Skip to main content
Uncensored — This model declares no content filter. SpicyAPI forwards the request without adding a platform content scan; legal and consent duties remain with the customer.

Jib Mix Qwen LoRA Review: Uncensored Photoreal Images With Your LoRAs

Load camera, portrait or realism LoRAs on a fast uncensored photoreal base; skip it if you expect style LoRAs to turn it into anime or paint.

Open the Playground

Spicy Index

67.0/ 100

Freedom Score

90.0/ 100

Content levels

Tested 6 of 6 probe types · 25 runs

  • Mild · The lightest prompts in the test

  • Moderate · Where filtered models start to refuse

  • Strong · Mid-scale; must be tested before a model is ranked

  • Very strong · Where most models start to hold back

  • Maximum · The top of the scale, scored in two halves: class A and class B

Price
$0.03 per image
Median time
8.9 s
Success rate
Collecting data

Jib Mix Qwen LoRA is the same photoreal checkpoint as Jib Mix Qwen, a community merge by the Civitai creator J1B on top of Alibaba's 20-billion-parameter Qwen-Image, with one addition: you can load up to three LoRA files of your own per request. SpicyAPI serves it as one text-to-image endpoint in nine aspect ratios up to 1536 pixels on the long side. It is uncensored, and on 5 October nothing we sent it was refused.

What Jib Mix Qwen LoRA made in our tests

Straight from our test runs. Copy a prompt or open it in the playground to try it yourself.

  • Text to image

    Studio portrait photo. A 36-year-old woman with silver-blonde hair pinned up wears an off-shoulder black velvet gown and long earrings, looking over her bare shoulder at the camera. Soft key light, dark backdrop, rim light on her hair.

    Try this model

  • Text to image

    Amateur phone photo with on-camera flash. A 31-year-old man with curly dark hair and a short beard in a denim jacket over a white T-shirt grins at the camera at a crowded night market, string lights and food stalls behind him.

    Try this model

  • Text to image

    Photorealistic photo. A 30-year-old woman with wavy auburn hair in an oversized white shirt sits on a hotel windowsill at night, city lights behind her, holding a glass of red wine and smiling at the camera.

    Try this model

  • Text to image

    Candid photo. A 34-year-old man with a shaved head in a fitted grey henley leans on a kitchen counter at night, laughing, a glass of whiskey in his hand, warm pendant lights.

    Try this model

  • Text to image

    Japanese modern anime style, an upper body shot of a 30-year-old woman with long black hair in a red evening dress standing on a rooftop at night. City lights glow behind her and she looks at the camera.

    Try this model

  • Text to image

    Japanese modern anime style, an upper body shot of a 30-year-old woman with long black hair in a red evening dress standing on a rooftop at night. City lights glow behind her and she looks at the camera.

    Try this model

  • Text to image

    Japanese modern anime style, an upper body shot of a 30-year-old woman with long black hair in a red evening dress standing on a rooftop at night. City lights glow behind her and she looks at the camera.

    Try this model

  • Text to image

    Japanese modern anime style, an upper body shot of a 30-year-old woman with long black hair in a red evening dress standing on a rooftop at night. City lights glow behind her and she looks at the camera.

    Try this model

Make your own with Jib Mix Qwen LoRA

Open the Playground

The verdict

The honest summary of the LoRA side is narrower than you might hope. LoRAs trained for Qwen-Image or Qwen-Image-2512 load without an error, from Hugging Face or from a Civitai direct link. Photographic, camera and portrait LoRAs change the picture in small, visible ways. Style LoRAs, such as a modern-anime one, barely moved this strongly photographic base: the image stayed a photograph. And a LoRA trained for a different base, FLUX, loaded with no error and changed nothing at all, down to the pixel.

Pick it when you want the Jib Mix look with a camera or portrait LoRA layered on top, or you want to test a Qwen-Image LoRA quickly at low cost. Skip it when your goal is a drawn or painted style, where an illustration model with a LoRA slot works better.

In our 5 October blind test it scored 61.6 out of 100. On our Freedom methodology it scored 90 out of 100.

