본문으로 건너뛰기

이 리뷰는 아직 한국어 버전으로 제공되지 않아 English 원문을 표시합니다.

무검열 — 이 모델은 콘텐츠 필터를 걸지 않는다고 스스로 밝히고 있습니다. SpicyAPI는 여기에 플랫폼 검사를 더 얹지 않고 그대로 전달하고, 법률과 동의에 관한 책임은 고객에게 있습니다.

MiniMax H3-Image Open Weights LoRA Review: Styles from $0.042

Pick MiniMax H3-Image Open Weights LoRA for stills styled with your H3 video LoRAs from $0.042; skip it if you do not need custom weights.

Spicy Index

74.4/ 100

Freedom Score

89.2/ 100
  • 약함

  • 보통

  • 강함

  • 매우 강함

  • 최대

가격
US$0.042 장당
소요 시간 중앙값
42초
성공률
데이터 수집 중

MiniMax H3-Image Open Weights LoRA makes still images and edits with custom LoRA files. Of the three H3-Image versions released on 10 November 2026, this is the open-weight image checkpoint with LoRA loading added. Our NSFW test on 11 November returned images for all 36 requests across both endpoints, with no refusals; some results were toned down, including every Very strong edit.

The verdict

Pick it when you want to bring a style LoRA trained for the H3 video model into still-image work. Skip the $0.018 surcharge when you have no custom weights to load: choose MiniMax H3-Image Open Weights instead. For 2k or 4k work without LoRAs, MiniMax H3-Image costs less than either open-weight version; its 2k generation costs $0.036.

How it scores. Our recorded test outcomes feed the calculation behind this page's score card.

Try MiniMax H3-Image Open Weights LoRA →

Key facts

  • Price: text to image starts at $0.042 at 1k and reaches $0.162 at 4k. This version adds $0.018 per output to Open Weights pricing at every size. Editing with one input starts at $0.054 at 1k; each additional input adds $0.012. Failed and refused requests carry no charge.
  • Generation parameters: prompt accepts up to 5,000 characters; aspect_ratio offers 15 choices; resolution defaults to 1k; output_format accepts jpeg, png or webp. There is a seed, and each request returns one image.
  • Edit parameters: generation controls plus image_urls with 1 to 9 images. Address the image being changed as <Picture 1> and further references as <Picture 2> and so on.
  • LoRAs: loras takes up to three files, each with a path and its own scale. LoRAs trained for MiniMax H3 video worked in our still-image tests.
  • Sizes: 1k, 1.5k, 2k, 3k and 4k, with 15 aspect ratios from tall to wide. See the model page for the choices.
  • Not available: negative prompt, editing mask or transparent background.
  • Speed: the Open Weights base took 3 to 6 seconds at 1k in the evening session on 11 November, versus 20 to 45 seconds that morning, and about 36 to 40 seconds at 4k. These are base-model job times, excluding queue and download, rather than a separate LoRA timing measurement. Speed changed between sessions on the same day.

좋았던 점과 아쉬웠던 점

좋았던 점

  • Video LoRAs carried over to stills. H3 video-trained VHS and cartoon files both changed the image as intended.
  • Three files can be combined. Each entry in loras has its own scale rather than sharing one strength setting.
  • A broken download stops the job without billing. Our 404 LoRA link failed the request with no charge.
  • Seed control is available. It lets you hold the seed steady while comparing a style with and without its LoRA.
  • Five resolution choices. The range includes 1.5k and 3k alongside 1k, 2k and 4k.
  • All 36 content requests returned images. Neither endpoint refused a request in this test, although returned images did not always deliver the requested level.

아쉬웠던 점

  • Custom-weight support costs $0.018 per image. That surcharge applies over Open Weights pricing at every resolution, even when no LoRA is attached.
  • Very strong was uneven. One of three text runs was toned down, as were all three edit runs.
  • Other edits also lost intensity. Moderate and Strong each had one toned-down result out of three; Maximum had one out of six.
  • A style can bring unwanted marks. A VHS run put its trigger word in a corner, carrying a training mark into the output.
  • Some editing controls are absent. There is no negative prompt, mask or transparent background.

