AI video at near cost. The receipt is the pitch.
Figs Healed renders good-enough AI video in bulk on open models and rented GPUs, and prices it from the measured cost up — not from what the market will bear. One standard clip — 5 seconds of video at 1280×720, 24 frames per second — costs one credit. Credits never expire while the service operates.
For faceless channels, TikTok Shop sellers, local businesses that churn through ads, and indie apps that need video through an API.
Private build. There is nothing to buy yet — the list opens first when the first public batch does.
The receipt
Every figure on this page is labeled — measured, from our own tests; typical, for the comparison; or a pack price. The measured figures come from our own test batches, run in October 2026 on a rented RTX 3090.
What a clip costs us — measured
Measured, October 2026, evidence batch: 10 standard clips (5.04 seconds, 121 frames at 1280×720, 24 fps) and one 1024×1024 still, rendered on a rented RTX 3090 at a live rate of $0.256/hr. Total batch cost $0.5356; divided by the 10 successful clips, the fully loaded cost is $0.054 per clip — provisioning (about 33 minutes, including a software repair), every render, one re-roll, and verification time included. The full batch report is published with the samples. An earlier smoke test measured ≈ $0.009 of instance time for a 2.04-second clip. The pack prices below sit above the fully loaded figure; the gap is overhead and margin.
What a keeper costs elsewhere — typical
Closed frontier models produce output ours does not match. This comparison is about price per usable clip for bulk content, product clips, and social video — the work Figs Healed is built for.
What a clip costs here
One credit per standard clip (5 seconds, 1280×720, 24 fps), from a pack whose credits never expire while the service operates.
How it works
1. Pick a look, or write a prompt
Bring your own prompt — a camera preset library is in build for the first public batch. Choose video or stills, and how many you need.
2. Your job enters the queue
Jobs run in order on a shared, capped pool; seeing your place in the queue is planned for the member view. Turnaround is batch-paced, not instant — that is the trade the price reflects.
3. A rented GPU renders it
We rent GPUs by the hour for each batch and shut them down when the batch ends. Open models do the rendering: Wan for video, FLUX for stills.
4. Download your clips
Finished clips and stills land in your account, ready to download. Usage terms publish with the first public batch.
Models
Open weights only.
Wan — video
The bulk clip workhorse: product shots, B-roll, faceless-channel footage.
FLUX — stills
Thumbnails, product images, and frames at about 2¢ each.
Plainly stated: these are not the closed frontier models, and the output is not their output. For bulk content, product clips, and social video, good enough is the standard we price to — and the price is the difference.
Pricing — credit packs
Credits never expire while the service operates. Buy a pack in a busy month; spend it in a slow one.
Starter
Creator
Studio
One credit = one standard clip (5 seconds, 1280×720, 24 fps). One still = 0.2222 credits — about 2¢. No subscription, no monthly minimum.
Packs are not on sale yet. Figs Healed is in private build, and early-access members are first in when the first public batch opens.
FAQ
Do credits expire?
No. Credits never expire while the service operates. Buy when you need them; use them when you make something.
What exactly is one credit?
One credit renders one standard clip with Wan: 5 seconds at 1280×720, 24 frames per second. A FLUX still is priced in credits too — see Pricing for the conversion.
Who is this for?
People and products that make video in volume and count the cost per clip: faceless channels, TikTok Shop sellers, local advertisers who refresh creative often, and indie apps that generate video through an API.
Is there an API?
In build — an API on the same queue and the same per-clip pricing, for builders. API access opens with the first public batch; join the list and tell us you build.
Why is it this cheap?
Open models carry no platform's per-generation fee, and rented GPUs are paid by the hour only while a batch runs. The measured fully loaded figure — $0.054 for a standard 5-second clip, from our October 2026 evidence batch — is in the receipt above; pack prices sit above measured cost, and the gap is overhead and margin, in the open, in the price.
When can I buy a pack?
When the first public batch opens. Join the early-access list and you will hear before it does. Nothing is charged before then.
Join the early-access list
Private build — the list opens with our first public batch. Interest only: no account, no charge, nothing to buy yet.