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.
Anima Turbo Review: Uncensored Anime Art, Fast and Low-Priced
Pick Anima Turbo for uncensored anime and illustration at a low per-image price; skip it for photorealism, readable text or precise edits.
Strong · Mid-scale; must be tested before a model is ranked
Delivered
Very strong · Where most models start to hold back
Delivered
Maximum · The top of the scale, scored in two halves: class A and class B
Mixed
Price
$0.012per image
Median time
13 s
Success rate
Collecting data
Anima Turbo is a 2-billion-parameter anime and illustration model from CircleStone Labs, built with Comfy Org on NVIDIA's Cosmos-Predict2 and a Qwen3 text encoder. Turbo is the distilled build, made to sample in 8 to 12 steps; the publisher calls it "only slightly worse than Anima-Aesthetic" on average while being much faster, and recommends starting with it. SpicyAPI serves it as two endpoints: text-to-image and image-to-image. It is uncensored: there is no content filter in front of it, and in our 5 October tests nothing we sent it was refused or toned down on text-to-image.
Straight from our test runs. Copy a prompt or open it in the playground to try it yourself.
Text to image
masterpiece, best quality, score_7, safe, 1girl, solo, mature female, adult woman, long wavy auburn hair, green eyes, off-shoulder knit sweater, bare shoulders, jeans, leaning on a rooftop railing, city skyline, golden hour, wind in hair, soft smile, looking at viewer, cinematic lighting, depth of field
masterpiece, best quality, safe. An anime illustration of a handsome man in his thirties working as a bartender late at night. He has short dark hair and light stubble, the sleeves of his white shirt are rolled up and the collar is open, and he is pouring a drink into a glass behind a dim wooden bar. Rows of amber bottles glow on the shelves behind him, and a warm rim light falls across his face and forearms.
masterpiece, best quality, score_7, safe, 1girl, 1boy, hetero, adult couple, mature female, mature male, summer festival at night, fireworks in the sky, paper lanterns, she wears a navy yukata with her hair up, he wears a grey yukata with the collar loose, standing close together on a wooden bridge, she looks back at him over her shoulder, warm lantern light, river reflections
masterpiece, best quality, safe, anime illustration, 1girl, solo, adult woman, mature female, short dark wavy hair, dark green velvet dress with long sleeves, holding a vintage microphone on a stand, jazz club stage, warm spotlight from above, dark background, soft smile, looking to the side
masterpiece, best quality, safe, anime illustration, 1boy, solo, adult man, mature male, short dark hair, stubble, white bathrobe, profile view, one hand touching a large window, city at dusk outside, warm lamp light behind him, blue evening light on the glass, quiet mood
masterpiece, best quality, score_7, safe, 1girl, solo, adult woman, mature female, short silver hair, blue eyes, beige trench coat, scarf, standing on a city street in autumn, falling leaves, looking at viewer, soft smile
masterpiece, best quality, score_7, safe, retro artstyle, 1990s (style), anime screenshot, 1girl, solo, adult woman, mature female, short silver hair, blue eyes, beige trench coat, scarf, standing on a city street in autumn, falling leaves, looking at viewer, soft smile
masterpiece, best quality, score_7, safe, (chibi:2), 1girl, solo, adult woman, mature female, short silver hair, blue eyes, beige trench coat, scarf, standing on a city street in autumn, falling leaves, looking at viewer, soft smile
masterpiece, best quality, score_7, safe, monochrome, greyscale, manga, screentone, 1girl, solo, adult woman, mature female, short silver hair, blue eyes, beige trench coat, scarf, standing on a city street in autumn, falling leaves, looking at viewer, soft smile
masterpiece, best quality, score_7, safe, 1boy, solo, adult man, mature male, short black hair, glasses, grey turtleneck, sitting by a cafe window, rain outside, holding a coffee cup, reading a book, warm indoor light, cozy
masterpiece, best quality, safe. An anime illustration of a man in his thirties sitting by a cafe window on a rainy afternoon. He has short black hair and glasses and wears a grey turtleneck, holding a cup of coffee in one hand while reading a paperback. Warm light from inside the cafe falls across the table, and raindrops streak the glass behind him.
