Test log
ComfyUI test data
Raw results for the ComfyUI 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
- 6
- Method
- Published methodology
- Environment
- Google Chrome on desktop
Results
| ID | Test | Key measurement | Score |
|---|---|---|---|
| T01 | Style Range | – | n/a |
| T02 | Character & Subject Consistency | – | n/a |
| T03 | Editing, Inpainting & Image-to-Image | – | n/a |
| T04 | Resolution & Upscaling | Native size: 1,024 px (1.05 MP) | n/a |
| T05 | Prompt Adherence | Elements correct: 24 of 25 | 5/5 |
| T06 | Photorealism & Anatomy | Images that pass: 2 of 2 | 5/5 |
| T07 | Text in Images | Words right: 11 of 11 | 4/5 |
| T08 | Speed per Image | Median time per image: 6.2 s | 5/5 |
| T09 | Usage Rights, Watermarks & Privacy | Points met: 6 of 10 | 3/5 |
| T10 | Pricing & Value | Free runs: 5 in total | 2/5 |
| Overall (in the review) | 4.2/5 | ||
Style Range
n/aNot Tested- Method
- Ten style prompts for one lighthouse.
- Remark
- Not graded: no style prompt was run.
Raw notes
Ten style prompts for one lighthouse.
Not run: ten images need ten runs and the free plan gives 5 in total; 4 were used on the fixed prompts.
- Run on a paid plan.
Character & Subject Consistency
n/aNot Tested- Method
- A described woman and a described thermos in six scenes each.
- 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 need twelve runs and the free plan gives 5 in total. No reference feature was tried.
- Run on a paid plan.
Editing, Inpainting & Image-to-Image
n/aNot Tested- Method
- Twelve fixed edits.
- Remark
- Not graded: no edit was run.
Raw notes
Twelve fixed edits.
Not run: twelve edits need twelve runs and the free plan gives 5 in total. No edit was tried.
- Run on a paid plan.
Resolution & Upscaling
n/aNot Tested- Method
- Every saved image was measured and the aspect ratios were read in the app.
- Native size
- 1,024 px (1.05 MP)
- Remark
- Not graded: the scale needs an upscale compared with its original, and no upscale was run.
Raw notes
Every saved image was measured and the aspect ratios were read in the app. No upscale was run.
All 4 images came out at 1,024 by 1,024 pixels (1.05 megapixels) as PNG.
The resolution node lists the aspect ratios 1:1, 2:3, 3:2, 3:4, 4:3, 9:16 and 16:9 and a megapixels setting that stood at 1.0.
- No upscale: the 5 free runs were used for prompts, and a fifth run would have passed the 85 percent point.
Prompt Adherence
5/5Tested- Method
- 4 of the 40 fixed prompts (a portrait, a product shot, a set-wording image and a busy scene), typed in by script, first image kept, every written element counted at full size.
- Elements correct
- 24 of 25
- Busy scene
- 7 of 8
- Remark
- Scale 5: 90 percent or more of elements correct. The sample was 4 prompts and 25 elements, not 40 prompts and about 240.
Raw notes
4 of the 40 fixed prompts (a portrait, a product shot, a set-wording image and a busy scene), typed in by script, first image kept, every written element counted at full size.
24 of 25 elements were right (96 percent). The portrait 6 of 6, the olive oil bottle 6 of 6, the coffee shop 5 of 5, the kitchen scene 7 of 8.
Kitchen scene: four apples instead of three. The apples are red and on the left of the blue jug, the dog is under the table, the cat is on the chair and the curtains are yellow.
Added content that changed the scene: two paper notices on the coffee shop door with made-up lettering.
- 4 of 40 prompts, and one busy scene. The free plan has 5 runs in total.
Photorealism & Anatomy
5/5Tested- Method
- The portrait and the olive oil bottle (P1 and F1), one run each, inspected at 100 percent.
- Images that pass
- 2 of 2
- Remark
- Scale 5: 22 or more of 24 pass. The sample is 2 images, so 2 of 2 (100 percent) maps to the top band; with so few images the grade is a first reading, not a measurement of 24.
Raw notes
The portrait and the olive oil bottle (P1 and F1), one run each, inspected at 100 percent.
Passed (2): the man on the bench has even skin, round glasses, a believable beard and a knitted cardigan; the bottle has a cork, a believable reflection of the window and rosemary with sharp needles.
Failed (0). Hand faults 0 (the man's hands are out of frame). Face faults 0. Faults in text-like marks on objects 0.
- Two images in one run, not 24 images in two runs. No hands were visible, so the hardest anatomy check was not made.
