# Qwen Image X models

Canonical page: https://qwenimagex.com/

Generated from the live model configuration of qwenimagex.com. Credit prices per option are in pricing.md.

## Qwen Image 3 (default image model)

- Modes: Text to image, Image to image
- Reference images: up to 3
- Aspect ratios: auto, 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16
- Resolutions: 1K, 2K
- Outputs per run: 1 to 4
- Credits: 8 per image (1K) to 15 per image (2K)

### Our test runs

- Text to image, 1 output, 3:4, 1K: 8 credits, result in 253 s (run on 2026-11-10)
  - Prompt: "A café chalkboard menu on a brick wall, hand-lettered in white chalk, photographed straight on. Heading: "THE CORNER CUP". Each item on its own line with its price: "Flat White $4.50", "Oat Latte $5.25", "Cold Brew $4.75", "Matcha Latte $5.50", "Almond Croissant $3.95", "Banana Bread $3.25". A small line at the bottom: "Open daily 7am - 4pm"."
  - Result: Every word and price right; it added a lamp, a plant and dotted leaders. The slowest run of the test.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/best/best-menu-qwen.webp

## HappyHorse 1.1 (default video model)

- Modes: Text to video, Image to video, Reference to video
- Image to video inputs: first frame (required)
- Reference images: up to 9
- Aspect ratios: 16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 9:21, 21:9
- Resolutions: 720p, 1080p
- Duration: 3 to 15 seconds
- Credits: 25 per second (720p) to 30 per second (1080p), minimum 75 per run

## Qwen Image 2.1

- Modes: Text to image, Image to image
- Reference images: up to 16
- Aspect ratios: auto, 1:1, 3:2, 2:3, 4:3, 3:4, 5:4, 4:5, 16:9, 9:16, 2:1, 1:2, 3:1, 1:3, 21:9, 9:21
- Resolutions: 1K, 2K, 4K
- Outputs per run: 1 to 4
- Credits: 6 per image (1K) to 20 per image (4K)

## Qwen Image 2

- Modes: Text to image, Image to image
- Reference images: up to 10
- Aspect ratios: 1:1, 1:4, 1:8, 2:3, 3:2, 3:4, 4:1, 4:3, 4:5, 5:4, 8:1, 9:16, 16:9, 21:9, auto
- Outputs per run: 1 to 4
- Credits: 6 per image

## Nano Banana 2

- Modes: Text to image, Image to image
- Reference images: up to 14
- Aspect ratios: 1:1, 1:4, 1:8, 2:3, 3:2, 3:4, 4:1, 4:3, 4:5, 5:4, 8:1, 9:16, 16:9, 21:9, auto
- Resolutions: 1K, 2K, 4K
- Outputs per run: 1 to 4
- Credits: 8 per image (1K) to 20 per image (4K)

## Nano Banana

- Modes: Text to image, Image to image
- Reference images: up to 10
- Aspect ratios: 1:1, 9:16, 16:9, 3:4, 4:3, 3:2, 2:3, 5:4, 4:5, 21:9, auto
- Outputs per run: 1 to 4
- Credits: 4 per image

## GPT Image 2.5 Flare

- Modes: Text to image, Image to image
- Reference images: up to 16
- Aspect ratios: auto, 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16, 21:9
- Resolutions: 1K, 2K, 4K
- Outputs per run: 1 to 4
- Credits: 3 per image (1K) to 16 per image (4K)

