Subscore

ExperienceTesting methodology

Experience measures what it is actually like to use the image generator.

An app can create great images and still be frustrating when generations take forever, fail regularly, or make you jump through several screens just to create one picture.

We check speed, failures, where images can be generated, whether custom prompts are accepted, whether editing is available, and whether adult image generation is supported.

7 evidence groups
7 scored tests

Experience is organized into 7 evidence groups. Each group contains one scored test: Speed, Failures, Chat Generation, Separate Generator, Custom Prompts, Image Editing, and NSFW Support.

How the Experience score is calculated

Every test gets a score from 0 to 10.

We multiply each test score by how much it counts. We then add all the points together to calculate the final Experience score.

The live system stores four tests at a weight of 13 and three tests at a weight of 12. We turn those into public percentages that add up to 100%. Experience makes up 33% of the Images score.

14.77%Speed
14.77%Failures
14.77%Chat Generation
14.77%Separate Generator
13.64%Custom Prompts
13.64%Image Editing
13.64%NSFW Support

Weighted evidence groups (combined 100%)Experience score33% of Images score

View exact calculation

Each test gets a score from 0 to 10. We multiply that score by how much the test counts. We then add all the points together.

Experience makes up 33% of the Images score. Images makes up 15% of the overall performance score.

Scored tests and weights

Scored test	How much it counts	See scoring
Speed	14.77%	View
Failures	14.77%	View
Chat Generation	14.77%	View
Separate Generator	14.77%	View
Custom Prompts	13.64%	View
Image Editing	13.64%	View
NSFW Support	13.64%	View
Total	100%

How the score is calculated

Each test gets a score from 0–10
Test score × how much it counts
Example: 8 × 14.77% = 1.18 points
We do this for every test
We add all the points together
Final Experience score
Counts for 33% of Images

Example calculation

We multiply each test score by how much it counts. We then add all the points together.

Scored test	Test score	How much it counts	Calculation	Points added
Speed	8.00	14.77%	8.00 × 14.77%	1.18
Failures	9.00	14.77%	9.00 × 14.77%	1.33
Chat Generation	10.00	14.77%	10.00 × 14.77%	1.48
Separate Generator	10.00	14.77%	10.00 × 14.77%	1.48
Custom Prompts	10.00	13.64%	10.00 × 13.64%	1.36
Image Editing	5.00	13.64%	5.00 × 13.64%	0.68
NSFW Support	5.00	13.64%	5.00 × 13.64%	0.68
Final Experience score	—	100%	Add all points	8.19/10

Special cases

Not Applicable

If a test does not apply, we remove it and spread its weight across the remaining tests.

Unknown

If we cannot verify a result, the test receives a score of 0.

Manual adjustment

In rare cases, we may adjust a score when the calculated result is clearly unfair. We always record the reason.

Why Experience matters

Images are one of the most popular features in AI girlfriend apps.

But the finished image is only part of the experience.

A generator is not very useful when every image takes several minutes, half the attempts fail, or you cannot write your own prompt.

This matters even more because many apps charge credits for every generation. A failed image or heavily restricted prompt can mean spending more money just to get something usable.

Ease of use also matters because many people using these apps are beginners. They should not need to understand advanced prompting or search through several menus before creating their first image.

A strong image generator should be quick, reliable, and easy to use while still giving you enough control.

What good Experience looks like

A high Experience score means images generate quickly, failures are rare, and the app gives you several useful ways to create and edit images. A lower score means generations are slow, regularly fail, or give you very little control.

Main benchmarks

GoodTypicalWeak

Generation speed

20 sec or less21–40 secOver 40 sec

Failed attempts

20% or less21–40%Over 40%

Custom Prompts

YesLimitedNo

Minimum useful standard

Chat Generation: Yes
Separate Generator: Yes
Custom Prompts: Yes
Image Editing: Yes
Failures: 20% or less

Common red flags

Generations regularly get stuck or produce nothing
Failed attempts still use credits
You can only choose from preset prompts
Images can only be created through one limited part of the app
Editing simply creates a completely different image
Adult image support is advertised but repeatedly blocked

How we test Experience

We use a paid account and test every image-generation tool available.

For Speed, we time the image-generation attempts. The timer starts when generation is submitted and stops when the finished image becomes available.

