---|
- | Higgsfield subscription | ~$0.30–0.50 in credits | tiers $15–129/mo; credits don't roll over |
+ | All-in-one subscription (Higgsfield) | ~$0.30–0.50 in credits | tiers $15–129/mo; credits don't roll over |
| fal.ai — Seedance 2.0 Fast | ~$0.11 | cheapest respectable social clip |
| fal.ai — Kling | ~$0.35 | pay-per-second, no minimums |
| fal.ai — Veo 3.1 Fast | ~$0.50 (720p) | 4K tier ~3× |
| RunComfy machines | $0.99–9.59/hr + Pro $20/mo | custom ComfyUI workflows; cold-start overhead makes single clips uneconomic |
| Comfy Cloud | $20/mo ≈ 180 GPU-min | official ComfyUI cloud, credit-based |
| ComfyDeploy | — | defunct; stack was open-sourced, managed service gone |
Recompute before deciding: prices move quarterly. The *structure* (subscription vs
per-use vs machine-rental) is what stays stable.
## Graduation triggers
- **To fal.ai**: monthly credit top-ups exceed ~$20 two months running, or you need one
premium model your subscription lacks.
- **To cloud ComfyUI**: you need the same character across 10+ videos via LoRA, or a
ControlNet/IPAdapter/motion-control chain no API engine exposes, or 100+ clips/month of
one repeatable workflow where machine-hours beat per-clip pricing.
- **Back down**: if a rented workflow runs less than monthly, cancel the machine and take
the per-use price. Idle infrastructure is the most expensive engine.
## What not to do
- Don't install vendor API-wrapper skill packs for engines you don't pay for — every one
of them is lock-in to that vendor's endpoint with no portable workflow definition.
- Don't run modern video models on integrated GPUs (see rule 4 above).
- Don't hold two overlapping subscriptions to hedge; pick one default and use per-use
APIs as the hedge.
## Credits
The single-dispatcher/model-leaf routing pattern used by `video-gen` is adapted from the
router design in agentspace-so's RunComfy skill pack — pattern absorbed, dependency skipped.
- Disclosure: I am a Higgsfield Ambassador; Higgsfield links on my site may be affiliate links.
- This skill's routing logic does not depend on that relationship — the decision tree sends
- you to fal.ai or ComfyUI whenever those genuinely win.
+ Disclosure: the author is a Higgsfield Ambassador, and Higgsfield links on frankx.ai may be
+ affiliate links. The routing logic here does not depend on that relationship: the decision tree
+ sends you to fal.ai or ComfyUI whenever those genuinely win.