### Our test runs

- Image to image, 3:4, 1K, 1 output, background type auto, 1 reference images: 3 credits, result in 138 s (run on 2026-11-10)
  - Prompt: "Put this same woman on a rainy Tokyo side street at night, neon signs and paper lanterns behind her, wearing a black raincoat and holding a clear umbrella. Keep her face, hair, freckles and nose stud exactly the same. Candid photo, shallow depth of field."
  - Result: Close shot under the umbrella: face, freckles and nose stud all kept, neon signs in Japanese behind her. Input sent as a JPEG copy of the selfie.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/yourself/me-tokyo-flare.webp
- Image to image, 3:4, 1K, 1 output, background type auto, 1 reference images: 3 credits, result in 105 s (run on 2026-11-10)
  - Prompt: "Turn this into an autumn park photoshoot of the same woman: golden leaves, a camel wool coat and a cream scarf, sitting on a wooden bench in warm late-afternoon light. Keep her face, hair, freckles and nose stud exactly the same."
  - Result: Close portrait, chin on hand, coat and scarf; face, freckles and stud kept. Input sent as a JPEG copy of the selfie.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/yourself/me-autumn-flare.webp
- Image to image, 3:4, 1K, 1 output, background type auto, 1 reference images: 3 credits, result in 61 s (run on 2026-11-10)
  - Prompt: "Make this a 1990s high school yearbook portrait of the same woman: mottled blue studio backdrop, soft flash, slight film grain, a denim jacket over a white tee. Keep her face, hair, freckles and nose stud exactly the same."
  - Result: Mottled blue backdrop and denim jacket; face, freckles and stud kept. Input sent as a JPEG copy of the selfie.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/yourself/me-yearbook-flare.webp
- Text to image, 3:4, 1K, 1 output, background type auto: 3 credits, result in 54 s (run on 2026-11-10)
  - Prompt: "A candid street portrait of an elderly fisherman in his seventies on a harbour pier at dawn: weathered skin with deep wrinkles, white stubble, a navy wool beanie and a yellow oilskin jacket, light rain. Shot on a 50mm lens at f/1.8, natural colour, no retouching."
  - Result: Side glance with very fine skin and beard detail and rain on the jacket; grey light, no dawn.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/best/best-portrait-flare.webp

## Nano Banana Pro

- Modes: Text to image, Image to image
- Reference images: up to 8
- Aspect ratios: 1:1, 9:16, 16:9, 3:4, 4:3, 3:2, 2:3, 5:4, 4:5, 21:9, auto
- Resolutions: 1K, 2K, 4K
- Outputs per run: 1 to 4
- Credits: 20 per image (1K) to 25 per image (4K)

### Our test runs

- Text to image, 3:4, 1K, 1 output: 20 credits, result in 54 s (run on 2026-11-10)
  - Prompt: "A candid street portrait of an elderly fisherman in his seventies on a harbour pier at dawn: weathered skin with deep wrinkles, white stubble, a navy wool beanie and a yellow oilskin jacket, light rain. Shot on a 50mm lens at f/1.8, natural colour, no retouching."
  - Result: Broad smile, sharp skin texture and rain on the jacket; grey morning, no dawn colour.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/best/best-portrait-nbpro.webp

## Nano Banana 2.1

- Modes: Text to image, Image to image
- Reference images: up to 10
- Aspect ratios: 1:1, 1:4, 1:8, 2:3, 3:2, 3:4, 4:1, 4:3, 4:5, 5:4, 8:1, 9:16, 16:9, 21:9, auto
- Resolutions: 1K, 2K, 4K
- Outputs per run: 1 to 4
- Credits: 3 per image (1K) to 8 per image (4K)