For Failures, we count attempts that fail completely, stay stuck, produce no image, or produce an unusable result.

We then test whether images can be generated inside chat and through a separate image generator.

We enter three different free-form prompts to check whether custom prompting is available.

For Image Editing, we try three simple tasks: change the clothing, change the background, and change the pose.

Finally, we review the platform’s rules and complete three allowed adult-image tests where legally appropriate.

Limitations and edge cases

Generation speed can change depending on the time of day, server demand, internet connection, and the number of images requested at once.

A slow result does not always mean the generator itself is bad. Timing several attempts helps reduce the effect of one unusually slow generation.

Some features may also be available only with certain characters, subscriptions, or devices.

A separate generator and chat generation are both useful, but they serve different users. Beginners may prefer creating an image directly in chat, while more experienced users may prefer a dedicated generator with more controls.

Image Editing on this page only checks whether basic editing features exist. We test whether those edits are accurate under the Images Accuracy methodology.

NSFW Support can also change when a platform updates its rules. Our result reflects what the platform allowed on the recorded test date.

Evidence groups

Experience has 7 evidence groups made up of 7 scored tests.

Speed measures how long the image generator normally takes to produce a finished image.

Fast generation makes it easier to try ideas and correct a bad result without waiting several minutes every time.

Speed

14.77%of Speed (14.77% total)

How long the image generator normally takes to produce a finished image.

How we test

We time the image-generation attempts. The timer starts when we submit the generation and stops when the finished image is available. We use the median time. The median is the middle result after placing all generation times in order. This prevents one unusually slow or fast attempt from controlling the score.

What counts

Time spent waiting after pressing Generate
Loading and processing time
Time until the finished image can be viewed
Normal generations made through the paid account

What does not count

Time spent writing the prompt
Time spent downloading the image
Image-editing time
A confirmed internet or platform outage
Waiting caused by the tester leaving the page

Result shown

Median generation time: 14 seconds

Scoring

Faster generation means a higher score.

Median generation time	Score
10 or fewer	10/10
11–20	8/10
21–40	6/10
41–90	4/10
91–180	2/10
181+	0/10

A median time of 14 seconds scores 8/10.

Evidence group 2 of 7

Failures measures how often an image-generation attempt does not produce a usable result.

A generator can be fast but still waste your time and credits when too many attempts fail.

Failures

14.77%of Failures (14.77% total)

How often an image-generation attempt does not produce a usable result.

How we test

We record every failed attempt during the image-generation test. An attempt fails when it produces no image, remains stuck, shows a generation error, or produces a result that is completely unusable.

What counts

An error message instead of an image
A generation that never finishes
A blank or corrupted result
An image that is so broken it cannot reasonably be used
An attempt that uses credits but produces nothing useful

What does not count

An image we simply do not like
A prompt detail being missed
A minor visual problem
A usable image with weak composition
A result that takes a long time but eventually finishes

Result shown

Failure rate: 10%

2 failed attempts out of 20

Scoring

Fewer failures means a higher score.

Failure rate	Score
0%	10/10
20%	8/10
40%	6/10
60%	4/10
80%	2/10
100%	0/10

A failure rate of 10% scores 9/10.

The exact percentage is used between these points.

Evidence group 3 of 7

Chat Generation measures whether you can request an image without leaving the conversation.

This is usually the easiest option for beginners because the image can be connected directly to what you are talking about.

Chat Generation

14.77%of Chat Generation (14.77% total)

Whether you can request an image without leaving the conversation.

How we test

We request one image in three separate chats. We check whether the request works and whether the image appears inside the conversation.

What counts

Requesting an image through a normal chat message
An image created from the current conversation
The result appearing inside the chat
The feature working with normal paying characters

What does not count

Opening a separate generator
Clicking an outside link
Receiving a pre-made profile picture
A text reply claiming an image was sent when none appears
A request that uses credits but produces no image

Result shown

Yes — images were generated in all 3 chats

Scoring

We use the result categories below.

Result	Score
Yes — all three attempts worked	10/10
Limited — only some attempts worked or important restrictions applied	5/10
No — chat generation was unavailable	0/10
Unknown	0/10

In this example, Chat Generation scores 10/10.