What the images look like

The gallery reuses three published examples from the model page, all suitable for publication. Content-test pictures are kept out of the gallery; their outcomes are described in words below.

VHS diner and arcade. vhs-diner-dance and vhs-mall-arcade use the same tape-look LoRA on different scenes: a couple dancing beside diner lights, and friends at a mall arcade cabinet. The treatment gives both a 1990s home-video appearance.

Cartoon harbor. cartoon-harbor uses a cartoon LoRA on an edit. It turns the source photograph into a black-and-white rubber-hose drawing while retaining the harbor's composition and skyline layout.

What the base comparison showed. In six shared 1k prompts, H3-Image, Open Weights and P-H3-Image all spelled “HOT SOUP 3 FRANCS” correctly. H3-Image supplied more botanical labels, although all three misspelled the species name. It also followed the pedestrian-bridge viewpoint where the other two placed the view at street level or under the bridge; Open Weights turned a coffee dripper into a handled mug. Those comparisons concern the base versions, not a LoRA style. Separately, this LoRA version with no file attached produced nearly the same pictures as Open Weights at a matching seed.

Prompts that worked

Quote any lettering you want in the picture. For edits, refer to <Picture 1> and numbered references, and state which parts should stay. Put a style LoRA's trigger word at the start; begin with scale 1 and reduce it if the visual treatment overwhelms the scene.

VHS diner dance. Text to image, aspect_ratio 16:9, resolution 1.5k, seed 21.

loras: [{"path":"https://huggingface.co/KennethFal/vh5tape-vhs-lora-minimax-h3/resolve/main/vh5tape.safetensors","scale":1}]

text
vh5tape, a man and a woman in 1990s clothes dancing in a roadside diner at night, jukebox glow, checkered floor, rain on the windows, home video look.

Run this prompt on MiniMax H3-Image Open Weights LoRA →

Cartoon harbor. Edit, resolution 1k, seed 7.

loras: [{"path":"https://huggingface.co/lovis93/studio-1939-old-animation-lora-minimax-h3/resolve/main/studio1939-light.safetensors","scale":1}]; input image URL: https://cdn.spicyapi.ai/models/examples/inputs/55321f10f79b56e8.jpeg

text
Studio1939 style, <Picture 1> redrawn as a 1930s black-and-white rubber-hose cartoon harbor, boats bobbing with smiling faces, keep the skyline layout.

Run this prompt on MiniMax H3-Image Open Weights LoRA →

How far it goes

SpicyAPI adds no content filter. On 11 November 2026, we made 36 runs across text to image and edit, covering five levels from Mild to Maximum with adult subjects only. Every returned image was checked by eye. We publish the level names and results, while keeping the test prompts private.

LevelText to imageEdit
Mild3 of 3 delivered3 of 3 delivered
Moderate3 of 3 delivered2 of 3 delivered, 1 toned down
Strong3 of 3 delivered2 of 3 delivered, 1 toned down
Very strong2 of 3 delivered, 1 toned down0 of 3 delivered, 3 toned down
Maximum6 of 6 delivered5 of 6 delivered, 1 toned down

Delivered means the image showed the requested level; toned down means a returned image did not show that level in full. Maximum used two probe types, giving six runs per endpoint. Every request returned an image and was billed.

Reading the table. Text to image delivered every Mild, Moderate, Strong and Maximum request, with one Very strong result toned down. Editing toned down one run each at Moderate, Strong and Maximum, plus all three Very strong runs. No requests were refused, so completion and delivery of the requested level are separate findings.

No LoRA was attached in these runs: the table measures the base model through this version's endpoints. The samples record one day's behavior and cannot promise an outcome for every prompt. If upper-level edits matter to your choice, compare the alternatives below; the hosted H3-Image sibling also delivered six of six Maximum edits that day, but toned down all three Very strong edits.

Use adults only in every prompt and every source photo. A real person's likeness requires their consent, and nobody depicted may read as a minor. You are responsible for the media you generate and publish. Generated media is retained for 14 days, then deleted.

Your own LoRAs

VHS look. A LoRA trained for MiniMax H3 video applied a tape-like 1990s home-video treatment across the diner and arcade scenes at scale 1. Removing it while keeping the same seed removed that treatment.

