Source: https://docs.sogni.ai/models/minimax-h3/ # MiniMax H3 and H3 Turbo video Sogni runs the open-weight **MiniMax H3 FL2VA** model through ComfyUI for text-to-video, first-frame image-to-video, and first/last-frame video, plus the separate **MiniMax H3 Ref2VA** checkpoint for multi-reference video. For FL2VA, **FastH3 Turbo** is the fastest choice. It runs the [FastVideo FastH3 4-step Preview v1 VSA DataFree](https://huggingface.co/FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree) checkpoint and renders up to **2× faster than the LightX2V 4-step Turbo** and up to **6× faster than Standard** for comparable 768p, 15-second requests. FastH3 supports the same custom LoRA catalog without changing engines. Base jobs can run on 24 GB-class GPUs, including the **NVIDIA GeForce RTX 4090 and RTX 3090**, when at least 23 GB of usable VRAM is available; jobs with a custom LoRA route to workers with at least 32 GB. The familiar four-step **LightX2V Turbo** remains available as a distinct creative look, alongside eight-step Balanced and maximum-quality Standard. Video and synchronized stereo audio are generated together in one pass. Sogni has written authorization from MiniMax to offer MiniMax H3 through the Sogni platform. MiniMax H3 is a separate MiniMax model, not a Seedance derivative. Its short-form multimodal scope is comparable to Seedance 2.0, while the Sogni deployment uses released H3 weights on Sogni Supernet workers. ## [#](https://docs.sogni.ai/models/minimax-h3/#workflow-ids)Workflow IDs | Mode | Model ID | Inputs | | --- | --- | --- | | Text to video | `minimax-h3-fl2va-fp8_t2v` | Prompt | | Image to video | `minimax-h3-fl2va-fp8_i2v` | Prompt + first frame | | First/last frame | `minimax-h3-fl2va-fp8_flf2v` | Prompt + first and last frames | | Reference to video | `minimax-h3-ref2va-fp8_r2v` | Prompt + 0–9 images, up to 3 videos and 3 audio clips (12 files total; at least one image or video) | | Balanced text to video | `minimax-h3-fl2va-fp8_t2v_balanced` | Prompt | | Balanced image to video | `minimax-h3-fl2va-fp8_i2v_balanced` | Prompt + first frame | | Balanced first/last frame | `minimax-h3-fl2va-fp8_flf2v_balanced` | Prompt + first and last frames | | Balanced reference to video | `minimax-h3-ref2va-fp8_r2v_balanced` | Prompt + references as above | | LightX2V Turbo text to video | `minimax-h3-fl2va-fp8_t2v_turbo` | Prompt | | LightX2V Turbo image to video | `minimax-h3-fl2va-fp8_i2v_turbo` | Prompt + first frame | | LightX2V Turbo first/last frame | `minimax-h3-fl2va-fp8_flf2v_turbo` | Prompt + first and last frames | | LightX2V Turbo reference to video | `minimax-h3-ref2va-fp8_r2v_turbo` | Prompt + references as above | | FastH3 Turbo text to video | `minimax-h3-fastvideo-int8_t2v_turbo` | Prompt | | FastH3 Turbo image to video | `minimax-h3-fastvideo-int8_i2v_turbo` | Prompt + first frame | | FastH3 Turbo first/last frame | `minimax-h3-fastvideo-int8_flf2v_turbo` | Prompt + first and last frames | ## [#](https://docs.sogni.ai/models/minimax-h3/#product-availability)Product availability The **Sogni API** accepts every exact model ID in the table. **Sogni Web** exposes Standard, Balanced, LightX2V Turbo, and FastH3 Turbo for text-to-video, image-to-video, and first/last-frame creation. Reference-to-Video exposes its dedicated Standard, Balanced, and LightX2V Turbo modes; FastH3 does not support Reference-to-Video. The **[Sogni Creative Agent Skill](https://docs.sogni.ai/sogni-intelligence/creative-agent-skill/)** (3.39.0 and later) exposes friendly selectors for Standard, Balanced, LightX2V Turbo, FastH3 Turbo, and dedicated Ref2VA workflows. Its `minimax-h3-turbo` selectors continue to mean LightX2V Turbo; use `minimax-h3-fasth3-turbo`, or the explicit `minimax-h3-fasth3-t2v-turbo`, `minimax-h3-fasth3-i2v-turbo`, and `minimax-h3-fasth3-flf2v-turbo`, for FastH3. FastH3 has no r2v selector. ## [#](https://docs.sogni.ai/models/minimax-h3/#which-speed-should-i-choose)Which speed should I choose? Choose **FastH3 Turbo** for the quickest drafts, prompt iteration, timing tests, and other work where turnaround matters most. It renders up to twice as fast as LightX2V Turbo and up to 6× faster than standard 544/768p H3 on a 15-second clip in Sogni's qualified production configuration (shorter clips see a smaller multiplier), and it accepts the same custom LoRA catalog on workers with at least 32 GB of VRAM. Sogni Web uses FastH3 as the default Turbo engine, with a FastH3 switch under Turbo that returns to LightX2V. Choose **LightX2V Turbo** when you prefer its familiar four-step look. The model you select always determines the engine; attaching or removing a LoRA does not switch it. Actual wall-clock time varies with mode, input conditioning, clip length, resolution, worker hardware, and queue state. Choose **Balanced** for an eight-step middle ground, or standard **20-step H3** when fine visual detail and audio polish matter more than iteration speed. At 544/768p-class output, LightX2V Turbo is 6 Spark per second and FastH3 is 4 Spark per second, against Standard H3's 16 — see [Pricing](https://docs.sogni.ai/models/minimax-h3/#pricing). On Sogni's primary **NVIDIA RTX PRO 6000** H3 workers, a highest-quality 15-second standard H3 video typically takes about **15 minutes** once rendering starts; FastH3 Turbo brings the same clip to roughly **two minutes**. Raw render time is the same on Unlimited and Unlimited Pro. Pro changes how much work can run and where it sits in the subscription queue, not the speed of one GPU render. ## [#](https://docs.sogni.ai/models/minimax-h3/#what-fasth3-is)What FastH3 is FastH3 Turbo runs **FastVideo FastH3 4-step Preview v1 VSA DataFree**, the recommended FastH3 Preview v1 checkpoint from the [FastVideo](https://github.com/hao-ai-lab/FastVideo) project at Hao AI Lab. It is a distillation of MiniMax H3 that generates synchronized video and audio in four transformer forwards instead of standard H3's 20 steps. _VSA_ is FastVideo's video sparse attention (the checkpoint was trained with VSA-H3 at 90% sparsity), and _DataFree_ means the distillation used data-free DMD2 rather than an external training set. Sogni runs Kijai's INT8 ConvRot conversion of the step-1300 checkpoint (about 21 GB), which is what fits the 24 GB-class worker pool. The checkpoint inherits the MiniMax H3 Community License. The [Hao AI Lab FastH3 preview post](https://haoailab.com/blogs/fasth3-preview/) covers the training details. ## [#](https://docs.sogni.ai/models/minimax-h3/#generation-limits)Generation limits - Fixed **24 fps** with native 32 kHz stereo audio. - **124–362 frames** on the `124 + n×17` frame grid, approximately 5.17–15.08 seconds. - Size presets cover **544/768p-class** output. The **480p lower tier** is no longer offered as a preset or default (since 2026-09-10); it remains available only as a custom size. Dimensions use a **32 px grid**, at least 480 px per side, up to 1344 px per side and no more than 1,032,192 rendered pixels. - **1080p and 2K output** are two-stage FastH3 deliveries (`outputScale: 2` on the FastH3 IDs only): the clip renders on a canvas and arrives at exactly twice its width and height with the same frame count, 24 fps and audio. **2K** doubles a 768p canvas (1344×768 arrives as 2688×1536); **1080p** doubles a 544 px short-edge canvas (960×544 arrives as 1920×1088). See [1080p and 2K output](https://docs.sogni.ai/models/minimax-h3/#1080p-and-2k-output). - Standard H3 uses fixed 20-step `res_multistep` / `simple` sampling. Balanced