Try Jib Mix Qwen LoRA →

Key facts

  • Content freedom: with no LoRA loaded, the ladder gave the same verdicts as the base endpoint, pixel for pixel at the same seed. Single-subject content from L1 Suggestive to L3, consensual restraint and gore rendered as asked; the L4 Intense probe was softened. A nine-ratio sweep delivered 9 of 9.
  • LoRA slots: up to three per request, each a direct link to a .safetensors file with a strength from 0 to 4. A useful range in our tests was about 0.5 to 1.5.
  • Scores: the Spicy Index and Freedom Score are on this page's score card, calculated by the server. We quote only the components we measured.
  • Speed: 38 completed runs on 5 October took 3.2 to 36 seconds of model time, median 5.6. A LoRA file used for the first time took 10 to 25 seconds.
  • Sizes: nine aspect ratios up to 1536 pixels: 1328 x 1328 at 1:1, 1104 x 1472 at 3:4, 1536 x 864 at 16:9, 1536 x 656 at 21:9.
  • Price: billed per image at one flat price whatever the ratio, format or number of LoRAs loaded; failed requests are not charged. The live price is on the Jib Mix Qwen LoRA model page.
  • Not available: negative prompt, image input, steps, guidance or sampler settings, and more than one image per request.

What we liked and what held it back

What we liked

  • LoRAs from Hugging Face and Civitai both load. Direct links to Qwen-Image and Qwen-Image-2512 LoRAs worked, and so did a Civitai download link.
  • Photo and portrait LoRAs show. On one prompt and seed, a phone-camera snapshot LoRA, a portrait-retouch LoRA, a flash-photo LoRA and a realism LoRA each changed the face and the light visibly, though subtly.
  • Stacking works. Two LoRAs together (phone camera at 0.8 and portrait at 0.6) produced a clean candid photograph.
  • Same speed as the base. The median was 5.6 seconds, close to the plain Jib Mix Qwen endpoint.
  • Nothing was refused. The ladder behaved exactly like the base endpoint, and adding a LoRA did not introduce a filter.
  • Errors say what is wrong. A login-gated link and a web-page link each came back with a plain message, and the failed requests were not charged.
  • Repeatable. The same request twice gave a pixel-identical image.

What held it back

  • Style LoRAs barely work here. A modern-anime LoRA changed about 99 percent of the pixels at strength 0.5 and 1, yet the image stayed a photograph at 0.5, 1 and 2. At 4 the picture broke into a distorted close-up with oversized eyes. A Qwen-Image-2512 pixel-art LoRA also changed pixels without producing a pixel-art look.
  • A LoRA for another base does nothing, silently. A FLUX LoRA loaded without an error and the output was pixel-identical to running with no LoRA. Match the LoRA to a Qwen-Image base.
  • The effect of a good LoRA is small. Camera and portrait LoRAs nudge faces and light; they do not reinvent the picture.
  • Strength 4 breaks the image. Stay between 0.5 and 1.5 unless the LoRA's own page says otherwise.
  • Links must be direct, public files. A repository page fails with "you should provide the direct download link", and a login-gated file fails with a 401.
  • More than three LoRAs, or a strength above 4, is rejected at submission.
  • No negative prompt, no image input, one image per request.

What the images look like

The gallery shows published examples from the model page next to our own comparison images. Every image shows clothed adults, suggestive at most.

A portrait-retouch LoRA. The studio portrait of a woman in an off-shoulder velvet gown, looking over her shoulder against a dark backdrop, was made with a portrait LoRA at strength 1. The skin is smooth but still has texture, and the rim light on her pinned-up hair is where the prompt put it.

A phone-snapshot LoRA. The man at a night market in a denim jacket has the hard flash and slightly crooked framing of a phone photo. The woman on the hotel windowsill, made with the same LoRA, shows the curlier hair and harder light we saw when comparing it against no LoRA.

Two LoRAs stacked. A laughing man in a grey henley at a kitchen counter used the phone-camera LoRA at 0.8 and the portrait LoRA at 0.6. It reads as a candid photograph rather than a styled one.

The control and the style test. We ran one prompt and one seed with no LoRA, with a modern-anime LoRA at 1, and at 4. The no-LoRA image and the strength-1 image are both photographs of a woman in a red dress on a rooftop; the strength-4 image is a distorted close-up. The FLUX LoRA run is byte-for-byte the same file as the control.

