VEED Fabric 1.0 price per second: fal list vs Sume rate
fal lists VEED Fabric 1.0 at $0.15 per second at 720p. Sume bills veed/fabric-1.0 at $0.1875 at 720p and $0.10 at 480p, rounded up per audio second.
On fal, VEED Fabric 1.0 is listed at $0.15 per second for 720p and $0.08 per second for 480p. On Sume, the same model is called veed/fabric-1.0 and is billed at $0.1875 per audio second at 720p and $0.10 at 480p, with the duration rounded up to whole seconds. The Sume price includes Sume's margin and the job handling around it; the 720p Sume figure is 25 percent above fal's list.
Fabric takes one still image plus one audio file and returns a talking video, so the bill follows the length of your audio, not a number of generations.
The two price lists side by side
Both columns are per second of output.
| Resolution | fal list | Sume | 60 s audio on fal | 60 s audio on Sume |
|---|---|---|---|---|
| 480p | $0.08 / s | $0.10 / s | $4.80 | $6.00 |
| 720p | $0.15 / s | $0.1875 / s | $9.00 | $11.25 |
What you get for the difference
Calling fal directly means your own account, key and queue handling. Calling Sume means one API key and one job envelope shared with every other Sume model: status_url, result_url, an optional signed webhook, an Idempotency-Key so a retry does not bill twice, and a durable media.sume.com result URL instead of a provider link. If you only ever run Fabric, fal is cheaper. If Fabric is one step in a pipeline of voice, captions and timeline, the single contract is usually worth more than the margin.
The request
The canonical route is POST /v1/veed/fabric-1.0. Send audio_url and a measured duration_seconds, plus exactly one visual source: an image_url of a still you have inspected, or an avatar_handle when the person named a specific avatar. The older POST /v1/avatar-1.0/image-to-video alias accepts the same body.
| Limit | Value |
|---|---|
| Audio length | Up to 300 seconds |
| Audio source | Sume-hosted, under 10 MB |
| Billing unit | ceil(duration_seconds) times the resolution rate |
| Maximum charge | $56.25 at 300 s and 720p |
When Fabric is the wrong choice
Fabric animates a talking face to supplied audio. If you want Sume to write the performance from a script, use the Avatar 1.0 talking-video route, billed at $0.184 to $0.55 per second by tier. If your audio is longer than 300 seconds, split it first. Prices on both sides change; the fal page was read on the date in the table caption. Before you commit to either, run one real audio file of the length you expect through both and compare the invoice line with the table above, because duration is rounded up per clip and many short clips round up more than a few long ones.
Sources
Related posts
More in Pricing
- Veo 3.1 price per second: Google's API vs fal, audio on and off
Veo 3.1 Standard is $0.40 a second at 720p and 1080p on Google's API; fal splits audio on and off. A dated table, 8-second clip costs, and the Omni alternative.
- Veo 3.1 audio block: not charged. What happens on a failed Sume job?
Google says a Veo 3.1 video blocked by an audio error is not charged. Sume reserves cost at submit and releases or refunds the hold when a job fails.
- Video inspect transcript: the one-minute reserve and duration_seconds
transcribe true in video-inspect costs $0.01 per audio minute. Omit duration_seconds and Sume reserves one minute; the hint maxes at 600 seconds.
- Video-to-video AI cost per clip: Recast and Genjutsu
Video-to-video on Sume costs $0.375 to $0.85 per second: a 10-second Recast clip at 768p is $3.75 and a 10-second Genjutsu clip at 720p is $8.51.
Written by Sume