### Our test runs

- Image to image, 3:4, 1K, 1 output, 1 reference images: 3 credits, result in 48 s (run on 2026-11-10)
  - Prompt: "Put this same woman on a rainy Tokyo side street at night, neon signs and paper lanterns behind her, wearing a black raincoat and holding a clear umbrella. Keep her face, hair, freckles and nose stud exactly the same. Candid photo, shallow depth of field."
  - Result: Wide street shot: she stands small under the umbrella among lanterns and neon. The face reads as her, but the nose stud is too small to see at this distance. Input sent as a JPEG copy of the selfie.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/yourself/me-tokyo-nb21.webp
- Image to image, 3:4, 1K, 1 output, 1 reference images: 3 credits, result in 38 s (run on 2026-11-10)
  - Prompt: "Turn this into an autumn park photoshoot of the same woman: golden leaves, a camel wool coat and a cream scarf, sitting on a wooden bench in warm late-afternoon light. Keep her face, hair, freckles and nose stud exactly the same."
  - Result: Full bench scene in a camel coat and cream scarf; the face is a little fuller than in the selfie and the stud is gone. Input sent as a JPEG copy of the selfie.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/yourself/me-autumn-nb21.webp
- Image to image, 3:4, 1K, 1 output, 1 reference images: 3 credits, result in 38 s (run on 2026-11-10)
  - Prompt: "Make this a 1990s high school yearbook portrait of the same woman: mottled blue studio backdrop, soft flash, slight film grain, a denim jacket over a white tee. Keep her face, hair, freckles and nose stud exactly the same."
  - Result: The most 1990s look (grain, flash, mottled backdrop), but the hair turned black with tight curls and the stud disappeared: the biggest drift of the four. Input sent as a JPEG copy of the selfie.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/yourself/me-yearbook-nb21.webp
- Text to image, 3:4, 1K, 1 output: 3 credits, result in 35 s (run on 2026-11-10)
  - Prompt: "A café chalkboard menu on a brick wall, hand-lettered in white chalk, photographed straight on. Heading: "THE CORNER CUP". Each item on its own line with its price: "Flat White $4.50", "Oat Latte $5.25", "Cold Brew $4.75", "Matcha Latte $5.50", "Almond Croissant $3.95", "Banana Bread $3.25". A small line at the bottom: "Open daily 7am - 4pm"."
  - Result: Every word and price right, in neat sign-painter lettering with dotted leaders.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/best/best-menu-nb21.webp

## Seedream 5.5

- Modes: Text to image, Image to image
- Reference images: up to 10
- Aspect ratios: 1:1, 1:2, 2:1, 1:3, 3:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 9:21, 21:9
- Resolutions: 1K, 1.5K, 2K
- Credits: Text to image: 4 per image (1K) to 7 per image (2K); Image to image: 2 per input image to 7 per image (2K)

## Seedream 5.0 Pro

- Modes: Text to image, Image to image
- Reference images: up to 10
- Aspect ratios: 1:1, 1:2, 2:1, 1:3, 3:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 9:21, 21:9
- Resolutions: 1K, 2K
- Credits: 9 per image (1K) to 20 per image (2K)

### Our test runs

- Text to image, 3:4, 1K: 9 credits, result in 35 s (run on 2026-11-10)
  - Prompt: "A candid street portrait of an elderly fisherman in his seventies on a harbour pier at dawn: weathered skin with deep wrinkles, white stubble, a navy wool beanie and a yellow oilskin jacket, light rain. Shot on a 50mm lens at f/1.8, natural colour, no retouching."
  - Result: The only one with a dawn sky; the skin is a little smoother than in the other three.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/best/best-portrait-sdpro.webp
- Image to image, 1:1, 1K, 1 reference images: 9 credits, result in 116 s (run on 2026-10-22)
  - Prompt: "Keep the mug exactly as it is. Place it on a wet black rock beside a small waterfall, with morning mist in the background."
  - Result: Mug shape and handle kept, green a shade darker; mossy rock by a forest waterfall in mist.
  - Image: https://cdn.qwenimagex.com/assets/landing/ai-image-studio/models/seedream-5-pro/t4-edit.webp
- Text to image, 1:1, 1K: 9 credits, result in 38 s (run on 2026-10-22)
  - Prompt: "Studio product photo of a matte ceramic coffee mug in sage green on a light oak table, soft window light from the left, plain white background, no logos."
  - Result: Matte sage mug on light oak in soft light against a white background.
  - Image: https://cdn.qwenimagex.com/assets/landing/ai-image-studio/models/seedream-5-pro/t3-product.webp
- Text to image, 1:1, 1K: 9 credits, result in 40 s (run on 2026-10-22)
  - Prompt: "A candid photo of an elderly fisherman mending a net on a wooden dock, early morning light, shallow depth of field, shot on a 50mm lens."
  - Result: Elderly fisherman in a blue polo mending a net on a dock; bright sun rather than early-morning light.
  - Image: https://cdn.qwenimagex.com/assets/landing/ai-image-studio/models/seedream-5-pro/t2-portrait.webp