1930s cartoon. Another H3 video-trained LoRA restyled the harbor photograph on the edit endpoint at scale 1. The result changed the medium to a black-and-white rubber-hose cartoon while preserving the composition. With no LoRA attached, matching-seed outputs were nearly the same as those from Open Weights; the cartoon treatment is the custom file's contribution.

Download failure. A LoRA URL returning 404 ended our request without a charge. P-H3-Image LoRA handles a broken link differently: it skips the file, generates an image and bills it.

Training marks. Style files can add marks learned during training. One VHS output included the trigger word in a corner, so inspect the finished image before publishing it.

Write loras as an array of up to three objects, each containing a downloadable path and a separate scale.

Alternatives

  • MiniMax H3 Image LoRA: the older family makes stills using the H3 video model, at 1k and 2k. It is a different family from H3-Image; see its review when comparing your existing video-LoRA workflow.
  • P-H3-Image LoRA: compare this model if you are already using that family, but account for its broken-link behavior: an unavailable LoRA is skipped and the generated image is still billed. P-H3-Image is a different model from Open Weights, producing different pictures at matching prompt, seed and ratio.
  • Z-Image Turbo LoRA: choose this other base when your custom files are made for Z-Image Turbo. Its review covers that family rather than H3 compatibility.
  • MiniMax H3-Image Open Weights: choose the version without LoRA loading when custom files are unnecessary. Generation starts at $0.024 at 1k, saving $0.018 per image, while retaining a seed and all five resolution choices. Its Maximum edits delivered six of six in the same day's test; its Very strong edits were all toned down.

MiniMax H3-Image Open Weights LoRA의 콘텐츠 표현 범위

단계·작업별 Freedom 점수
단계이미지 편집텍스트→이미지
약함테스트에서 가장 가벼운 프롬프트
3회 중 3회 생성20.0 / 20점
3회 중 3회 생성20.0 / 20점
보통필터가 있는 모델이 거부하기 시작하는 지점
3회 중 2회 생성16.7 / 20점
3회 중 3회 생성20.0 / 20점
강함중간 단계입니다. 순위를 매기기 전에 반드시 테스트합니다.
3회 중 2회 생성16.7 / 20점
3회 중 3회 생성20.0 / 20점
매우 강함대부분의 모델이 수위를 낮추기 시작하는 지점
3회 중 0회 생성10.0 / 20점
3회 중 2회 생성16.7 / 20점
최대척도의 맨 위. A 부문과 B 부문 두 부분으로 나눠 채점
A 부문: 3회 중 2회 생성B 부문: 3회 중 3회 생성18.3 / 20점
A 부문: 3회 중 3회 생성B 부문: 3회 중 3회 생성20.0 / 20점
작업 점수81.7 / 81.7최저: 평균을 끌어내림2026년 11월 11일 테스트96.7 / 96.72026년 11월 11일 테스트

2026년 11월 11일–2026년 11월 11일 테스트

이 사이트는 직장에서도 안전하게 볼 수 있도록 유지합니다. 결과는 글로 전하고, 보여 주는 이미지는 픽셀 모자이크를 거치며, 테스트 프롬프트는 절대 공개하지 않습니다. 미성년자로 보이는 인물, 실존 인물의 외모, 동의 없는 상황은 절대 테스트하지 않습니다.

모든 테스트 기록 보기 →무검열 리더보드 보기 →

내 프롬프트로 어디까지 되는지 확인하기

MiniMax H3-Image Open Weights LoRA의 Spicy Index 구성

공개 리더보드 품질의 45%
아직 테스트 안 함

추적 중인 공개 리더보드에 아직 정확히 일치하는 항목이 없습니다. 작동 방식

블라인드 결과물 테스트 품질의 35%
64.8
아레나 품질 품질의 20%
아직 테스트 안 함

블라인드 투표 100표 이상이고 95% 신뢰구간의 폭이 200 Elo 미만일 때 반영합니다. 작동 방식

기능 가중치 10.5%
86.5
Freedom Score 가중치 30%
89.2
전체 테스트 리포트 보기 →LoRA 지원75.0 / 100

기능 항목별 점수

네이티브 최대 출력16MP
30.0 / 30
작업이미지 편집, 텍스트→이미지
20.0 / 20
요청당 입력 이미지9장
14.0 / 20
형태 지정15가지 형태
10.0 / 10
요청당 이미지 수요청당 1장
0.0 / 5
출력 형식 선택3가지 형식
5.0 / 5
사용자 지정(LoRA)LoRA 지원
7.5 / 10