Text in Images
4/5Tested- Method
- One of the six set-wording prompts: the coffee shop with a sign reading "Harbour Bean Coffee Roasters" and a board reading "Open Daily 7 am to 5 pm".
- Words right
- 11 of 11
- Images fully correct
- 0 of 1
- Remark
- Scale 4: 85 to 94 percent of words, or four images fully correct. All words were right but the image carried extra text, which the method counts as a fault, so I placed it one band below the top.
Raw notes
One of the six set-wording prompts: the coffee shop with a sign reading "Harbour Bean Coffee Roasters" and a board reading "Open Daily 7 am to 5 pm".
Every one of the 11 words was spelled right, in the place asked for: the sign above the door and the board beside it.
The image was not fully correct: two paper notices on the door glass carry made-up lettering that I never asked for.
- One of six prompts; the poster, the book cover, the jar label, the handwritten note and the street sign were not run.
Speed per Image
5/5Tested- Method
- Each of the 4 runs was timed from pressing Run to the finished image on screen.
- Median time per image
- 6.2 s
- Slowest image
- about 12 s
- Remark
- Scale 5: a median under 10 seconds.
Raw notes
Each of the 4 runs was timed from pressing Run to the finished image on screen.
Median 6.2 seconds, average 6.9, fastest 3.1, slowest about 12.
The portrait took 4.1 seconds, the olive oil bottle 3.1, the coffee shop 8.2 and the kitchen scene about 12, read from the screen recording because my timer caught the wrong image on that run.
No run failed. There was no queue on the free plan at the time.
- 4 runs, not 40. The Krea-2 template is a fast model; a heavier model would take longer.
Usage Rights, Watermarks & Privacy
3/5Tested- Method
- The terms of service (read 6 November 2026), the pricing page and the saved files were checked against the ten points.
- Points met
- 6 of 10
- Remark
- Scale 3: 6 or 7 of 10 met. Two more points could not be tested on a free plan.
Raw notes
The terms of service (read 6 November 2026), the pricing page and the saved files were checked against the ten points.
Met (6): the terms say you keep all right, title and interest in the Output, on every plan; so commercial use is not blocked on free or paid; the terms say Comfy will not use Input or Output to train generative AI or diffusion models; Customer Data is treated as confidential; my downloaded files carry no visible watermark.
Not met (2): the terms and pricing page do not say what the models were trained on; the PNG files carry the prompt and the workflow as text but no content credentials.
Not tested (2): a watermark on paid downloads, and whether the app refuses prompts about real people. The terms ban synthetic media made to impersonate a real person without consent.
- Terms read by me, not a lawyer. Third-party partner models are covered by their own terms. No paid plan was bought.
Pricing & Value
2/5Tested- Method
- The free tier was used until I reached the 85 percent point, and the pricing page and its FAQ were read on 6 November 2026.
- Free runs
- 5 in total
- Cheapest plan
- $20 a month
- Remark
- Scale 2: credit costs unclear. The scale's top marks need a dollar cost per image, which the pages do not give.
Raw notes
The free tier was used until I reached the 85 percent point, and the pricing page and its FAQ were read on 6 November 2026.
Free: 5 runs in total. The counter fell from 5 to 1 over my 4 images. The pricing page names no reset.
The cheapest plan is Standard at $20 a month ($16 a month paid yearly), with 50,400 credits for the year when you pay yearly. Creator is $35 and Pro $100 a month.
What one image costs in credits is not stated: the FAQ says credits "map to GPU runtime" and gives only a five-second video, about 11 credits, as a reference. The free plan shows runs, not credits, so I could not read the cost of my own images.
- No plan bought, so the credit cost of an image and the renewal terms were not seen at the checkout.
Coverage
Tested
- 4 of the 40 fixed prompts, first image kept: 1 portrait, 1 product shot, 1 set-wording image, 1 busy scene
- The time of each run and the runs-left counter before and after
- The saved PNG files measured and searched for a watermark and for content credentials
- Terms of service, the pricing page and its FAQ read in full on 6 November 2026
Not tested or limited
- The free plan gives 5 runs in total, not a daily allowance: I made 4 images, so 4 of the 40 prompts were run
- Photorealism used 2 images (a portrait and a product shot), not 24
- Style range, consistency and editing were not run: they need 10, 12 and 12 images or more than the 5 runs. Resolution was not graded because no upscale was run
- No plan was bought: paid downloads, a watermark on paid files and the credit cost of one image were not seen
- 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 and editing on a paid plan
- Photorealism with 24 images, and the other 36 prompts
- Upscaling, and the credit cost of one image on Standard
- Cancellation and a watermark check after buying Standard