## Seedream 5.0 Flash

- Modes: Text to image, Image to image
- Reference images: up to 10
- Aspect ratios: 1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2, 21:9
- Resolutions: 1K, 2K
- Outputs per run: 1 to 4
- Credits: 3 per image (1K)

### Our test runs

- Image to image, 3:4, 1 output, 1K, 1 reference images: 3 credits, result in 16 s (run on 2026-11-10)
  - Prompt: "Put this same woman on a rainy Tokyo side street at night, neon signs and paper lanterns behind her, wearing a black raincoat and holding a clear umbrella. Keep her face, hair, freckles and nose stud exactly the same. Candid photo, shallow depth of field."
  - Result: Kept the selfie framing and the face almost exactly, stud and freckles included; the street only shows behind her. Input sent as a JPEG copy of the selfie.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/yourself/me-tokyo-sdflash.webp
- Image to image, 3:4, 1 output, 1K, 1 reference images: 3 credits, result in 14 s (run on 2026-11-10)
  - Prompt: "Turn this into an autumn park photoshoot of the same woman: golden leaves, a camel wool coat and a cream scarf, sitting on a wooden bench in warm late-afternoon light. Keep her face, hair, freckles and nose stud exactly the same."
  - Result: Same face and stud on the bench in coat and scarf; framing still close to the selfie. Input sent as a JPEG copy of the selfie.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/yourself/me-autumn-sdflash.webp
- Image to image, 3:4, 1 output, 1K, 1 reference images: 3 credits, result in 14 s (run on 2026-11-10)
  - Prompt: "Make this a 1990s high school yearbook portrait of the same woman: mottled blue studio backdrop, soft flash, slight film grain, a denim jacket over a white tee. Keep her face, hair, freckles and nose stud exactly the same."
  - Result: Face and stud kept; light denim jacket on a blue backdrop. Input sent as a JPEG copy of the selfie.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/yourself/me-yearbook-sdflash.webp
- Text to image, 16:9, 1 output, 1K: 3 credits, result in 21 s (run on 2026-11-10)
  - Prompt: "An anime character sheet of a young woman bike courier with short silver hair, a red bomber jacket, goggles pushed up on her forehead and a messenger bag: front, side and back views standing in a row, plus three small face close-ups (happy, surprised, angry). Clean line art, flat cel shading, plain white background."
  - Result: Clean sheet: front, side and back views plus three labelled faces; it added a grey hoodie.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/best/best-anime-sdflash.webp

## Grok Imagine Image 2.0

- Modes: Text to image, Image to image
- Reference images: up to 3
- Aspect ratios: 2:1, 20:9, 19.5:9, 16:9, 4:3, 3:2, 1:1, 2:3, 3:4, 9:16, 9:19.5, 9:20, 1:2, auto
- Resolutions: 1K, 2K
- Outputs per run: 1 to 4
- Credits: Text to image: 8 per image (1K, low) to 20 per image (2K, medium); Image to image: 10 per image (1K, low) to 20 per image (2K, medium)

### Our test runs

- Text to image, 16:9, 1K, quality low, 1 output: 8 credits, result in 22 s (run on 2026-11-10)
  - Prompt: "An anime character sheet of a young woman bike courier with short silver hair, a red bomber jacket, goggles pushed up on her forehead and a messenger bag: front, side and back views standing in a row, plus three small face close-ups (happy, surprised, angry). Clean line art, flat cel shading, plain white background."
  - Result: Polished art in two rows; it invented a "RUSH COURIER" logo on the jacket and bag.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/best/best-anime-grok.webp

## FLUX 3 Klein

- Modes: Text to image, Image to image
- Reference images: up to 10
- Aspect ratios: 1:1, 4:3, 3:4, 3:2, 2:3, 16:9, 9:16, 21:9, 9:21, auto
- Resolutions: 1K, 2K, 4K
- Outputs per run: 1 to 4
- Credits: 2 per image (1K) to 6 per image (4K)

