Test log
Vheer test data
Raw results for the Vheer 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
- 8
- Method
- Published methodology
- Environment
- Google Chrome on desktop
Results
| ID | Test | Key measurement | Score |
|---|---|---|---|
| T01 | Editing, Inpainting & Image-to-Image | – | n/a |
| T02 | Prompt Adherence | Elements correct: 186 of 209 | 4/5 |
| T03 | Photorealism & Anatomy | Images that pass: 20 of 24 | 4/5 |
| T04 | Text in Images | Words right: 45 of 61 | 3/5 |
| T05 | Style Range | Styles that hit: 10 of 10 | 5/5 |
| T06 | Character & Subject Consistency | Consistent images: 6 of 12 | 3/5 |
| T07 | Resolution & Upscaling | Native size: 1,248 px square | 3/5 |
| T08 | Speed per Image | Median per image: 17.4 s | 4/5 |
| T09 | Usage Rights, Watermarks & Privacy | Points met: 5 of 8 checked | 3/5 |
| T10 | Pricing & Value | Free Quality images: No cap seen | 4/5 |
| Overall (in the review) | 4.5/5 | ||
Editing, Inpainting & Image-to-Image
n/aNot Tested- Method
- Twelve fixed edits: four masked fills, four written edits, two restyles of my reference photos and two frame expansions.
- Remark
- Not graded: no edit was run.
Raw notes
Twelve fixed edits: four masked fills, four written edits, two restyles of my reference photos and two frame expansions.
Not run. The Edit button in the image viewer opens a crop, filter and drawing editor, with no change by prompt or mask; the Image to Image page takes a prompt and an upload, but I did not run the twelve edits on it.
- Run Image to Image with the four reference photos and the written edits on a later pass.
Prompt Adherence
4/5Tested- Method
- All 40 fixed prompts (6 portraits, 6 product shots, 6 set-wording images, 6 busy scenes, 10 lighthouse styles and 6 character scenes), typed in by script on the Vheer Quality model, first image kept, every written element counted at full size.
- Elements correct
- 186 of 209
- Busy scenes
- 36 of 42
- Remark
- Scale 4: 80 to 89 percent of elements correct. 89.0 percent sits at the top of that band; 90 percent would be a 5.
Raw notes
All 40 fixed prompts (6 portraits, 6 product shots, 6 set-wording images, 6 busy scenes, 10 lighthouse styles and 6 character scenes), typed in by script on the Vheer Quality model, first image kept, every written element counted at full size.
186 of 209 elements were right (89.0 percent). Portraits 35 of 35, products 32 of 32, styles 39 of 40, character scenes 26 of 30, busy scenes 36 of 42, set-wording images 18 of 30.
Busy scenes: kitchen 7 of 8 (two apples are clear, the third is hidden), market 4 of 7 (a pile of about eleven oranges, not five in a row, and the bicycle leans on a different stall), living room 4 of 6 (three green cushions and one white, not two and two), beach 7 of 7, desk 6 of 6, street 8 of 8.
Added content that changed the scene: invented lettering on the set-wording images, counted in the text test.
- The count is mine, made by eye at full size; the first image of each prompt counts, with no rerolls.
Photorealism & Anatomy
4/5Tested- Method
- The six portrait and six product prompts were each run twice (24 images), and every image was inspected at 100 percent for faults that give a generated image away.
- Images that pass
- 20 of 24
- Hand and face faults
- 0
- Remark
- Scale 4: 18 to 21 of 24 pass.
Raw notes
The six portrait and six product prompts were each run twice (24 images), and every image was inspected at 100 percent for faults that give a generated image away.
20 of 24 images pass as photographs. The four that fail are all warped lettering: the chef's jacket badge twice, the olive oil label and the numeral on the watch dial.
Hand faults 0. Face faults 0. Faults in text-like marks 4. Skin, eyes, teeth and hands passed in every portrait.
- The same model was used for all images; premium models are locked on the free plan.
Text in Images
3/5Tested- Method
- The six set-wording prompts (coffee shop sign and board, honey label, night market poster, book cover, fridge note, street signs), 61 words as the brief lists them, every word checked at full size.
- Words right
- 45 of 61
- Images fully correct
- 0 of 6
- Remark
- Scale 3: 70 to 84 percent of words correct.
