MiniMax H3 + H3 Turbo

About

MiniMax H3 is a next-generation, open-weights, general-purpose multimodal video model—not a Seedance derivative—with short-form capabilities comparable to Seedance 2.0. On Sogni, H3 supports opt-in mature-content workflows while remaining subject to Sogni's Terms of Use and MiniMax's license and safety requirements. It creates the moving image and stereo soundtrack in one generation, so you can direct the action, camera, dialogue, ambience, sound effects, and music in the same prompt and get a complete audiovisual shot instead of a silent clip that still needs a soundtrack.

Choose how much control and speed you want. Start from text for a blank canvas, animate a starting image when the composition is already set, provide opening and closing frames when the shot must arrive at a specific visual, or condition a scene on labelled image, video, and audio references. H3 Turbo adds 4-step text-to-video, image-to-video, and first-and-last-frame workflows; Ref2VA remains a standard 20-step workflow.

Use H3 Turbo for drafts and rapid prompt iteration. Sogni's warm reference runs measured about 3.5× faster image-to-video and 3.7× faster text-to-video, conservatively summarized as about 3.5× faster; actual time varies by mode, input, clip length, resolution, worker hardware, and queue state. The upstream v0.1 release is a preview and may trade some fine visual detail and audio polish for speed, so standard H3 remains the quality-first choice.

On Sogni, H3 produces roughly 5–15 seconds of 768p-class video at 24 fps with 32 kHz stereo audio. Sogni Web offers six square, landscape, cinema, and portrait presets; direct SDK/API jobs accept a 32 px grid up to 1344 px per axis within the model's pixel budget. MiniMax reports stable dialogue support in 11 languages.

All seven current H3 modes are available through the Sogni API and Creative Agent Skill: four standard workflows (text-to-video, image-to-video, first-and-last-frame, and Ref2VA) plus three Turbo workflows (text-to-video, image-to-video, and first-and-last-frame). Sogni Web currently exposes standard text-to-video, image-to-video, and Ref2VA. Ref2VA supports up to 9 images, 3 videos, and 3 audio clips (12 files total, with at least one visual reference).

Choose your H3 workflow and speed

Use standard 20-step H3 when detail and audio polish come first, or a 4-step Turbo FL2VA workflow for about 3.5× faster iteration. Ref2VA multi-reference video currently has no Turbo variant.

Prompting tips

For Creative Agent, write a compact shot brief: what we see, what changes, how the camera moves, and what we hear. The agent expands it into H3's production prompt. Direct SDK users should follow MiniMax's structured prompt guide.

• Build one strong moment — Give the clip a clear subject, action, setting, mood, and final beat instead of cramming in an entire story.
• Time a multi-beat shot — For a sequence, use short ranges such as 0–3 seconds and 3–7 seconds so each action, cut, or camera change has a place.
• Direct the camera — Name the framing and one deliberate move—a slow push-in, orbit, pan, tracking shot, handheld drift, or locked-off composition.
• Write the soundtrack — Describe dialogue, voice character, music, ambience, and synchronized sound effects alongside the visual action.
• Keep dialogue performable — Put spoken lines in quotation marks, identify the speaker, and keep the word count realistic for a 5–15 second clip.
• Connect opening and closing frames — Use one continuous shot when possible, and describe the visible motion that carries the composition from the opening image to the closing image.
• Protect your anchors — For products, characters, posters, and UI, explicitly preserve identity, shape, brand colors, layout, or typography while describing what may move.
• Assign every reference a job — For Ref2VA, refer to uploads as <Picture 1>, <Video 1>, or <Audio 1> and state exactly which identity, motion, location, voice, or sound each reference controls.

Read the Sogni Engineering field guide, Your Prompt Is Now a Director, for side-by-side H3 and H3 Turbo tests across text-to-video, image-to-video, first-and-last-frame, and reference-to-video workflows.

Standard H3 or 4-step H3 Turbo?

Choose standard 20-step H3 when fine visual detail and audio polish matter most. Choose H3 Turbo for drafts, prompt iteration, timing tests, and other work where turnaround matters most. Sogni's warm reference runs measured about 3.5× faster image-to-video and 3.7× faster text-to-video, conservatively summarized as about 3.5× faster. Actual time varies with workflow, input conditioning, clip length, resolution, worker hardware, and queue state.

LightX2V labels the current v0.1 Turbo weights a preview and notes that image detail still needs improvement. Turbo is a speed/quality trade-off, not a higher-quality replacement for standard H3.