### Our test runs

- Image to image, 3:4, 1K, 1 output, 1 reference images: 2 credits, result in 24 s (run on 2026-11-10)
  - Prompt: "Put this same woman on a rainy Tokyo side street at night, neon signs and paper lanterns behind her, wearing a black raincoat and holding a clear umbrella. Keep her face, hair, freckles and nose stud exactly the same. Candid photo, shallow depth of field."
  - Result: Mid shot on a wet street; the face is close to hers, the stud too small to see. Input sent as a JPEG copy of the selfie.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/yourself/me-tokyo-klein.webp
- Image to image, 3:4, 1K, 1 output, 1 reference images: 2 credits, result in 24 s (run on 2026-11-10)
  - Prompt: "Turn this into an autumn park photoshoot of the same woman: golden leaves, a camel wool coat and a cream scarf, sitting on a wooden bench in warm late-afternoon light. Keep her face, hair, freckles and nose stud exactly the same."
  - Result: Whole bench in view; the face came out slimmer than hers. Input sent as a JPEG copy of the selfie.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/yourself/me-autumn-klein.webp
- Image to image, 3:4, 1K, 1 output, 1 reference images: 2 credits, result in 31 s (run on 2026-11-10)
  - Prompt: "Make this a 1990s high school yearbook portrait of the same woman: mottled blue studio backdrop, soft flash, slight film grain, a denim jacket over a white tee. Keep her face, hair, freckles and nose stud exactly the same."
  - Result: Smoother skin; the face is close to hers, the stud missing. Input sent as a JPEG copy of the selfie.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/yourself/me-yearbook-klein.webp
- Text to image, 16:9, 1K, 1 output: 2 credits, result in 24 s (run on 2026-11-10)
  - Prompt: "An anime character sheet of a young woman bike courier with short silver hair, a red bomber jacket, goggles pushed up on her forehead and a messenger bag: front, side and back views standing in a row, plus three small face close-ups (happy, surprised, angry). Clean line art, flat cel shading, plain white background."
  - Result: All three views, but the three faces are squeezed beside the boots and one overlaps a leg.
  - Image: https://cdn.qwenimagex.com/assets/landing/aipicturegenerator/best/best-anime-klein.webp

## Ideogram 4.5

- Modes: Text to image, Image to image
- Reference images: up to 5
- Aspect ratios: 1:1, 3:4, 4:3, 9:16, 16:9
- Outputs per run: 1 to 4
- Credits: 6 per image (low) to 50 per image (high)

## Veo 3.1

- Modes: Text to video, Image to video
- Image to video inputs: first frame (required), last frame (optional)
- Aspect ratios: 16:9, 9:16, auto
- Resolutions: 720p, 1080p, 4k
- Credits: Text to video: 250 per clip (720p) to 380 per clip (4k); Image to video: 250 per clip (720p) to 370 per clip (4k)

## Veo 3.1 Fast

- Modes: Text to video, Image to video, Video to video
- Image to video inputs: first frame (required), last frame (optional)
- Reference images: up to 3
- Aspect ratios: 16:9, 9:16, auto
- Resolutions: 720p, 1080p, 4k
- Credits: 60 per clip (720p) to 180 per clip (4k)

### Our test runs

- Text to video, 16:9, 720p: 60 credits, result in 68 s (run on 2026-10-21)
  - Prompt: "A tall guy in a grocery store reaches for the last box of cereal on the top shelf — just as a woman on her tiptoes stretches for the same box from the other side. They both grab it at the same time, look at each other, then burst out laughing. He gestures "go ahead." She hesitates, then splits it — hands him half. Handheld camera, fluorescent supermarket lighting, natural and candid feel."
  - Result: An 8-second 720p clip: a man and a woman reach for the same cereal box on a top shelf, laugh, and each end up holding a half-open box in the aisle. The box packaging shows garbled text. Audio was generated but not reviewed.
  - Video: https://cdn.qwenimagex.com/assets/landing/voo31/veo-3-1-lite-vs-fast/fast-cereal.mp4

## Veo 3.1 Lite

- Modes: Text to video, Image to video
- Image to video inputs: first frame (required), last frame (optional)
- Aspect ratios: 16:9, 9:16, auto
- Resolutions: 720p, 1080p, 4k
- Credits: 30 per clip (720p) to 150 per clip (4k)