Evidence group 4 of 7

Separate Generator measures whether the app has a dedicated image-generation tool outside the chat.

A separate generator usually gives users more space for prompts, settings, styles, and other controls.

Separate Generator

14.77%of Separate Generator (14.77% total)

Whether the app has a dedicated image-generation tool outside the chat.

How we test

We check whether the platform has a separate image-generation page or tool. When available, we create three images through it.

What counts

A dedicated image-generation page
A separate creation tool with its own controls
Three completed image generations
A tool available to normal paying users

What does not count

Generating images only inside chat
A marketing demo
A gallery without a Generate button
A tool that repeatedly fails
An outside service owned by another company

Result shown

Yes — all 3 images were created through a dedicated generator

Scoring

We use the result categories below.

Result	Score
Yes — all three attempts worked	10/10
Limited — restrictions or failures affected the feature	5/10
No — no separate generator was available	0/10
Unknown	0/10

In this example, Separate Generator scores 10/10.

Evidence group 5 of 7

Custom Prompts measures whether you can write your own image instructions.

Preset buttons are easy to use, but they can become limiting when you want a specific outfit, pose, location, or scene.

Custom Prompts

13.64%of Custom Prompts (13.64% total)

Whether you can write your own image instructions.

How we test

We enter three different free-form prompts. We check whether each prompt is accepted and whether major restrictions affect what can be entered.

What counts

A free-text prompt box
Users writing their own image instructions
All three prompts being accepted
Normal safety rules that are clearly explained

What does not count

Only choosing from preset buttons
A prompt selected automatically by the app
Editing a character description
A text box that cannot be submitted
Prompts so restricted that most normal instructions are blocked

Result shown

Yes — all 3 free-form prompts were accepted

Scoring

We use the result categories below.

Result	Score
Yes — all three prompts were accepted	10/10
Limited — prompts worked but were heavily restricted	5/10
No — custom prompts were unavailable	0/10
Unknown	0/10

In this example, Custom Prompts scores 10/10.

Evidence group 6 of 7

Image Editing measures whether you can make basic changes to a generated image.

Editing can save time and credits because you do not need to regenerate the full picture just to change one detail.

Image Editing

13.64%of Image Editing (13.64% total)

Whether you can make basic changes to a generated image.

How we test

We try three basic editing tasks: change the clothing, change the background, and change the pose. We record how many editing types are supported.

What counts

A tool that changes the clothing
A tool that changes the background
A tool that changes the pose
Editing through normal user controls
A finished edited image

What does not count

Regenerating a completely new image
Editing through outside software
Changing the prompt before the first generation
A button that appears but does not create an edit
Manual photo editing performed by support

Result shown

Limited — clothing and background could be changed, but pose editing was unavailable

2 of 3 editing types supported

Scoring

We use the result categories below.

Result	Score
Yes — all three editing tasks were supported	10/10
Limited — only some tasks worked or important restrictions applied	5/10
No — image editing was unavailable	0/10
Unknown	0/10

In this example, Image Editing scores 5/10.

Evidence group 7 of 7

NSFW Support measures whether the image generator supports adult content under the platform’s current rules.

Because AI girlfriend apps are adult platforms, many users want to know what type of image content is actually allowed before paying.

NSFW Support

13.64%of NSFW Support (13.64% total)

Whether the image generator supports adult content under the platform’s current rules.

How we test

We read the platform’s current rules and help pages. Where legally appropriate, we complete three allowed adult-content tests. We check whether the requests are accepted and whether important restrictions affect the feature.

What counts

Adult image generation clearly allowed by the platform
Requests that follow the platform’s stated rules
Finished images produced through normal user access
Clear information about what is and is not allowed

What does not count

Requests that break the platform’s rules
Illegal content
Marketing claims we cannot verify
Adult-ready-made characters without image generation
A request that is accepted but produces no result

Result shown

Limited — adult images were supported, but important prompt restrictions applied

Scoring

We use the result categories below.

Result	Score
Yes — adult image generation worked under the stated rules	10/10
Limited — support existed with important restrictions	5/10
No — adult image generation was not supported	0/10
Unknown — we could not verify the rules or result	0/10

In this example, NSFW Support scores 5/10.

Back to Images