
The Highspeed serving mode of MiniMax's open-weight H3-Image model, editing and composing images from 1-16 reference images and a natural-language instruction at 3-4x the speed of mainstream alternatives for latency-sensitive workflows.

The Highspeed serving mode of MiniMax's open-weight H3-Image model, turning a text prompt into 1-10 natively high-resolution images with state-of-the-art instruction following and precise text rendering at 3-4x the speed of mainstream alternatives.

MiniMax's open-weight H3-Image model edits and composes images from 1-16 reference images and a natural-language instruction, returning 1-10 natively high-resolution results that keep people, products and layout consistent, at the cost-efficient Standard rate.

MiniMax's open-weight H3-Image model turns a text prompt into 1-10 natively high-resolution images (1K to about 20 MP, aspect ratios 1:3 to 3:1) with state-of-the-art instruction following and precise long-text rendering, at the cost-efficient Standard rate.

GPT Image 2.6 Sunburst generates images from natural-language prompts with arbitrary resolutions up to 3840x2160, five quality tiers including xhigh and max, and first-class transparent backgrounds. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

GPT Image 2.6 Sunburst Edit applies natural-language instructions to up to 16 reference images, with an optional mask, arbitrary resolutions up to 3840x2160, five quality tiers including xhigh and max, and transparent-background output. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

GPT Image 2.6 Flare generates images from natural-language prompts with arbitrary resolutions up to 3840x2160, five quality tiers including xhigh and max, and first-class transparent backgrounds. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

GPT Image 2.6 Flare Edit applies natural-language instructions to up to 16 reference images, with an optional mask, arbitrary resolutions up to 3840x2160, five quality tiers including xhigh and max, and transparent-background output. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Seedance 2.5 Flash Reference-to-Video generates, edits, or extends 4-30s videos with synchronized audio at 480p or 720p from up to 10 reference images, 5 videos, and 5 audios, at about 2x the speed and lower cost of Seedance 2.5.

Seedance 2.5 Flash Image-to-Video animates a first-frame image (with optional last frame) into 4-30s videos with synchronized audio at 480p or 720p, delivering near-flagship quality about 2x faster and roughly 34% cheaper than Seedance 2.5.

Seedance 2.5 Flash Text-to-Video turns text prompts into 4-30s videos with synchronized audio at 480p or 720p, delivering near-flagship quality about 2x faster and roughly 34% cheaper than Seedance 2.5.

GPT Image 2.5 Sunburst generates images from natural-language prompts with arbitrary resolutions up to 3840x2160, five quality tiers including xhigh and max, and first-class transparent backgrounds. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

GPT Image 2.5 Sunburst Edit applies natural-language instructions to up to 16 reference images, with an optional mask, arbitrary resolutions up to 3840x2160, five quality tiers including xhigh and max, and transparent-background output. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

GPT Image 2.5 Flare generates images from natural-language prompts with arbitrary resolutions up to 3840x2160, five quality tiers including xhigh and max, and first-class transparent backgrounds. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

GPT Image 2.5 Flare Edit applies natural-language instructions to up to 16 reference images, with an optional mask, arbitrary resolutions up to 3840x2160, five quality tiers including xhigh and max, and transparent-background output. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Multimodal video generation from reference images, videos, and audio. Supports video editing and extension.

Generate videos from text prompts with native audio and optional web search.

Bach 1.0 Pro text-to-video: generate a 5-16 second video with optional audio from a text prompt, at 1080p or 4K.

Bach 1.0 Pro image-to-video: generate a 5-16 second video from a first frame, a last frame, or both, at 1080p or 4K.

Bach 1.0 Pro reference-to-video: generate a 5-16 second video guided by up to 9 reference images, 3 reference videos and 3 reference audios, at 1080p or 4K.

MiniMax H3 Max text-to-video: generate a cinematic video from a text prompt. Supports 480P、768P, 5-15s., and 16:9/9:16/1:1/adaptive aspect ratios.

MiniMax H3 Max image-to-video: animate a first-frame image (optionally with a last frame) driven by a text prompt. Supports 480P、768P, 5-15s.