## Gemini Omni 1.2 Flash

- Modes: Text to video, Image to video, Reference to video, Video to video
- Image to video inputs: first frame (required), last frame (optional)
- Reference images: up to 6
- Reference videos: up to 1
- Aspect ratios: 16:9, 9:16
- Resolutions: 360p, 720p, 1080p, 4k
- Duration: 3 to 15 seconds
- Credits: 6 per second (360p) to 50 per second (4k)

## Seedance 2

- Modes: Text to video, Image to video, Video to video
- Image to video inputs: first frame (required), last frame (optional)
- Reference images: up to 9
- Reference videos: up to 3
- Reference audio: up to 3
- Aspect ratios: 1:1, 4:3, 3:4, 16:9, 9:16, 21:9, adaptive
- Resolutions: 480p, 720p, 1080p, 4k
- Duration: 4 to 15 seconds
- Credits: Text to video, Image to video: 20 per second (480p) to 210 per second (4k); Video to video: 15 per second (480p, with reference_video_urls) to 190 per second (4k, without reference_video_urls)

## Seedance 2 Fast

- Modes: Text to video, Image to video, Video to video
- Image to video inputs: first frame (required), last frame (optional)
- Reference images: up to 9
- Reference videos: up to 3
- Reference audio: up to 3
- Aspect ratios: 1:1, 4:3, 3:4, 16:9, 9:16, 21:9, adaptive
- Resolutions: 480p, 720p
- Duration: 4 to 15 seconds
- Credits: Text to video, Image to video: 15 per second (480p) to 25 per second (720p); Video to video: 7 per second (480p, with reference_video_urls) to 25 per second (720p, without reference_video_urls)

## Seedance 2 Mini

- Modes: Text to video, Image to video, Video to video
- Image to video inputs: first frame (required), last frame (optional)
- Reference images: up to 9
- Reference videos: up to 3
- Reference audio: up to 3
- Aspect ratios: 1:1, 4:3, 3:4, 16:9, 9:16, 21:9, adaptive
- Resolutions: 480p, 720p
- Duration: 4 to 15 seconds
- Credits: Text to video, Image to video: 4 per second (480p) to 9 per second (720p); Video to video: 3 per second (480p, with reference_video_urls) to 8 per second (720p, without reference_video_urls)

## Grok Imagine Video 1.5

- Modes: Text to video, Image to video
- Image to video inputs: first frame (required)
- Aspect ratios: 1:1, 16:9, 9:16, 3:2, 2:3, auto
- Resolutions: 480p, 720p
- Duration: 1 to 15 seconds
- Credits: 3 per second (480p) to 5 per second (720p), minimum 3 per run

## MiniMax H3 Max

- Modes: Text to video, Image to video, Reference to video, Video to video
- Image to video inputs: first frame (required), last frame (optional)
- Reference images: up to 9
- Reference videos: up to 3
- Reference audio: up to 3
- Aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, adaptive, auto
- Resolutions: 480P, 768P, 1080P, 2K
- Duration: 5 to 15 seconds
- Outputs per run: 1 to 4
- Credits: Text to video, Image to video, Reference to video: 10 per second (480P) to 35 per second (1080P); Video to video: 20 per second (480P) to 140 per second (2K)

## MiniMax H3 Max Turbo

- Modes: Text to video, Image to video
- Image to video inputs: first frame (required), last frame (optional)
- Aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16
- Resolutions: 480P, 768P, 1080P
- Duration: 2 to 15 seconds
- Outputs per run: 1 to 4
- Credits: 5 per second (480P) to 20 per second (1080P)

## Wan 3.0

- Modes: Text to video, Image to video, Video to video, Reference to video
- Image to video inputs: first frame (required), last frame (optional)
- Reference images: up to 10
- Reference videos: up to 5
- Reference audio: up to 5
- Aspect ratios: adaptive, 16:9, 4:3, 1:1, 3:4, 9:16
- Resolutions: 480P, 720P, 1080P
- Duration: 2 to 30 seconds
- Credits: Text to video, Image to video: 8 per second (480P) to 30 per second (1080P), minimum 16 per run; Video to video: 8 per second (480P, with reference_video_urls) to 30 per second (1080P, without reference_video_urls), minimum 16 per run; Reference to video: 8 per second (480P) to 30 per second (1080P), minimum 16 per run