uses eight-step Euler/simple sampling at a video shift of 6. FastH3 Turbo uses its qualified four-step Euler/simple recipe, while LightX2V Turbo uses its qualified four-step ER-SDE/simple recipe. Guidance is fixed at 1. - Balanced accepts an optional `shift` from 4 to 12 in steps of 0.5 (default 6). Higher values spend more of the eight steps on overall layout and motion, lower values more on fine detail; if a Balanced render drifts, warps, or loses its subject, try 12 with the same seed. Values below 6 have not been reviewed. Other modes ignore `shift`. - Ref2VA reference sets: up to 9 images, 3 videos (24 fps, 2–15 s each and 15 s combined, each with an optional soundtrack), and 3 standalone audio clips — at most 12 reference files, with at least one visual reference (image or video). Audio alone is rejected. ## [#](https://docs.sogni.ai/models/minimax-h3/#pricing)Pricing MiniMax H3 is priced per delivered second at the standard 1 Spark = $0.005 peg. **FastH3 costs 4 Spark ($0.02) per output second at both resolution classes**, so an exact 192-frame, 8-second FastH3 clip costs **32 Spark ($0.16)**. LightX2V Turbo remains 4 Spark ($0.02) per second at 480p and 6 Spark ($0.03) at 544/768p-class output. Standard costs 10 Spark ($0.05) per second at 480p and 16 Spark ($0.08) at 544/768p-class output. Balanced costs 6 Spark ($0.03) per second at 480p and 10 Spark ($0.05) at 544/768p-class output. **FastH3 1080p and 2K add a flat surcharge per delivered second:** 10 Spark ($0.05) for a 544/768p-class canvas, making FastH3 **14 Spark ($0.07) per second**, and 6 Spark ($0.03) for a 480p-class canvas. Both 1080p (960×544 canvas) and 2K (1344×768 canvas) are in the 544/768p class, so an exact 141-frame, 5.88-second clip at either costs **82.25 Spark ($0.41)**. The resolution class is decided by the canvas that renders, not by the doubled delivery size. - **Resolution tier and speed tier set the rate.** Aspect ratio and mode do not change it within that tier. - **1080p and 2K are a per-second surcharge, not a multiplier.** The same 10 Spark ($0.05) per second is added to the FastH3 rate at either size. They are offered on FastH3 only; LightX2V Turbo, Balanced, Standard and every Reference-to-Video mode deliver their normal size. - **Ref2VA video input is billed by duration and resolution.** Each reference-video input second costs 10 Spark ($0.05) at 480p or 16 Spark ($0.08) at 544/768p-class output. Those full input rates apply to Standard, Balanced, and LightX2V Turbo; output-tier discounts apply only to generated output. FastH3 has no Ref2VA route. - **Standard and Balanced Ref2VA include five images.** Each additional reference image costs 16 Spark ($0.08). Reference audio is free, and LightX2V Turbo has no excess-image surcharge. - **You pay for the length that actually renders.** H3 snaps to the `124 + n×17` frame grid at 24 fps, so billable durations land on grid values between 5.17 s and 15.08 s. ### [#](https://docs.sogni.ai/models/minimax-h3/#480p-output-custom-size-only)480p output (custom size only) | Frames | Duration | Standard | Balanced | LightX2V Turbo | FastH3 | | --- | --- | --- | --- | --- | --- | | 124 | 5.17 s | 51.67 Spark / $0.2583 | 31 / $0.155 | 20.67 / $0.1033 | 20.67 / $0.1033 | | 141 | 5.88 s | 58.75 / $0.2938 | 35.25 / $0.1763 | 23.5 / $0.1175 | 23.5 / $0.1175 | | 175 | 7.29 s | 72.92 / $0.3646 | 43.75 / $0.2188 | 29.17 / $0.1458 | 29.17 / $0.1458 | | 192 | 8.00 s | 80 / $0.40 | 48 / $0.24 | 32 / $0.16 | **32 / $0.16** | | 243 | 10.13 s | 101.25 / $0.5063 | 60.75 / $0.3038 | 40.5 / $0.2025 | 40.5 / $0.2025 | | 362 | 15.08 s | 150.83 / $0.7542 | 90.5 / $0.4525 | 60.33 / $0.3017 | 60.33 / $0.3017 | ### [#](https://docs.sogni.ai/models/minimax-h3/#544768p-class-output)544/768p-class