Free-trial allowance per signup for an AI video app: Sume costs
What a free trial pack costs per signup and per 1,000 signups on Sume: three packs from $0.14 to $2.76, and cost per paying user at 2-10% conversion.

If your product gives each new signup a free taste of AI media and calls Sume underneath, three trial packs cost you $0.14, $0.56 and $2.76 per signup. At 1,000 signups that is $137, $556 and $2,760. At a 5% paid conversion, the cost of acquiring one paying user through the trial alone is $2.74, $11.11 and $55.20.
The pack is your design choice; the unit prices are Sume's. This post prices the packs so the allowance is a number you picked, not one that drifted.
What goes in a trial pack?
Pick the smallest deliverable that proves the product to the user. For a still-image product that is a few low-quality images; for a clip product it is one short clip; for a presenter product it is one short avatar take. Prices below come from the Sume pricing package: GPT Image 2.5 at low quality is $0.02475 and at medium $0.055625 at its default request; a five-second Grok Imagine clip is $0.0625 (Sume's Grok row needs one input image as the start frame, which is why the Starter pack includes images); a five-second H3 Max clip at 768p is $0.50; and the Avatar Standard tier is $0.184 per second without a product image.
| Pack | Cost per signup | Cost per 1,000 signups |
|---|---|---|
| Starter: 3 low-quality images + one 5 s Grok clip (one image is its start frame) | $0.1368 | $136.75 |
| Standard: 1 medium image + one 5 s H3 Max clip | $0.5556 | $555.63 |
| Premium: one 15 s Standard avatar clip (avatar already made) | $2.7600 | $2,760.00 |
What is the cost per paying user?
Divide the pack cost by the share of trial users who pay. The result is the trial's contribution to customer acquisition cost, before any ad spend. It assumes every signup uses the whole pack, which overstates the cost for a real funnel where many users do not.
| Pack | At 2% convert | At 5% convert | At 10% convert |
|---|---|---|---|
| Starter | $6.84 | $2.74 | $1.37 |
| Standard | $27.78 | $11.11 | $5.56 |
| Premium | $138.00 | $55.20 | $27.60 |
How do you stop a trial from running away?
A free allowance is a liability you control only if the request path enforces it. Sume reserves the estimate at submit and captures it on success, per Generation admission, so your own counter must track what you gave away.
- Count allowance in your database per user, and decrement before you call Sume, not after.
- Make the pack a fixed recipe (model, duration, resolution) so the price is known in advance and the user cannot pick the 1080p version.
- Use an idempotency key derived from user id and pack so a double click does not spend the allowance twice.
- Cap the daily total across all trial users in your own code, and fund a separate wallet for it.
- Pick the cheapest model that makes the trial convincing, then test whether a more expensive pack actually raises conversion.
Which pack should you pick?
Start with the cheapest pack that shows the output your users would pay for, and only move up if conversion data justifies it. A pack that costs five times more (Premium against Standard) has to convert about five times better to break even on cost per paying user. For presenter products, remember that an avatar is created once at $0.95 and reused, so the creation fee is a per-user setup cost if every signup needs their own avatar, and a shared one-time cost if you ship a stock presenter.
Sources
Related posts
More in Pricing
- Freelance editor's 3 revision rounds on a 30-second ad: $0.96-$1.26
A 30-second client ad costs $0.36 to build on Sume and $0.20 to $0.30 per revision round, so three rounds end at $0.96 to $1.26 depending on what changes.
- gpt-4o-mini-tts at $12 per 1M characters vs a Sume TTS job
OpenAI lists gpt-4o-mini-tts at $12 per 1M characters, about $0.012 per 1k. Sume TTS bills about $0.0475 per 1k characters, so OpenAI is cheaper on price alone.
- GPT-6.1 Sol Flex and Batch at half price: is there a Sume equivalent?
OpenAI prices GPT-6.1 Sol Batch and Flex at half of Standard and Fast at 2x. A Sume Format run exposes only a model field; the picker has a Fast box.
- GPT-6.1 Sol past 272K input tokens: the price doubles for the request
GPT-6.1 Sol charges 2x input and 1.5x output once a prompt passes 272K tokens. What that means for a long Format thread, with the vendor tables read Oct 3.
Written by Sume