### Our test runs

- Reference to video, 480P, 16:9, 10 s, 1 reference image urls: 80 credits, result in 141 s (run on 2026-11-05)
  - Prompt: "Widescreen 16:9 dance video on a cruise ship at golden hour, in the style of the 24k Magic AI dance trend: warm low sunlight, blue ocean behind white railings and white stairs, a warm wood deck. Image1 is the main dancer and Image2 makes a short cameo at the end; keep each person's face, hair and clothes from their image the whole time. 0-2 s: close-up of Image1 in a warm, softly lit ship lounge, smiling into the camera and tapping their chest to the beat. 2-6 s: low wide full-body shot on the sunny deck, Image1 dances alone with knee bends, sliding steps and shoulder bounces. 6-8 s: a shaded corridor that opens onto the sea, Image2 leans out from behind a door frame and dances in the doorway while Image1 keeps dancing far behind at the railing. 8-10 s: close-up of Image2 touching the top of their head and grinning. Exactly two people, five fingers on each hand, no captions or on-screen text."
  - Result: A 10-second 16:9 clip with hard cuts at about 2, 6 and 8 s. Both faces, hair and tops stay right (red bob, round glasses and mustard sweater; bald, bearded, green jacket over a white tee). Close-up of her in a warm ship lounge pointing at the camera (not the chest taps asked); a wide shot of her dancing alone on a sunny deck by the railing, with no stairs and an eye-level camera; a doorway shot where he stands and smiles with her dancing far behind at the railing (he does not lean out or dance much); a close-up of him touching his head and grinning. Jeans and white sneakers were made up below the chest. Audio was generated but not reviewed.
  - Video: https://cdn.qwenimagex.com/assets/landing/magic24kai/guides/prompt-test-wan.mp4
- Reference to video, 480P, 9:16, 10 s, 1 reference image urls: 80 credits, result in 160 s (run on 2026-11-04)
  - Prompt: "Vertical 9:16 music performance video in the style of a COLORS session: a seamless teal studio with one vintage silver microphone hanging from the ceiling on a long black cable, soft even light. Image1 raps first and Image2 takes the verse; keep each person's face, hair and clothes from their image the whole time. 0-4 s: waist-up two-shot, Image1 raps the hook toward the camera with hand gestures while Image2 stands beside them, nodding along with mouth closed. 4-5 s: Image2 turns to face Image1. 5-8 s: close-up of Image2 rapping the verse into the hanging microphone. 8-10 s: wide full-body shot of both on a low teal platform, Image2 still rapping while Image1 dances and nods along, impressed. Exactly two people and one microphone, five fingers on each hand, no captions or on-screen text."
  - Result: A 10-second 9:16 clip in a teal studio with one silver vintage mic on a long black cable. Both faces, hair and tops stay right (red bob, round glasses and mustard sweater; bald, bearded, green jacket over a white tee). Waist-up two-shot with her talking to the camera and gesturing; he turns toward her at about 3.5-4 s; a close-up of him at the mic from about 6 s (about a second later than asked); a wide full-body shot of both on a round teal platform from about 8 s, where the mic hangs far above their heads and she stands and smiles instead of dancing. Jeans, black trousers and white trainers were made up below the chest. Audio was generated but not reviewed.
  - Video: https://cdn.qwenimagex.com/assets/landing/ai58thfloor/guides/prompt-test-wan.mp4
- Image to video, 480P, adaptive, 5 s, first frame + last frame: 40 credits, result in 82 s (run on 2026-10-31)
  - Prompt: "The young man's calm face slowly breaks into a warm laugh as he lifts the coffee cup off the saucer and looks into the camera. Static camera. Soft cafe ambience, quiet chatter, a light clink of the cup."
  - Result: The same man and cafe throughout; he looks up, smiles and lifts the cup to chest height. The move between the two frames reads as one natural motion with no morphing. Audio was generated but not reviewed.
  - Video: https://cdn.qwenimagex.com/assets/landing/wan30ai/frames-to-video/small.mp4
- Image to video, 480P, adaptive, 5 s, first frame + last frame: 40 credits, result in 82 s (run on 2026-10-31)
  - Prompt: "She stands still on the sidewalk while her beige trench coat, jeans and sneakers change into a long red satin evening dress in one smooth shimmer that sweeps from her shoulders down to her feet. Static camera. Quiet street ambience and a soft whoosh as the fabric changes."
  - Result: Her face, hair, pose and the street stay put. The outfit swap is bridged by a golden glow that runs over her body from the shoulders down, then settles on the red dress; it reads as an effect, not a real costume change. Audio was generated but not reviewed.
  - Video: https://cdn.qwenimagex.com/assets/landing/wan30ai/frames-to-video/medium.mp4