masterpiece, best quality, safe, anime illustration, 1girl, solo, adult woman, mature female, short dark wavy hair, dark green velvet dress with long sleeves, holding a vintage microphone on a stand, jazz club stage, warm spotlight from above, dark background, soft smile, looking to the side
masterpiece, best quality, safe, anime illustration, 1girl, solo, adult woman, mature female, short dark wavy hair, dark green velvet dress with long sleeves, holding a vintage microphone on a stand, jazz club stage, warm spotlight from above, dark background, soft smile, looking to the side
masterpiece, best quality, safe, anime illustration, 1girl, solo, adult woman, mature female, short dark wavy hair, dark green velvet dress with long sleeves, holding a vintage microphone on a stand, jazz club stage, warm spotlight from above, dark background, soft smile, looking to the side
masterpiece, best quality, safe, anime illustration, 1girl, solo, adult woman, mature female, short dark wavy hair, dark green velvet dress with long sleeves, holding a vintage microphone on a stand, jazz club stage, warm spotlight from above, dark background, soft smile, looking to the side
It is also a narrow tool, and that is the whole review in two lines. Pick it when you want anime or illustration, you want content without a filter, and you want a low per-image price. Skip it when you need photorealism, more than a word or two of readable text, or an edit that follows an instruction.
Our blind output test scored it 44.7 out of 100 on 5 October, which is low, and the reason matters before you read anything into it. The prompt bank is the same one we run on every image family, and much of it asks for photographs; Anima answers those with anime by design. Edit consistency was also very low, because image-to-image at the default strength of 0.8 redraws the picture. For comparison on the same test, GPT Image 2 scored 81.2, Seedream 5.0 Flash 77.8 and Nano Banana 2 70.3. The test measures photoreal adherence, and Anima is an illustration model, so read the number as "not a camera" rather than "weak at drawing".
Content freedom: on 5 October, six probe types with three runs each per endpoint, random seeds, no LoRA. Text-to-image delivered 18 of 18, including L5 Extreme (consensual restraint between adults, and cinematic gore). Image-to-image delivered 15 of 18 as written and softened 3.
Scores: the Spicy Index and Freedom Score are on this page's score card, calculated by the server; we quote only the measured components in the text.
Price: billed per image by resolution tier. The 1.5k tier costs more than 1k, aspect ratio does not change the price, image-to-image costs the same as text-to-image, and failed requests are not charged. The live price is on the Anima Turbo model page.
Limits: no negative prompt, steps, CFG or sampler; output is always JPEG; one image per request; no masks or inpainting.
Speed: across 60 runs on 5 October, model time was 2 to 16 seconds per image and 52 of the 60 finished in under 10 seconds.
What we liked and what held it back
What we liked
Nothing was refused on text-to-image. All 18 runs across L1 to L5 came back as written. The platform adds no filter, and the model did not hold back.
Tags, sentences and mixes all work. Danbooru-style tag strings, natural-language paragraphs and combinations of both produced good results in our tests, so you can prompt it the way you already prompt.
A real style range from one set of words. Changing only the style tags on one character and one seed gave a clear 1990s cel-animation look, a chibi, and a black-and-white manga page.
It is quick. Two to sixteen seconds per image, with most under ten, makes iterating on a prompt cheap in time as well as money.
1.5k is native. The higher tier is generated at about 2.4 megapixels directly, not upscaled from the 1k image, up to 2048 pixels on the long side.
Fifteen aspect ratios. From 1:1 to 21:9 and tall 1:3, with the output sized to roughly one megapixel on the 1k tier whatever shape you pick.
Photo-to-illustration works well. At the right strength, image-to-image turns a photograph into clean anime while keeping its composition.
What held it back
It does not do photorealism. Ask for a photograph and you get an illustration. That is the model's design, and it is the main reason for the blind-test score above.
Text in images is limited to a word or two. It also sometimes draws a signature-like scribble or a line of pseudo-text in a corner, a habit picked up from artwork. We saw it in roughly one image in five. There is no negative prompt to suppress it, so change the seed or crop.
No negative prompt. The rating tag is your main lever for tone, and unwanted elements have to be avoided by wording alone.
Image-to-image restyles; it does not follow edit instructions. Asking an existing illustration for a new outfit, or to go from day to night, changed very little at strengths of 0.6 to 0.8, and at high strength it added pseudo-text. The default strength of 0.8 redraws the whole picture, so edit consistency is poor.
Same seed, similar image, not identical. The same request and seed gave us three slightly different renderings at different times, so treat a seed as a way to compare runs, not as a guarantee.
Refusal-free is not flawless. Anatomy and hands still need the occasional seed re-roll, like any image model.
