90 hook drafts at 360p, then 6 finals: a 30-product hook test
Thirty products times three hooks is 90 Omni drafts at 360p for $10.125; six winners at 1080p add $9.00. The whole test costs $19.125 on Sume.

The cheapest way to test hooks on Sume is to draft every hook at 360p on Gemini Omni Flash 1.1 and re-render only the winners. For 30 products with 3 hooks each, 90 drafts of 3 seconds cost $10.125, and 6 finals of 8 seconds at 1080p cost $9.00, so the whole test is $19.125.
Omni Flash 1.1 clips run 3 to 10 seconds at 360p, 720p, 1080p or 4K, with native audio always on. fal lists the model at $0.03, $0.10, $0.15 and $0.30 per output second for those resolutions (read 2026-10-07); Sume bills list times 1.25.
The arithmetic
A 3-second clip is the shortest the model accepts, so it is the floor for a draft. Resolution is the only variable that changes the rate.
| Step | Clips | Seconds each | Rate per second | Subtotal |
|---|---|---|---|---|
| Drafts, 360p | 90 | 3 | $0.0375 | $10.125 |
| Finals, 1080p | 6 | 8 | $0.1875 | $9.00 |
| Total | $19.125 |
What a 360p draft can and cannot tell you
A draft is good for the first three seconds: the opening framing, the first motion and the spoken line. It is poor for fine detail such as label text or skin texture, so judge hooks on pacing and clarity, not polish.
Re-rendering at 1080p is a new generation from the same prompt, not an enlargement, so the result can differ from the draft. Budget one extra final per winner if the exact draft look matters.
Finishing all 90 hooks at 1080p and 8 seconds would cost $135.00. The staged plan costs $19.125, a saving of $115.875.
- Hook set: 3 hooks per product, 30 products.
- Selection: pick one winner per five products, 6 total.
- Spend: each draft shows as reserved, then captured on success, in the usage ledger.
Scaling the test up or down
Each extra product with three 3-second drafts at 360p adds $0.3375. Going from 30 to 60 products doubles the draft line to $20.25 and leaves the finals line to you: at one final per five products that is 12 finals, or $18.00.
The ratio that matters is drafts to finals. At 15 drafts per final, the draft spend is $1.6875 for every $1.50 final, so drafts are about 112 percent of the cost of a final. If you pick one final per ten drafts the draft share doubles, and if you pick one per twenty it halves.
Keep a note of which hook won in each batch. A winning hook line from product A is often worth testing on product B at 360p before you commit finals, and that costs only $0.1125 per draft.
Limits of the cheap pass
Omni Flash 1.1 has native audio that is always on, so a draft carries a soundtrack you did not pick. Judge the hook on the picture and the spoken line, not on the music. The model accepts 360p, 720p, 1080p and 4K, 16:9 or 9:16, so pick the aspect ratio in the draft that you plan to ship; a 9:16 draft says nothing about a 16:9 final.
Wallet and concurrency
Ninety jobs exceed the accepted capacity on smaller plans. Free accepts 6 jobs at a time, Pro 24, Startup 48 and Scale 120, so submit in waves and read each job's status. Queued jobs hold their reserve, so 90 drafts at 360p hold up to $10.125 while they wait. How far $20 goes scales the same rates to a smaller wallet.
Sources
Related posts
More in Use cases
- One 58-second vertical master that fits Shorts, Reels, and TikTok ads
A 58-second 1080x1920 master fits YouTube Shorts, the Reels 3-minute line, TikTok reservation and TopView, and Google's Shorts-ad guide. Render it and verify.
- One ad, five cities: swap the backdrop per market with Omni edit
Localize a street-scene ad by editing the backdrop per market with Gemini Omni Flash. Prompt pattern, five 8-second edits at 720p for $5.00, and what to verify.
- One AI image a day for a year: 365 images cost $1.825 to $96.27
A daily image for 365 days costs $1.825 on Soul at 720p, $9.13 on Grok, $13.69 on Flux 2 Pro and $96.27 on ChatGPT Image 2. Table and a scheduling note.
- One AI video for TikTok, LinkedIn, Reels and Threads: limits table
Length, size and format limits from TikTok, LinkedIn, Meta and Threads pages in one table, and how to produce one master AI video that fits them with Sume.
Written by Sume