Test log
Mage test data
Raw results for the Mage review. The verdict is in the review.
- Tested
- 6 November 2026
- Tester
- Miles Foster
- Account
- Free demo account, signed up by RankDude, nothing paid to the brand
- Tests
- 10
- Screenshots
- 5
- Method
- Published methodology
- Environment
- Google Chrome on desktop
Results
| ID | Test | Key measurement | Score |
|---|---|---|---|
| T01 | Style Range | Styles run: 0 of 10 | n/a |
| T02 | Character & Subject Consistency | Images run: 0 of 12 | n/a |
| T03 | Editing, Inpainting & Image-to-Image | Edits run: 0 of 12 | n/a |
| T04 | Resolution & Upscaling | Native size: 896 x 1,152 px (1.03 MP) | n/a |
| T05 | Prompt Adherence | Elements correct: 55 of 56 | 5/5 |
| T06 | Photorealism & Anatomy | Images that pass: 4 of 5 | 4/5 |
| T07 | Text in Images | Words correct: 24 of 24 | 5/5 |
| T08 | Speed per Image | Median per image: 17.3 s | 4/5 |
| T09 | Usage Rights, Watermarks & Privacy | Points met: 6 of 8 checked | 3/5 |
| T10 | Pricing & Value | Free images I made: 9, with 57 Gems left | 4/5 |
| Overall (in the review) | 4.4/5 | ||
Style Range
n/aNot Tested- Method
- Ten style prompts for one lighthouse.
- Styles run
- 0 of 10
- Remark
- Not graded: no style prompt was run.
Raw notes
Ten style prompts for one lighthouse.
Not run: ten images cost 270 Gems and 57 were left after the other tests.
- Run on the next review of this tool.
Character & Subject Consistency
n/aNot Tested- Method
- A described woman and a described thermos in six scenes each.
- Images run
- 0 of 12
- Remark
- Not graded: no consistency prompt was run.
Raw notes
A described woman and a described thermos in six scenes each.
Not run: twelve images cost 324 Gems, more than the whole 300 Gem allowance.
- Run on a plan with Gems to spare.
Editing, Inpainting & Image-to-Image
n/aNot Tested- Method
- Twelve fixed edits.
- Edits run
- 0 of 12
- Remark
- Not graded: no edit was run.
Raw notes
Twelve fixed edits.
Not run: the plans page lists Inpainting and Refine under the Basic plan, and 57 Gems were left, below the 85 percent stop.
- Run on a plan with Gems to spare.
Resolution & Upscaling
n/aNot Tested- Method
- Every saved image was measured in pixels.
- Native size
- 896 x 1,152 px (1.03 MP)
- Remark
- Not graded: the scale needs an upscale compared with its original, and none was run.
Raw notes
Every saved image was measured in pixels. The upscaler was not run.
All 9 images came out at 896 by 1,152 pixels (1.03 megapixels) as JPEG, in a 4:5 frame, at the 1K setting the app showed.
The Image Enhancer app offers 2K, 4K and 6K. The plans page lists Enhance (2K) under Pro and Enhance (4K) under Pro Plus, so I did not run it on the free plan.
- No upscale, so no grade. Run on a paid plan.
Prompt Adherence
5/5Tested- Method
- 9 of the 40 fixed prompts (3 portraits, 2 product shots, 2 set-wording images, 2 busy scenes), typed in by script, first image kept, every written element counted at full size.
- Elements correct
- 55 of 56
- Busy scenes
- 15 of 15
- Remark
- Scale 5: 90 percent or more of elements correct. The sample was 9 prompts and 56 elements, not 40 prompts and about 240.
Raw notes
9 of the 40 fixed prompts (3 portraits, 2 product shots, 2 set-wording images, 2 busy scenes), typed in by script, first image kept, every written element counted at full size.
55 of 56 elements were right (98.2 percent). Portraits 17 of 18, product shots 11 of 11, set-wording images 12 of 12, busy scenes 15 of 15.
The one miss: the girl holds the red balloon string in her left hand, not the right hand the prompt asked for. Kitchen scene 8 of 8: three red apples left of the blue jug, a dog under the table, a cat on the chair, yellow curtains. Beach scene 7 of 7.
Nothing was added that changed the scene.
- 9 of 40 prompts, and only two of the six busy scenes. The two groups not run (styles and character scenes) and four busy scenes are the ones the free allowance could not pay for.
Photorealism & Anatomy
4/5Tested- Method
- Five portrait and product images (P1, P2, P3, F1, F4), one run each, inspected at 100 percent for faults that give a generated image away.
- Images that pass
- 4 of 5
- Hand and face faults
- 0
- Remark
- Scale 4: 18 to 21 of 24 pass. The sample is 5 images, so 4 of 5 (80 percent) stands for 19 to 20 of 24.
Raw notes
Five portrait and product images (P1, P2, P3, F1, F4), one run each, inspected at 100 percent for faults that give a generated image away.
Passed (4): the man on the bench, the laughing woman, the girl with the balloon and the olive oil bottle, with no hand, face or skin faults.
Failed (1): the watch dial has no 1 at the one o'clock position, only a tick, so its numerals read 10, 11, 12, then 2.
Hand faults 0. Face faults 0. Faults in text-like marks on objects 1.
- Five images in one run, not 24 images in two runs.
Text in Images
5/5Tested- Method
- Two of the six set-wording prompts: the coffee shop sign and board (11 words) and the night market poster (13 words), 24 words in all.