What the images look like
The gallery mixes published examples from the model page with our own test images. Everything in it shows clothed adults, suggestive at most.
Clean line work and mood lighting. The model-page examples cover a woman on a rooftop at golden hour, a bartender at 1.5k, a couple in yukata under fireworks and a seaside railway crossing with no people at 21:9. The faces are stylised and consistent, and backgrounds hold up at wide ratios.
Style tags move the look a long way. On one character and one seed, retro artstyle, 1990s (style), anime screenshot gave a clear 90s cel look, (chibi:2) gave a chibi, and monochrome, greyscale, manga, screentone gave a black-and-white manga page. Weighting like (chibi:2) works and needs higher weights than SDXL users expect. The watercolor and pencil-sketch tags we tried changed the rendering only slightly, so do not count on a traditional-media look from tags alone.
Tags against sentences. We prompted one scene, a man with glasses reading in a rainy cafe, once as tags and once as four sentences on the same seed. Both are good, and they differ in framing and detail rather than in quality. The sentence version is in the gallery next to the tag version.
The strength ladder. With a photograph of a singer as the source and the same prompt and seed each time, strength is the whole story:
at 0.3 the result is still a photograph;
at 0.6 it already reads as an illustration with the same details;
at 0.75 it is clean anime with the same composition;
from 0.9 details start to change;
at 1.0 the source is ignored and you get a new image from the prompt;
at 0 the source comes back untouched.
We would start at 0.75 for photo-to-anime and move in steps of 0.05.
Sizes. The 1k tier gives 1024 x 1024 at 1:1, 1360 x 768 at 16:9 and 1248 x 832 at 3:2; the 1.5k tier gives 1536 x 1536, 2048 x 1152 and 1872 x 1248. In image-to-image the output keeps the source's aspect ratio, so a 16:9 photo came back at 1360 x 752 on the 1k tier.
Prompts that worked
Use lowercase tags with spaces instead of underscores. The publisher's order is quality and rating tags, then head count (1girl, 1boy, no humans), character, series, artist, then general tags, and a good opening is masterpiece, best quality, optionally score_7. Artist styles need an @ in front of the name or the effect is weak. For natural language the publisher recommends at least two sentences, because very short prompts drift. Because there is no negative prompt, every prompt below starts with the safe rating tag. The model also knows sensitive and nsfw, and one stronger tag, which remove restraint; we do not print prompts that use them.
1. One character, tags only, easy to restyle.
text
masterpiece, best quality, score_7, safe, 1girl, solo, adult woman, mature female, short silver hair, blue eyes, beige trench coat, scarf, standing on a city street in autumn, falling leaves, looking at viewer, soft smile
Add retro artstyle, 1990s (style), anime screenshot for a cel look, or monochrome, greyscale, manga, screentone for a manga page.
2. Sentences, four of them.
text
masterpiece, best quality, safe. An anime illustration of a man in his thirties sitting by a cafe window on a rainy afternoon. He has short black hair and glasses and wears a grey turtleneck, holding a cup of coffee in one hand while reading a paperback. Warm light from inside the cafe falls across the table, and raindrops streak the glass behind him.
masterpiece, best quality, score_7, safe, 1girl, 1boy, hetero, adult couple, mature female, mature male, summer festival at night, fireworks in the sky, paper lanterns, she wears a navy yukata with her hair up, he wears a grey yukata with the collar loose, standing close together on a wooden bridge, she looks back at him over her shoulder, warm lantern light, river reflections
4. Photo to anime on the image-to-image endpoint. Send a photograph as image_url, set strength to 0.75 and describe the scene you want in the same composition.
text
masterpiece, best quality, safe, anime illustration, 1girl, solo, adult woman, mature female, short dark wavy hair, dark green velvet dress with long sleeves, holding a vintage microphone on a stand, jazz club stage, warm spotlight from above, dark background, soft smile, looking to the side
Weak: a one-line prompt such as "a girl in a park". Very short prompts drift. Strong: any prompt above, with quality tags, head count, a short description of the subject and one sentence of lighting. Every character is an adult in these prompts, and we say so in the tags; keep doing that.
How far it goes
On 5 October we ran our Freedom methodology on both endpoints: six probe types, three runs each, random seeds, no LoRA. The probe types are L1 Suggestive (lingerie, a slipping strap), L2 Partial, L3 Strong, L4 Intense (an intimate scene between two adults), and the two L5 Extreme types, consensual restraint between adults and cinematic gore.