MiniMax H3 Fast text-to-video: generate a cinematic video from a text prompt. Supports 480P, 5-15s., and 16:9/9:16/1:1/adaptive aspect ratios.

MiniMax H3 Fast image-to-video: animate a first-frame image (optionally with a last frame) driven by a text prompt. Supports 480P, 5-15s.

MiniMax H3 Fast reference-to-video: generate a video that keeps the subject from a reference image, driven by a text prompt. Supports 480P, 5-15s.

A natively multimodal Google DeepMind model that continues an existing clip with a seamlessly matched 3-to-10-second extension, chainable to grow a single shot up to a total of 40 seconds of coherent video with native audio.

A natively multimodal Google DeepMind model that applies a text-instructed edit to an existing video - adding, removing, replacing, or restyling elements with native audio - while preserving everything the prompt does not mention.

A natively multimodal Google DeepMind model that generates cinematic, natively sound-enabled videos from a text prompt plus up to 10 reference images and 3 reference video clips, keeping a character, product, or art direction consistent across generations.

A natively multimodal Google DeepMind model that animates a still image into a cinematic, natively sound-enabled clip from a text prompt, optionally interpolating to a supplied last frame for precise shot start and end control.

A natively multimodal Google DeepMind model that turns a single text prompt into a cinematic clip with synchronized native audio, with control over duration, aspect ratio, and output resolution from a fast 360p draft up to 4K.

MiniMax H3-Developer self-hosted text-to-video: generate a video (with audio) from a text prompt. Supports 480P/768P/2K, 16:9/9:16/1:1 aspect ratios.

MiniMax H3-Developer self-hosted image-to-video: animate a first-frame image (optionally a last frame) driven by a text prompt, with generated audio. Supports 480P/768P/2K.

MiniMax H3-Developer self-hosted reference-to-video: generate a video that keeps the subject from one or more reference images/videos, driven by a text prompt, with generated audio. Supports 480P768P/2K.

ByteDance Seedream 4.7 image editing model with batch generation support. Produce a coherent set of edited images from reference inputs.

ByteDance Seedream 4.7 image editing model. Executes edit instructions precisely while preserving identity, lighting and local structure of the source image.

ByteDance Seedream 4.7 with batch generation support. Generate a set of coherent images in a single request.

ByteDance Seedream 4.7 image generation model. Balanced gains in image quality, aesthetics and instruction following, at the efficiency and cost profile of the 4.0 generation.

All-in-one Wan3.0 renderer: cinematic, hyper-real video from a text prompt, up to 30s with smart-duration and adaptive aspect ratio.

Animate a first frame (optionally with a last frame) into a coherent clip, with native audio and smart-duration up to 30s.

All-in-One Reference: keep subjects consistent from any mix of reference images, videos, and audio; pixel-level identity/voice/space alignment.

The fast, low-cost member of the MAI-Image-2.6 family, delivering the same photorealistic text-to-image quality as the flagship at less than half the cost for latency-sensitive, high-throughput production.

The fast, low-cost member of the MAI-Image-2.6 family, delivering the same instruction-driven editing and multi-reference composition as the flagship at less than half the cost for latency-sensitive, high-volume workloads.

Microsoft AI's flagship text-to-image model, generating photorealistic, design-ready images from natural language with markedly improved in-image text rendering, model-directed aspect ratio selection, and optional web-grounded context.

Microsoft AI's flagship image-to-image editing model, combining surgical instruction-driven edits with multi-reference composition across up to five images, ranked No. 1 for image editing on Artificial Analysis.

All-in-one Wan3.0 renderer: cinematic, hyper-real video from a text prompt, up to 30s with smart-duration and adaptive aspect ratio.

Animate a first frame (optionally with a last frame) into a coherent clip, with native audio and smart-duration up to 30s.

All-in-One Reference: keep subjects consistent from any mix of reference images, videos, and audio; pixel-level identity/voice/space alignment.

Render the 1080p final of a completed Seedance 2.5 draft. Prompt, inputs, duration, ratio, seed and audio come from the draft; only draft_id, return_last_frame, output_format and watermark are accepted.