output | Frames | Duration | Standard | Balanced | LightX2V Turbo | FastH3 | | --- | --- | --- | --- | --- | --- | | 124 | 5.17 s | 82.67 Spark / $0.4133 | 51.67 / $0.2583 | 31 / $0.155 | 20.67 / $0.1033 | | 141 | 5.88 s | 94 / $0.47 | 58.75 / $0.2938 | 35.25 / $0.1763 | 23.5 / $0.1175 | | 175 | 7.29 s | 116.67 / $0.5833 | 72.92 / $0.3646 | 43.75 / $0.2188 | 29.17 / $0.1458 | | 192 | 8.00 s | 128 / $0.64 | 80 / $0.40 | 48 / $0.24 | **32 / $0.16** | | 243 | 10.13 s | 162 / $0.81 | 101.25 / $0.5063 | 60.75 / $0.3038 | 40.5 / $0.2025 | | 362 | 15.08 s | 241.33 / $1.2067 | 150.83 / $0.7542 | 90.5 / $0.4525 | 60.33 / $0.3017 | ### [#](https://docs.sogni.ai/models/minimax-h3/#fasth3-1080p-and-2k-544768p-class-canvas-delivered-at-twice-its-size)FastH3 1080p and 2K (544/768p-class canvas, delivered at twice its size) | Frames | Duration | FastH3 1080p or 2K | | --- | --- | --- | | 124 | 5.17 s | 72.33 Spark / $0.3617 | | 141 | 5.88 s | **82.25 / $0.4113** | | 175 | 7.29 s | 102.08 / $0.5104 | | 192 | 8.00 s | 112 / $0.56 | | 243 | 10.13 s | 141.75 / $0.7088 | | 362 | 15.08 s | 211.17 / $1.0558 | H3 runs on Sogni Supernet workers rather than an external vendor API, so it uses standard Spark access and does not require credit-card Premium Spark coverage. For a live quote at any frame count, use the [job estimate calculator](https://docs.sogni.ai/pricing/#calculator). ## [#](https://docs.sogni.ai/models/minimax-h3/#1080p-and-2k-output)1080p and 2K output **1080p** and **2K** are FastH3's two-stage deliveries and the best MiniMax H3 quality Sogni offers: they look sharper than a one-stage FastH3 render at a smaller size. FastH3 renders the clip on a canvas, then the worker enlarges the finished video latent 2× with the MiniMax H3 Latent Upscaler (Community), built on [Comfyui\_Minimax\_h3\_latent\_Upscaler](https://github.com/LBH-123-AI/Comfyui_Minimax_h3_latent_Upscaler) by LBH-123-AI. It re-encodes any first and last frames at the doubled size, refines the enlarged clip on overlapping tiles for two extra steps, and decodes it once. The delivered file is exactly twice the canvas width and height, with the same frame count, 24 fps, seed and stereo audio. | Output | Canvas that renders | Delivered | | --- | --- | --- | | 1080p | the aspect at a 544 px short edge, e.g. 960×544 | 1920×1088 | | 2K | the 768p canvas, e.g. 1344×768 | 2688×1536 | FastH3 at 2K is four times the pixels of the 1344×768 the Standard tier delivers, finishes about 20% sooner than Standard, and costs a little less. - **Where:** FastH3 text-to-video, image-to-video and first/last-frame (`minimax-h3-fastvideo-int8_t2v_turbo`, `_i2v_turbo`, `_flf2v_turbo`). Other tiers refuse `outputScale: 2`. - **API:** add `outputScale: 2` to a FastH3 video project. Send a 768p canvas such as 1344×768 for 2K, or a 544p canvas such as 960×544 for 1080p. The default, `1`, delivers the canvas size. - **Sogni Web:** choose **Output resolution → 1080p or 2K** in FastH3 mode. Clips are limited to 6 seconds there for now. - **Price:** the surcharge in [Pricing](https://docs.sogni.ai/models/minimax-h3/#pricing): +10 Spark ($0.05) per second at 1080p or 2K. - **Time:** measured warm for a 158-frame (6.6 s) clip on an RTX 5090: one-stage 1344×768 about 53 s, 1080p about 131 s, 2K about 208 s. - **Workers:** served by Comfy worker 1.0.215 and newer. Clips of 175 frames or more (about 7.3 seconds) go only to 48 GB-class workers, so they can wait longer than the same clip at the normal size. If no eligible worker is connected, the request is refused up front with a message saying so, and you can submit the normal size instead. MiniMax's own hosted H3-Context-IR model stage and H3-Regenerate-2K feature are still not published