A generator charges for attempts. A project needs usable results. If two images are generated but one is discarded, dividing the bill by two understates the cost of the picture that actually reaches the project.
The task is a small set of original fictional-adult portraits for an imaginary travel journal. There is no benchmark dataset here: all success rates below are explicit hypothetical examples, not observed FreeGF or OurDream results.
First write the acceptance rule
The portrait must show the same adult traveler, a blue coat and a visible map at a fictional station. The crop must retain the face and map. That is enough to make “usable” meaningful.
A nice-looking picture with the wrong coat may be unsuitable for continuity. A correct subject with an obscured map may fail the story. Count it as a rejected attempt for this task even if it would suit another project.
Keep the rule fixed while comparing costs. A provider should not appear more efficient merely because its outputs were judged against easier requirements.
Normalize inside each currency
FreeGF’s FAQ lists 4 credits per image. OurDream’s cost guide, edited March 2, 2026, lists 10 dreamcoins for two standard images and 10 for enhancing one image. The OurDream guide explicitly assumes a paid subscriber; a guest should not assume all those features are available.
The useful formula is:
Cost per retained image = all relevant generation and revision charges ÷ number of retained images.
Do not divide by every output returned. If nothing is retained, report the spending and zero usable results; there is no finite per-retained-image price to quote.
| Hypothetical result | Spending | Retained images | Effective unit cost |
|---|---|---|---|
| FreeGF: two attempts, both retained | 8 credits | 2 | 4 credits |
| FreeGF: two attempts, one retained | 8 credits | 1 | 8 credits |
| OurDream: one standard pair, both retained | 10 dreamcoins | 2 | 5 dreamcoins |
| OurDream: one standard pair, one retained | 10 dreamcoins | 1 | 10 dreamcoins |
| OurDream: that retained image also enhanced once | 20 dreamcoins | 1 | 20 dreamcoins |
The last row includes the original pair and the enhancement. Counting only the enhancement would hide the generation that made it possible.
These are within-provider calculations. Eight FreeGF credits and ten dreamcoins are not a cash comparison: acquiring the balances, any required subscription and the minimum purchase sizes all matter.
An enhancement has a different job from a new attempt
If the pose, identity and composition are already acceptable, an enhancement may be intended to improve the retained asset. If the scene contains the wrong subject or object, a new generation or an appropriate editing action may be needed instead.
OurDream lists enhancement and editing separately. Its guide gives different costs for named editing models. This article does not assume enhancement repairs an incorrect identity or that every tool preserves the source equally well.
FreeGF’s generator guide supplies a basic image rate, but the sources reviewed do not establish one universal repair price for all of its editing routes. Do not fill that gap with the 4-credit base-image figure.
Do not buy the first two images twice in your estimate
The OurDream subscriber guide lists a new chatbot at 10 dreamcoins with its first two images included. If those two initial images meet the brief, they belong to the setup action; adding another 10-coin standard pair to account for the same two results would double-count them. If neither is usable, subsequent image requests become additional attempts.
Its creator FAQ separately advertises a free-trial allocation and up to ten characters. That does not publish a current five-coin subscriber creation rate or establish that all paid media actions are available to a guest. Keep the actual entitlement attached.
The yield table assumes an existing suitable character. For a new one, start a separate setup row, count the included outputs once and add only later chargeable requests. That is the difference between an attempt budget and a misleading list of nominal feature prices.
Use a stop condition
For a project needing two portraits, stop when two meet the brief. Keep a small record of action, charge, output count and retained count. If three batches fail on the same requirement, revising that requirement’s instruction may be more useful than buying a fourth identical attempt.
FreeGF’s simple listed image rate is easy to use for a bounded attempt budget. OurDream’s subscriber documentation makes the pair size and several downstream revision charges explicit. Neither source proves the better retention rate.
The fair decision combines the actual account quote with the yield you observe on your own clearly defined task. This guide has not generated or purchased images on either platform.