Raw notes
The six set-wording prompts (coffee shop sign and board, honey label, night market poster, book cover, fridge note, street signs), 61 words as the brief lists them, every word checked at full size.
45 of 61 words were spelled right and in the place asked for (73.8 percent), and none of the six images was fully correct.
Sign "Harbur Bean Coffee Roasters"; board "Open Daily 7am 7o to 5pm" with an extra "7o". Label "Raw and Unfriuted Sanc 250 grams". Poster "Summmer Night Market", "Saturday 14 Junie from 6 pm" and "Free Entry for Everype". Cover "The Quiiet Orchord", "A Novle" and "by Hellen Marsh". Note "Please put the milk in fercfrccatior in the befou leave.". Street signs "Mill LaNe", "No Parking 8 am t6 pm" and "Bus Stopt".
Short, common words came out right; long words and the second and third lines went wrong. Every image carried extra or repeated lettering.
- Counting words is my call: "5pm" written without a space counts as right, and "Mill LaNe" counts as right because the letters are correct.
Style Range
5/5Tested- Method
- One subject, a lighthouse on a rocky coast at dusk, in ten styles named by technique: oil painting, watercolour, pencil sketch, flat vector, Japanese woodblock print, anime cel, pixel art, clay 3D render, 1970s film photograph and black ink line art.
- Styles that hit
- 10 of 10
- Remark
- Scale 5: all ten styles hit, subject kept.
Raw notes
One subject, a lighthouse on a rocky coast at dusk, in ten styles named by technique: oil painting, watercolour, pencil sketch, flat vector, Japanese woodblock print, anime cel, pixel art, clay 3D render, 1970s film photograph and black ink line art.
All ten styles show in the medium itself: brush marks, paper edges, graphite, solid shapes, a seal and paper on the woodblock print, flat colours, a visible pixel grid, clay rocks, film grain and ink lines.
The lighthouse and the rocks stayed in every image. Dusk light is clear in eight images and not visible in the black ink image on white paper; the pencil sketch shows it only as darker sky.
- The clay render is the weakest: the rocks look like clay, the lighthouse looks like stone.
Character & Subject Consistency
3/5Tested- Method
- A described woman (oval face, wavy auburn hair, a scar above her left eyebrow, a long green wool coat) and a described yellow thermos with a blue paper boat logo, each in six scenes.
- Consistent images
- 6 of 12
- Remark
- Scale 3: 6 to 8 of 12 consistent.
Raw notes
A described woman (oval face, wavy auburn hair, a scar above her left eyebrow, a long green wool coat) and a described yellow thermos with a blue paper boat logo, each in six scenes. The free text-to-image page has no character or reference feature, so the description was repeated word for word; each image was compared with the first.
6 of 12 images are consistent: the first woman and the second (same face, hair, coat and scar), and four of the six thermos images (yellow body, black lid and a clean blue boat).
The woman: the face changes in the beach image, and in the library, train and bridge images the scar moves to the forehead, between the eyebrows or the cheek. The thermos: in the rain and train images the logo carries extra lettering that was not in the description.
- The six thermos images were saved under shifted file names in the first run and matched to their prompts by eye.
Resolution & Upscaling
3/5Tested- Method
- Every saved image was measured, the 13 frame shapes were listed from the app, a 16:9 image was made to read its size, and the Download button was used once and compared with the stored file.
- Native size
- 1,248 px square
- Frame shapes
- 13
- Remark
- Scale 3: 1 megapixel or more native, with no upscaler.
Raw notes
Every saved image was measured, the 13 frame shapes were listed from the app, a 16:9 image was made to read its size, and the Download button was used once and compared with the stored file.
Every 1:1 image is 1,248 by 1,248 pixels (1.56 megapixels) as JPEG. The 16:9 image is 1,248 by 696 pixels (0.87 megapixels). The app offers 13 shapes, from 1:1, 1:2 and 2:1 to 16:9, 21:9 and 9:21.
The Download button gave the same file, byte for byte, with no watermark and no content credentials label in the file. I found no upscale button in the viewer, so no upscale was run.
- Only two sizes were measured, 1:1 and 16:9.
Speed per Image
4/5Tested- Method
- Each of the 40 first-run prompts was timed from pressing Generate to the full-size image on screen.