as open weights, so these outputs are a learned latent enlargement plus refinement of the FastH3 render, not MiniMax's proprietary second stage. If MiniMax releases that stage as open weights, Sogni plans to integrate it, and like other open-weight model updates it would be included automatically for existing subscribers. ## [#](https://docs.sogni.ai/models/minimax-h3/#unlimited-throughput-and-fair-use)Unlimited throughput and fair use Both [Sogni Unlimited plans](https://docs.sogni.ai/pricing/unlimited-plan-details/) cover standard H3 and H3 Turbo under credit-free fair use. The render time for one job is the same on both plans, but Unlimited Pro adds higher subscription queue priority, an additional standard H3 concurrency slot, and **twice the H3 fair-use capacity** of Unlimited. Sogni intentionally does not publish a fixed number of 15-second H3 videos per 24 hours. The average paid subscriber does not encounter daily throttling, but concurrency is not the only control: paid plans also use adaptive daily fair-use throughput limits to keep the shared network responsive. Exact thresholds are unpublished to prevent gaming and can shift with GPU supply and demand. Within each plan tier, daily usage is a scheduling signal so accounts with lighter use are favored and more subscribers receive fast service with minimal queue time. The platform is designed for an individual creator to keep a substantial queue moving throughout the day. Unlimited can queue up to 8 videos at once and Unlimited Pro up to 24; the API and [Creative Agent Skill](https://docs.sogni.ai/sogni-intelligence/creative-agent-skill/) make it practical to automate a personal production queue. Fair use still applies: unattended continuous 24/7 infrastructure and multi-user production workloads require pay-as-you-go Premium Spark or an Enterprise arrangement. ## [#](https://docs.sogni.ai/models/minimax-h3/#javascript-sdk-example)JavaScript SDK example Use the exact FastH3 workflow ID. Sampling settings are owned by the workflow, so omit `steps`, `guidance`, sampler, and scheduler overrides: ``` const project = await sogni.projects.create({ type: 'video', network: 'fast', modelId: 'minimax-h3-fastvideo-int8_t2v_turbo', positivePrompt: `integrated_multimodal_description: [Shot 1] A cinematic tracking shot moves through a rainy night market as vendors serve customers beneath glowing awnings. overall_soundscape: Steady rain, footsteps through shallow puddles, and layered crowd ambience. non_diegetic_music: A restrained electronic pulse builds beneath the scene.`, duration: 8, generateAudio: true, tokenType: 'spark' }); const urls = await project.waitForCompletion(); ``` For [1080p or 2K output](https://docs.sogni.ai/models/minimax-h3/#1080p-and-2k-output), add `outputScale: 2` to a FastH3 request: keep `width: 1344, height: 768` for 2K, or send `width: 960, height: 544` for 1080p. For first-frame FastH3 animation, use `minimax-h3-fastvideo-int8_i2v_turbo` with `referenceImage`. For a FastH3 transition between two anchors, use `minimax-h3-fastvideo-int8_flf2v_turbo` with `referenceImage` and `referenceImageEnd`. Use the corresponding `minimax-h3-fl2va-fp8_*_turbo` ID when you want LightX2V Turbo. Include `referenceVideoDurations` in matching order for an accurate Ref2VA preflight quote; final billing uses the accepted reference-video duration. ## [#](https://docs.sogni.ai/models/minimax-h3/#creative-agent-skill-examples)Creative Agent Skill examples The Creative Agent Skill exposes friendly selectors for every current workflow. The generic `minimax-h3` and `minimax-h3-turbo` selectors infer text-to-video, image-to-video, or first/last-frame mode from the supplied frames; standard Ref2VA remains explicit. ``` # Standard quality-first