Prompts that worked

Write full sentences and start with the kind of picture. State an age for every person. LoRAs trained for photography respond to camera words, so "amateur phone photo with on-camera flash" or "candid photo" gave the camera LoRAs something to latch onto.

1. A studio portrait for a portrait LoRA.

text
Studio portrait photo. A 36-year-old woman with silver-blonde hair pinned up wears an off-shoulder black velvet gown and long earrings, looking over her bare shoulder at the camera. Soft key light, dark backdrop, rim light on her hair.

Run this prompt on Jib Mix Qwen LoRA →

2. A flash snapshot for a camera LoRA.

text
Amateur phone photo with on-camera flash. A 31-year-old man with curly dark hair and a short beard in a denim jacket over a white T-shirt grins at the camera at a crowded night market, string lights and food stalls behind him.

Run this prompt on Jib Mix Qwen LoRA →

3. A candid at night.

text
Photorealistic photo. A 30-year-old woman with wavy auburn hair in an oversized white shirt sits on a hotel windowsill at night, city lights behind her, holding a glass of red wine and smiling at the camera.

Run this prompt on Jib Mix Qwen LoRA →

4. A candid with two LoRAs.

text
Candid photo. A 34-year-old man with a shaved head in a fitted grey henley leans on a kitchen counter at night, laughing, a glass of whiskey in his hand, warm pendant lights.

Run this prompt on Jib Mix Qwen LoRA →

Weak: asking for "anime style" and hoping a style LoRA will carry it. Strong: a photographic prompt plus a camera or portrait LoRA, checked on one image before you run a batch.

How far it goes

On 5 October we ran a content ladder with no LoRA loaded. The results matched the base endpoint pixel for pixel on the same seed. L1 Suggestive (lingerie, a slipping strap), L2 Partial and L3 Strong rendered, and so did a male L3 probe. Consensual restraint between adults and cinematic gore rendered as written. L4 Intense (an intimate scene between two adults) was softened: all three seeds drew two adults without a refusal, but the scene was posed rather than carried out and anatomy was muddled.

We also tried a third-party Qwen-Image adult LoRA on the L4 probe. It changed the composition into a posed couple and still did not draw the scene. So with this base, a LoRA is not a reliable way past the top rung. Nothing was refused and nothing was blocked by a filter; our Freedom methodology scored the family 90 out of 100, and the page label is Uncensored.

LoRA weights load with no review of what they contain, so what they draw is your responsibility. Adults only, in every prompt, every LoRA and every output. Never prompt for a real person's likeness without consent, never depict anyone who could read as a minor, and keep what you generate private unless you have the right to share it.

Bringing your own LoRA

Each request takes up to three LoRA entries. Each entry has a direct HTTPS link to a .safetensors file in the field path and a strength in scale, from 0 to 4. Use a link to the file itself, not to the repository page.

What loads. Qwen-Image LoRAs and Qwen-Image-2512 LoRAs both load, because they share an architecture, from Hugging Face or from a Civitai download link. A LoRA built for FLUX also "loads", but it has no effect.

What shows. On one prompt and seed, four photographic LoRAs (phone camera, portrait retouch, flash photo and realism) each changed the face and the light, visibly but by a small amount. A Civitai style LoRA nudged skin and light slightly. A modern-anime LoRA and a pixel-art LoRA changed pixels but not the style.

What to do first. Render once with no LoRA, once with it at 1, and compare. If nothing changes, the LoRA was probably trained for a different base. If the picture breaks, lower the strength.

What fails. A gated repository fails with Failed to load LoRA and a 401 message; a repository page fails with a message asking for the direct download link; four LoRAs or a strength of 4.5 are rejected at submission. Failed requests are not charged.

Alternatives

How far Jib Mix Qwen LoRA goes

Each level, on each task, based on confirmed results. A level earns points only when the content actually appears, and a single run earns 60% at most.