## Kling 4.0

- Modes: Text to video, Image to video, Reference to video
- Image to video inputs: first frame (required), last frame (optional)
- Reference images: up to 10
- Aspect ratios: 16:9, 9:16, 1:1, 21:9, auto
- Resolutions: 720p, 1080p
- Duration: 3 to 30 seconds
- Credits: 30 per second (720p) to 40 per second (1080p), minimum 80 per run

## Kling 4.0 Flash

- Modes: Video to video
- Reference images: up to 4
- Reference videos: up to 1
- Resolutions: 720p
- Credits: 15 per second

## FLUX 3

- Modes: Text to video, Image to video, Video to video
- Reference images: up to 10
- Reference videos: up to 1
- Aspect ratios: auto, 21:9, 2:1, 16:9, 4:3, 1:1, 3:4, 9:16
- Resolutions: 720p, 1080p
- Duration: 5 to 20 seconds
- Credits: Text to video, Image to video: 35 per second (720p) to 60 per second (1080p), minimum 175 per run; Video to video: 6 per second, minimum 12 per run

### Our test runs

- Text to video, 16:9, 720p, 5 s, audio on: 175 credits, result in 77 s (run on 2026-10-20)
  - Prompt: "Night market food stall, handheld medium shot. An older street cook tosses noodles in a flaming wok; the flames flare up and light her face. She looks up at a customer just off camera and says with a grin: "Two minutes. Best noodles on this street." Sizzling wok, crowd chatter, a distant scooter horn. Warm neon signs and drifting steam."
  - Result: An open-air night street with the crowd and scooters on screen. Her toss is gentler, closer to the prompt's wording, and she ends on a grin. Audio was generated but not reviewed.
  - Video: https://cdn.qwenimagex.com/assets/landing/flux3max/compare/pair1-flux3.mp4
- Text to video, 16:9, 720p, 5 s, audio on: 175 credits, result in 62 s (run on 2026-10-20)
  - Prompt: "Slow 180-degree orbit around a pair of matte black over-ear headphones resting on a concrete pedestal in a minimal studio. Soft key light from the left, a thin rim light traces the ear cups, dust drifts through the beam. The orbit ends front-on as the light warms slightly. A low ambient synth hum."
  - Result: The camera covers the full half circle, from side profile to front-on, but the light stays cool and neutral instead of warming. Audio was generated but not reviewed.
  - Video: https://cdn.qwenimagex.com/assets/landing/flux3max/compare/pair2-flux3.mp4

## Tools

- Background Remover (image): 1 credit per run
- Image Extender (image): 2 credits per run
- Image Upscaler (image): 2 credits (2x) or 4 credits (4x) or 8 credits (8x) per run
- Photo Restore (image): 6 credits per run
- Motion Control (video): 20 credits (720p) or 40 credits (1080p) per second of video, minimum 45 credits per run
- Lip-Sync Video (video): 5 credits (720p) or 10 credits (1080p) per second of video, minimum 15 credits per run
- Video Upscaler (video): 2 credits (1080p) or 3 credits (2k) or 6 credits (4k) per second of video