- Median per image
- 17.4 s
- Slowest image
- 41.4 s
- Remark
- Scale 4: a median of 10 to 19.9 seconds.
Raw notes
Each of the 40 first-run prompts was timed from pressing Generate to the full-size image on screen.
Median 17.4 seconds, average 19.8, fastest 15.4, slowest 41.4 (the 1970s film lighthouse).
37 of the 40 images took 15 to 25 seconds, and none failed. There was no queue message on the free plan, and the balance of 50 credits never moved.
- All times are from one connection on one afternoon, in a browser that also showed adverts.
Usage Rights, Watermarks & Privacy
3/5Tested- Method
- The terms (effective 15 January 2025, read 6 November 2026) and the downloaded file were checked against the ten-point list.
- Points met
- 5 of 8 checked
- Remark
- Scale 3: 6 or 7 of 10. 5 of 8 checked is 62 percent, so about 6 of 10.
Raw notes
The terms (effective 15 January 2025, read 6 November 2026) and the downloaded file were checked against the ten-point list. Eight points could be checked on the free plan.
Met (5): commercial use is allowed on the paid plan; the terms say you keep full ownership of your content; no visible watermark on free downloads; no public gallery was seen and the terms say files are deleted every 6 hours; and the terms say your content is not used to train AI models.
Not met (3): free users may use Quality images only for personal, non-commercial purposes; no source for training data is stated; the downloaded file carries no content credentials label.
Not tested (2): a watermark on paid downloads, and whether images of real people and brands are refused.
- Terms read by me, not a lawyer. The terms say files are deleted every 6 hours while the app says history is saved for signed-in users; I did not test which is true.
Pricing & Value
4/5Tested- Method
- I generated until the allowance stopped me, read the model list for credit costs, and read the plans page in both billing views.
- Free Quality images
- No cap seen
- Flux Klein on Premium
- about $0.04
- Remark
- Scale 4: $0.02 to $0.049 an image on the cheapest plan that allows commercial use.
Raw notes
I generated until the allowance stopped me, read the model list for credit costs, and read the plans page in both billing views.
Free: 50 credits a month and the Vheer Quality model costs 0 credits, so I made more than 60 images and the balance stayed at 50. Flux Klein costs 4 credits an image, so the 50 credits pay for 12 of those. Premium models are locked.
The cheapest paid plan is Premium at $9.99 a month ($8.33 a month paid yearly) with 1,000 credits and commercial use. A Flux Klein image at 4 credits costs about $0.04 on the monthly price, about $0.033 yearly; the Quality model stays at 0 credits.
Costs on top: credit packs from 400 credits for $3.99 to 7,000 for $69.99, and paid plans remove the adverts. The terms offer a 7-day refund.
- No plan bought. The credit cost of a premium model image was not read because those models are locked.
Coverage
Tested
- All 40 fixed prompts, first image kept, on the Vheer Quality model
- The 12 portrait and product prompts a second time (24 images in all) for the photorealism test
- Six scenes of a described woman and six of a described thermos for the consistency test
- Time of every image, the 13 frame shapes, the native size of a 1:1 and a 16:9 image
- The Download button, compared byte for byte with the stored image, and a check of the file for a content credentials label
- The terms (effective 15 January 2025), the plans page in both billing views and the model list
Not tested or limited
- The twelve fixed edits were not run, so editing is not graded; the Edit button is a crop, filter and drawing editor, and Image to Image was not tested with my reference photos
- No upscale was run: I found no upscale button in the image viewer, so resolution is graded on native size only
- The premium models (GPT Image 2, Nano Banana, SeeDream, Minimax) are locked on the free plan, so every image came from the free Vheer Quality model, not from the best model the site lists
- In the first batch the product scenes and the second photo run were saved against the wrong prompt after an advert thumbnail was caught as a result; I matched each file to its prompt by eye and re-ran one image, and the speed test uses only the 40 first-run prompts
- No plan was bought: watermarks on paid downloads, paid-plan speed and the cancel path were not tested; whether prompts naming real people are refused was not tested
Still to test
- Editing with Image to Image and my reference photos
- The premium models on a paid plan
- Paid-plan speed and watermark
- Whether a request for a real person is refused