- Words correct
- 24 of 24
- Images fully correct
- 2 of 2
- Remark
- Scale 5: 95 percent or more of words correct. Two of two images fully correct; the scale asks for five or six of six, so the grade rests on a small sample.
Raw notes
Two of the six set-wording prompts: the coffee shop sign and board (11 words) and the night market poster (13 words), 24 words in all.
Every one of the 24 words was spelled right, in the place asked for.
Both images were fully correct, with no invented text. The sign reads "Harbour Bean Coffee Roasters" and the board "Open Daily 7 am to 5 pm". The poster carries "Summer Night Market", "Saturday 14 June from 6 pm" and "Free Entry for Everyone" in the right order from top to bottom.
The two images took 17.3 and 20.3 seconds, the same as the other prompts.
- Two of six prompts; the book cover, the fridge note, the street sign and the jar label were not run.
Speed per Image
4/5Tested- Method
- Eight of the nine generations were timed from pressing Generate to the full-size image on screen, in Fast mode at 1x.
- Median per image
- 17.3 s
- Slowest image
- 21.3 s
- Remark
- Scale 4: a median of 10 to 19.9 seconds.
Raw notes
Eight of the nine generations were timed from pressing Generate to the full-size image on screen, in Fast mode at 1x.
Median 17.3 seconds, average 18.1, fastest 16.2, slowest 21.3.
The two set-wording images took 17.3 and 20.3 seconds; the portraits and products took 16.2 to 21.3.
No generation failed. The first image was not timed because my screen recorder failed that run for lack of disk space.
- 8 generations, not 40, all in Fast mode, which is paid for in Gems.
Usage Rights, Watermarks & Privacy
3/5Tested- Method
- The terms of service (read 6 November 2026), the plans page and the downloaded file were checked against the ten-point list.
- Points met
- 6 of 8 checked
- Remark
- Scale 3: 6 or 7 of 10. 6 of 8 checked is 75 percent, which stands for 7 or 8 of 10; I took the lower figure because two points could not be checked.
Raw notes
The terms of service (read 6 November 2026), the plans page and the downloaded file were checked against the ten-point list. Eight points could be checked on the free plan.
Met (6): commercial use is allowed on every plan (the plans page says "Commercial license on every plan"); the terms say what happens to the output; the terms say images can be used commercially; the free download had no visible watermark; the home page says "Private by default" and the viewer has a Post button I never pressed; the downloaded file carried content credentials.
Not met (2): the terms let Mage use content that finds its way into training data, with no opt-out I could find, and Mage states no source for its training data.
Not tested (2): a watermark on paid downloads, and whether real people and brands are refused when asked for.
- Terms read by me, not a lawyer. No paid plan was bought. The terms also say users give up copyright, so anyone may use your images, which is a different thing from owning them.
Pricing & Value
4/5Tested- Method
- The free allowance was used until the 85 percent mark, and the plans page and the Gem packs were read.
- Free images I made
- 9, with 57 Gems left
- Cost per image from a Gem pack
- about $0.027
- Remark
- Scale 4: $0.02 to $0.049 an image. The free plan gives about 11 images once, not 10 a day, so it cannot reach the top mark.
Raw notes
The free allowance was used until the 85 percent mark, and the plans page and the Gem packs were read.
Free: a one-time 300 Gems, no daily refill. At 27 Gems an image I made 9 images and had 57 Gems left. Gems from the sign-up bonus expire after 90 days.
A Gem costs $0.001 in the smallest pack ($10 for 10,000 Gems), so one Mango 3 Turbo image costs about $0.027. The bigger packs add bonus Gems: $20 for 20,500, $40 for 42,000 and $100 for 110,000.
Memberships: Basic is $10 a month with 500 Gems and unlimited generations on base models, Pro is $30 with 1,500 Gems and unlimited Mango, Pro Plus is $60 with 3,500 Gems, Max is $200 with 15,000 Gems.
- No plan or pack bought; the price per image uses the Gem cost shown in the app and the pack price on the Gems page.
Coverage
Tested
- 9 of the 40 fixed prompts, first image kept: 3 portraits, 2 product shots, 2 set-wording images, 2 busy scenes
- Time and Gem cost of every image, read from the app and from the Gem transactions list
- The product's own Download button, opened and compared with the stored file, including its content credentials
- The terms of service, the plans page and the Gem packs read in full
Not tested or limited
- The free plan's 300 Gems paid for 9 images at 27 Gems each: 9 of the 40 prompts were run, and I stopped at 57 Gems left, the 85 percent mark
- I used Mango 3 Turbo at 27 Gems an image. The flagship Mango 3 costs 68 Gems, which would have paid for 3 images only. A different model could grade differently
- The frame was 4:5 (896 by 1,152 pixels) because the model's default was kept, not 1:1
- The first image was made while my screen recorder failed for lack of disk space, so 8 of the 9 images were timed
- Style range, consistency and editing were not run, and the upscaler (the Image Enhancer app, 2K on Pro and 4K on Pro Plus) was not run, so resolution is not graded
- Photorealism used one run of 5 images, not two runs of 12 prompts
- No plan was bought: paid watermark rules, the paid file set and cancellation were not tested. Prompts naming real people or brands were not run, so the enforcement point of the rights checklist was not tested
Still to test
- Style range, consistency, editing and the upscaler on a paid plan
- Photorealism with 24 images, and the other 31 prompts
- The flagship Mango 3 model and the GPT Image models
- Watermark rules and cancellation after a paid month