Freedom points by level and task
LevelText to image
MildThe lightest prompts in the test
4/4 delivered20.0 / 20 pts
ModerateWhere filtered models start to refuse
4/4 delivered20.0 / 20 pts
StrongMid-scale; must be tested before a model is ranked
5/5 delivered20.0 / 20 pts
Very strongWhere most models start to hold back
0/5 delivered10.0 / 20 pts
MaximumThe top of the scale, scored in two halves: class A and class B
Class A: 3/3 deliveredClass B: 4/4 delivered20.0 / 20 pts
Task score90.0 / 90.0Tested Oct 5, 2026

Tested Oct 5, 2026 to Oct 5, 2026

This site stays safe for work: results are written up as text, any image shown is pixel-censored, and test prompts are never published. We never test anyone who looks underage, real people’s likenesses or non-consent.

See every test record →See the uncensored leaderboard →

Run your own prompt and see how far it goes

Open the Playground

How Jib Mix Qwen LoRA earns its Spicy Index

Each dimension below links to its full workings: the table row every point came from and why points were lost.

Public leaderboards 45% of quality
Not yet tested

No exact match on a tracked public leaderboard yet. How it works

Blind output test 35% of quality
61.6
Arena quality 20% of quality
Not yet tested

Counts once the model has 100+ blind votes and a 95% interval under 200 Elo. How it works

Capability 10.5% weight
32.5
Freedom Score 30% weight
90.0
Open the full test report →LoRA support75.0 / 100

Capability, item by item

Points from the public scoring table. Upscaled tiers and limits we set ourselves don’t count.

Native max outputNot declared
0.0 / 30
TasksText to image
10.0 / 20
Input images per requestNone
0.0 / 20
Shape control9 shapes
10.0 / 10
Images per request1 image per request
0.0 / 5
Output format choice3 formats
5.0 / 5
Customization (LoRA)LoRA supported
7.5 / 10

What Jib Mix Qwen LoRA can do on SpicyAPI

Straight from the live model specification on SpicyAPI. Some creators offer more in their own apps; this is what you can call through our API today.

Capabilities by endpoint
FeatureText to image
Aspect ratios1:1 · 16:9 · 9:16 · 4:3 · 3:4 · 3:2 · 2:3 · 21:9 · 9:21
Output formatsjpeg · png · webp
Seed controlYes
Negative promptNo
See every parameter and code sample →

How long it really takes

Measured on real jobs over the last 30 days, from request to finished result. Times include our own test runs.

Median time
8.9 s
90th percentile
36 s
Success rate
Collecting data

A time appears once at least 10 completed jobs are in. Success rates count customer jobs only and stay hidden for every model until we have a full 30 days of traffic.

Models to consider instead of Jib Mix Qwen LoRA

What do you think?

Was this review useful?

Log in to vote.

Comments

Comments are checked automatically and by our team. We remove spam, offensive content, personal attacks and anything that names a real person without consent. Report a comment and we review it within one business day.

Tried this model? Share what worked. You need a free account to comment, so we can keep spam out.

Jib Mix Qwen LoRA FAQ

Try Jib Mix Qwen LoRA yourself

Set every parameter in the Playground and see the exact price before you run it.

$0.03 per image

More reviews

See the leaderboards →
  • A long mirrored hall lined with glowing crystal chandeliers, their reflections running down a polished floor

    Seedream 5.0 Pro

    Pick Seedream 5.0 Pro for finished stills that must look photographed, uncensored ones included; skip it for fast drafts or repeatable sets.

    Spicy Index
    75.6
    Freedom Score
    76.9

    Tested Sep 24, 2026

    Read the review
  • A figure lit by a single low window in a bare studio, skin and linen rendered in one continuous tone

    Qwen Image 2.1

    Pick Qwen Image 2.1 for uncensored text-to-image and 10-reference editing at $0.024 a still; its LoRA sibling posted a cleaner ladder record in our tests.

    Spicy Index
    78.0
    Freedom Score
    96.3

    Tested Sep 24, 2026

    Read the review
  • A pale sculptural ring and a curved red strip

    Seedream 5.0 Flash

    Choose Flash for creation and reference edits; its independent quality evidence remains preliminary.

    Spicy Index
    73.9
    Freedom Score
    62.2

    Tested Sep 30, 2026

    Read the review