WAND-Vega on Sume: price ladder from Lite to Pro 1.5

Sume's three WAND-Vega rows bill from $0.072125 per second (Lite at 768p) to $1.3475 per second (Pro 1.5 at 4K), and each takes text-to-video requests of 4 to 15 seconds. They are Tencent-hosted rows that appear in a workspace only if GET /v1/videos/models returns their ids, so check the catalog first.

The numbers below come from Sume's pricing code, provider list price times 1.25. I read no vendor page for this post, so it makes no claim about WAND-Vega outside Sume.

The ladder

Each row has its own resolution set. Lite offers 768p, 1080p, 2K and 4K. Pro 1.0 and Pro 1.5 offer 720p, 1080p, 2K and 4K. All three take 16:9, 9:16, 21:9, 4:3, 1:1 and 3:4, with an audio toggle through generate_audio.

Prices below are billed per second. Multiply by the duration, which must be a whole number from 4 to 15.

A 5-second clip at each step

The table shows the unrounded arithmetic for a 5-second clip: the billed rate times 5. Sume's displayed charge may round differently, so read usage.cost on the finished job.

The spread is the point. A Lite 1080p clip costs about a fifth of a Pro 1.5 clip at the same resolution, and Pro 1.0 sits between them.

Where they sit against Sume's other rows

For a 5-second 720p clip, Wan 3.0 bills $0.625, which is cheaper than Pro 1.0 ($0.764375) and Pro 1.5 ($1.14263125) and more than Lite's 768p ($0.360625). Wan also takes image and reference inputs, while WAND-Vega rows are text-to-video only. If you need references, WAND-Vega is not the row.

If you simply want the lowest price for a text-only prompt, Lite is the cheapest row in this family and Mingmou and Hunyuan (a different post) are cheaper still at 720p.

Using them

Send model set to the row id, a prompt, an integer duration from 4 to 15, and a resolution from that row's list. Sume rejects size, seed and provider.options for every video model. Keep the Idempotency-Key header on the submit, as the Errors and rate limits page advises for retries, and compare usage.cost with the table on your first clip.

How to pick a rung

Start at the cheapest rung that matches the final resolution. Draft prompts on Lite at 768p, where a 5-second clip bills $0.360625, and only move to Pro 1.0 or Pro 1.5 once the prompt is settled. Because the rows share aspect ratios and the same 4 to 15 second range, the same request body works across all three; only model and resolution change.

Pro 1.5 at 4K for 15 seconds bills $20.2125, the top of the ladder. That is why it pays to draft low. A 15-second Lite draft at 768p bills $1.081875, so you can run about eighteen drafts for the price of one top-rung render.