text-to-video sogni-agent --video -m minimax-h3 --duration 10 "" # 4-step Turbo image-to-video sogni-agent --video -m minimax-h3-i2v-turbo --ref first.png --duration 8 "" # 4-step Turbo first-and-last-frame video sogni-agent --video -m minimax-h3-flf2v-turbo --ref first.png --ref-end last.png --duration 8 "" # FastH3 Turbo (FastVideo FastH3 4-step Preview v1); --ref and --ref-end infer I2VA, L2VA, or FL2VA sogni-agent --video -m minimax-h3-fasth3-turbo --duration 8 "" sogni-agent --video -m minimax-h3-fasth3-i2v-turbo --ref first.png --duration 8 "" # Standard Ref2VA with labelled image, video, and audio references sogni-agent --video -m minimax-h3-r2v --ref identity.png --ref-video motion.mp4 --ref-audio voice.m4a "" ``` Use `minimax-h3-turbo` when you want the Skill to infer LightX2V Turbo t2v/i2v/flf2v automatically, and `minimax-h3-fasth3-turbo` for the same inference on FastH3. Use `minimax-h3-r2v-turbo` for Turbo Ref2VA; FastH3 rejects `--workflow r2v`. ## [#](https://docs.sogni.ai/models/minimax-h3/#worker-hardware)Worker hardware FastH3 Turbo supports custom H3 LoRAs on workers with at least **32 GB of VRAM** without switching engines. Base text-to-video, image-to-video, and first/last-frame jobs can additionally run on eligible CUDA 13 workers with at least **23 GB of usable VRAM**, including the **NVIDIA GeForce RTX 4090 and RTX 3090**. LightX2V Turbo, Standard, Balanced, and every Reference-to-Video mode require at least **32 GB of VRAM**. Reference-to-Video always uses its dedicated checkpoint, and CUDA 12 workers do not advertise FastH3 because that track lacks the required optimized INT8 kernels. Sogni Comfy Worker applies Blackwell-specific runtime tuning to the RTX 5090 and RTX PRO 6000 Blackwell, but H3 is not restricted to Blackwell. See the [Fast Worker requirements](https://docs.sogni.ai/run-a-worker/fast-worker/about-sogni-fast-worker/). Sogni's FastH3 [1080p and 2K output](https://docs.sogni.ai/models/minimax-h3/#1080p-and-2k-output) is a learned latent enlargement plus a short tiled refinement of the FastH3 render, served by the same workers. MiniMax's hosted H3-Context-IR model stage and its H3-Regenerate-2K feature are not open weights and are not part of the current release; Sogni plans to integrate that stage if MiniMax publishes it, and it would be included automatically for existing subscribers. Unless a release is specifically marked otherwise, newly supported open-weight models and related updates are automatically included in both existing and new subscriptions. The separate H3 Ref2VA reference-to-video checkpoint **is** already included and runs on the same 32 GB worker class. The hosted model stage is distinct from the ordered Context-IR prompt document used by the released workflows. ## [#](https://docs.sogni.ai/models/minimax-h3/#prompting)Prompting Base and Turbo text-to-video, image-to-video, and first/last-frame workflows use MiniMax's ordered three-field prompt contract: ``` integrated_multimodal_description: [Shot 1] ... overall_soundscape: ... non_diegetic_music: ... ``` Image-to-video and first/last-frame prompts prepend their mode-specific alignment instruction, followed by a blank line before the three fields. Use `[Shot N]` notation, keep stable `(S1)`, `(S2)`, and subsequent speaker IDs across shots, and write exact dialogue as `[Language] words`. Put negative direction inside the structured prompt because H3 has no negative-prompt field. Ref2VA uses a separate six-field contract in this order: `subject_definitions`, `summary`, `retention_analysis`, `detailed_description`, `overall_soundscape`, and `non_diegetic_music`. Its references are loose references rather than locked frames. Address them with H3's 1-based per-type labels — ``, `