Text-to-image delivered 18 of 18. Every probe at every level came back as written, nothing softened and nothing refused, which makes the endpoint score 100.
Image-to-image delivered 15 of 18 and softened 3. L1 to L4 were all delivered 3 of 3. On consensual restraint, 2 runs delivered and 1 was softened. On cinematic gore, 1 delivered and 2 were softened. The gore result is a test artefact more than a policy: the edit source was a knight in full armour and strength 0.8 kept the armour, so injuries could not appear. That puts the endpoint score at 95 and the family Freedom score at 97.5. The full run is in the 5 October Freedom report.
Two earlier checks on the same day agree. In a first pass of 25 requests across L1 to L4, with and without LoRAs and with female and male subjects, we saw zero refusals and zero substitutions, and L4 was rendered as written on the first try. Another 60 aspect-ratio runs used an L2-level prompt, and all 60 delivered. Image-to-image started from a clothed source, and asking for L3 or L4 at strength 0.8 to 0.9 was followed.
The rating tags steer tone more than anything else you can set. With a clothed description, safe and sensitive gave almost the same picture on one seed, and the publisher documents nsfw and a stronger tag as the ones that remove restraint. Because the model is trained on anime and illustration, what comes out reads as illustration, never as a photograph of a real person. The three gallery images tied to this section are suggestive at most: a woman in a slip dress seen from behind, two adults dancing close in a dim bar, and a man in an unbuttoned shirt in a doorway.
Adults only, in every image you generate and publish. Never prompt for a real person's likeness without consent, never depict anyone who could read as a minor, and keep your own outputs private unless you have the right to share them; you are responsible for what you generate and publish.
Alternatives
Your own style or character LoRAs on the same base?Anima Turbo LoRA takes up to three LoRA files per request. A LoRA trained for a different base loaded without error and changed nothing in our test, so use ones trained for Anima.
A different anime look, text-to-image only?Prefect Pony XL is a community anime checkpoint prompted with quality tags, and it does not take an image input.
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
Level
Image editing
Text to image
MildThe lightest prompts in the test
3/3 delivered20.0 / 20 pts
4/4 delivered20.0 / 20 pts
ModerateWhere filtered models start to refuse
3/3 delivered20.0 / 20 pts
4/4 delivered20.0 / 20 pts
StrongMid-scale; must be tested before a model is ranked
5/5 delivered20.0 / 20 pts
6/6 delivered20.0 / 20 pts
Very strongWhere most models start to hold back
4/4 delivered20.0 / 20 pts
7/7 delivered20.0 / 20 pts
MaximumThe top of the scale, scored in two halves: class A and class B
Class A: 2/3 deliveredClass B: 1/3 delivered15.0 / 20 pts
Class A: 3/3 deliveredClass B: 3/3 delivered20.0 / 20 pts
Task score
95.0 / 95.0Lowest: pulls the average downTested Oct 5, 2026
100.0 / 100.0Tested 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.
Measured on real jobs over the last 30 days, from request to finished result. Times include our own test runs.
Median time
13 s
90th percentile
24 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.
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.
Anima Turbo FAQ
Yes. There is no content filter in front of it, and on 5 October its text-to-image endpoint delivered all 18 Freedom probes, up to L5 Extreme. Image-to-image softened 3 of 18, mostly because of how the gore edit was set up.
The same prompt bank is used for every image family, and it leans on photographic prompts that Anima turns into anime on purpose. Edit consistency also scored very low because the default image-to-image strength of 0.8 redraws the picture. On that test GPT Image 2 scored 81.2, Seedream 5.0 Flash 77.8 and Nano Banana 2 70.3. The test measures photoreal adherence; Anima is an illustration model.
Either, or both. In our tests tag strings, sentences and mixes all worked. If you use sentences, write at least two, because very short prompts drift.
It restyles rather than edits. Strength 0.75 turns a photo into clean anime with the same composition, but asking an illustration for a new outfit or a day-to-night change changed very little at 0.6 to 0.8. There are no masks and no inpainting.
Only a word or two. It may also draw a signature-like scribble in a corner, roughly one image in five in our runs. Change the seed or crop the corner.
It is billed per image by resolution tier, and the 1.5k tier costs more than 1k. Aspect ratio does not change the price, image-to-image costs the same as text-to-image, and failed requests are not charged. The live price is on the model page.
Try Anima Turbo yourself
Generate in Studio in a few clicks, or set every parameter in the Playground and see the exact price before you run it.
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.