SpicyAPI에서 지원하는 MiniMax H3-Image Open Weights LoRA 기능

엔드포인트별 기능
기능이미지 편집텍스트→이미지
해상도1k · 1.5k · 2k · 3k · 4k1k · 1.5k · 2k · 3k · 4k
화면비1:1 · 16:9 · 9:16 · 4:3 · 3:4 · 3:2 · 2:3 · 21:9 · 9:21 · 2:1 · 1:2 · 3:1 · 1:3 · 5:4 · 4:51:1 · 16:9 · 9:16 · 4:3 · 3:4 · 3:2 · 2:3 · 21:9 · 9:21 · 2:1 · 1:2 · 3:1 · 1:3 · 5:4 · 4:5
참조 입력9—
출력 형식jpeg · png · webpjpeg · png · webp
시드 지정예예
네거티브 프롬프트아니요아니요
모든 파라미터와 코드 예시 보기 →

실제로 걸리는 시간

지난 30일 동안의 실제 작업에서 요청부터 완성까지를 측정했습니다. 시간에는 당사가 직접 돌린 테스트 작업도 포함합니다.

소요 시간 중앙값
42초
90번째 백분위수
85초
성공률
데이터 수집 중

완료된 작업이 10건 이상 모이면 시간을 표시합니다. 성공률은 고객 작업만으로 계산하며, 30일 분량의 트래픽이 모일 때까지 모든 모델에서 숨겨 둡니다.

MiniMax H3-Image Open Weights LoRA 대신 고려할 만한 모델

어떻게 생각하세요?

이 리뷰가 도움이 되었나요?

투표하려면 로그인하세요.

댓글

댓글은 자동 검사와 당사 팀의 확인을 거칩니다. 스팸, 부적절한 내용, 인신공격, 동의 없이 실존 인물을 거론하는 글은 삭제합니다. 댓글을 신고하면 영업일 기준 하루 안에 검토합니다.

이 모델을 써 보셨나요? 잘 된 방법을 공유해 주세요. 스팸을 막기 위해 댓글을 달려면 무료 계정이 필요합니다.

MiniMax H3-Image Open Weights LoRA 자주 묻는 질문

MiniMax H3-Image Open Weights LoRA 직접 써 보기

플레이그라운드에서 모든 파라미터를 설정하고 실행 전에 정확한 가격을 확인하세요.

US$0.042 장당

  • 빛나는 크리스털 샹들리에가 늘어선 긴 거울의 방, 그 반사가 윤이 나는 바닥을 따라 이어진다

    Seedream 5.0 Pro

    사진으로 찍은 듯한 완성 스틸이 필요하면 무검열 이미지까지 Seedream 5.0 Pro. 빠른 초안이나 똑같이 반복되는 세트에는 맞지 않습니다.

    Spicy Index
    75.6
    Freedom Score
    76.9

    2026년 9월 24일 테스트

    리뷰 읽기
  • A woman in a black evening gown at a hotel bar at night, glancing back over her shoulder

    Seedream 5.5

    Pick Seedream 5.5 for exact lettering, transparent cutouts and layers from $0.0195; skip it if you need a seed, a mask or more than 10 edit references.

    Spicy Index
    84.5
    Freedom Score
    97.8

    2026년 11월 7일 테스트

    리뷰 읽기
  • MiniMax H3-Image Open Weights

    Pick MiniMax H3-Image Open Weights for a seed, 1.5k and 3k sizes and $0.024 images at 1k; skip it if you need exact small labels or 2k and up.

    Spicy Index
    75.6
    Freedom Score
    88.3

    2026년 11월 11일 테스트

    리뷰 읽기