Make a finished beat, not a silent motion test

MiniMax H3 generates picture and synchronized stereo sound together: dialogue, ambience, effects, and music can all be directed in the same brief. It is useful for short brand films, dialogue scenes, product and fashion clips, animated posters and titles, motion design, music visuals, and stylized social video.

Choose the control your shot needs

Start from text when the scene is still an idea. Supply an opening frame when identity, product shape, palette, or composition is established. Add a closing frame when a transition must land at a specific visual. Standard Ref2VA conditions a scene on labelled uploads such as <Picture 1>, <Video 1>, and <Audio 1>; it accepts up to nine images, three videos, and three audio clips, with no more than 12 files total and at least one visual reference. Turbo currently has no Ref2VA variant.

A general-purpose multimodal model—not a derivative

MiniMax H3 is MiniMax's next-generation open-weights model, with a short-form multimodal scope comparable to Seedance 2.0. Sogni supports opt-in mature-content generation with H3; "uncensored" describes that wider adult creative scope, not an absence of rules. Sogni's Terms of Use and the applicable MiniMax and LightX2V licenses still apply.

Clear deployment boundaries

Sogni Web supports standard H3 text-to-video, image-to-video, and Ref2VA reference-to-video. The Sogni API and Creative Agent Skill additionally support standard first-and-last-frame video and the exact Turbo t2v, i2v, and flf2v IDs listed above. Sogni generates 768p-class output and does not expose MiniMax's hosted editing or 2K regeneration stages.

Flat per-second pricing

Standard H3 costs 16 Spark ($0.08) per second of finished video; H3 Turbo costs 6 Spark ($0.03) per second — about 2.67× cheaper. Within each speed tier, duration is the only price factor: layout, workflow, and reference count do not change the rate. The 24 fps frame grid produces billable clips from roughly 5.17 seconds (82.67 Spark / $0.41 standard, 31 Spark / $0.155 Turbo) to 15.08 seconds (241.33 Spark / $1.21 standard, 90.5 Spark / $0.4525 Turbo).

Sources and licenses

Review the official MiniMax H3 model card, MiniMax H3 Community License, LightX2V H3 Turbo model card, and LightX2V Turbo implementation. Sogni has received written authorization from MiniMax to offer H3 through the platform.

Pricing

Standard MiniMax H3 is billed at a flat 16 Spark ($0.08) per second of finished video; H3 Turbo is 6 Spark ($0.03) per second — about 2.67× cheaper. Within each speed tier, cost scales with clip duration only — layout and workflow do not change the per-second rate, and references are never charged extra. Use pay-as-you-go Spark, or eligible fair-use coverage on an active Unlimited plan.

API

Start with a natural creative brief. Creative Agent expands it into the production prompt MiniMax H3 needs and runs the matching workflow.

const response = await fetch('https://api.sogni.ai/v1/chat/completions', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
    Authorization: `Bearer ${process.env.SOGNI_API_KEY}`,
  },
  body: JSON.stringify({
    messages: [{ role: 'user', content: "Create an 8-second MiniMax H3 video from this brief: A ceramic artist opens a glowing kiln as the camera slowly pushes in; fire crackles, tools clink softly, and a restrained string score begins" }],
    sogni_tools: 'creative-agent',
    sogni_tool_execution: true,
  }),
});

if (!response.ok) throw new Error(await response.text());
const { choices } = await response.json();
console.log(choices[0].message.content);

curl https://api.sogni.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $SOGNI_API_KEY" \
  -d '{
    "messages": [{ "role": "user", "content": "Create an 8-second MiniMax H3 video from this brief: A ceramic artist opens a glowing kiln as the camera slowly pushes in; fire crackles, tools clink softly, and a restrained string score begins" }],
    "sogni_tools": "creative-agent",
    "sogni_tool_execution": true
  }'

Creative Agent accepts the short brief above and prepares MiniMax H3's structured production prompt. Calling a worker directly? Follow the MiniMax H3 integration guide and use one of these exact model IDs: minimax-h3-fl2va-fp8_t2v, minimax-h3-fl2va-fp8_i2v, minimax-h3-fl2va-fp8_flf2v, minimax-h3-ref2va-fp8_r2v, minimax-h3-fl2va-fp8_t2v_turbo, minimax-h3-fl2va-fp8_i2v_turbo, minimax-h3-fl2va-fp8_flf2v_turbo. Full reference at docs.sogni.ai.