# H3 Max Turbo Reference to Video > H3 Max Turbo generates video from image, video, and audio references, with prompt-guided control over subjects, style, motion, and voice. Built on fal’s post-trained MiniMax H3 and optimized inference stack. ## Overview - **Endpoint**: `https://fal.run/minimax/h3-max-turbo/reference-to-video` - **Model ID**: `minimax/h3-max-turbo/reference-to-video` - **Category**: image-to-video - **Kind**: inference **Description**: H3 Max Turbo Reference to Video generates videos from image, video, and audio references with text-prompt control. Use Image 1, Image 2, Video 1, and Audio 1 in your prompt to identify the supplied assets and describe how their subjects, style, motion, or voice should guide the result. Provide up to 9 images, 3 video clips, and 3 audio clips, with no more than 12 files in total. Video and audio clips must each be 2–15 seconds long, with at most 15 seconds combined per modality. You can supply an ordered reference_conditions list instead of the reference URL lists when cross-modality ordering matters; do not combine the two input formats. Generate 0.92–15 seconds of video at 480P or 768P, or use 1080P latent refinement from a native 768P source. Outputs may run up to approximately 0.7 seconds longer than the requested duration. Supported aspect ratios are adaptive, 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16. Choose disabled, balanced, or quality prompt expansion to control how much the model rewrites your prompt. Seed control is available, and the safety checker is enabled by default. **Tags**: reference-to-video, character-consistency, multi-reference, stylized, transform ## Pricing - **Price**: $0.05 per seconds For more details, see [fal.ai pricing](https://fal.ai/pricing). ## API Information This model can be used via our HTTP API or more conveniently via our client libraries. See the input and output schema below, as well as the usage examples. ### Input Schema The API accepts the following input parameters: - **`prompt`** (`string`, _required_): Text prompt for video generation. Refer to reference assets by their modality and order in the reference lists: Image 1, Image 2, Video 1, Audio 1, and so on. - Examples: "Image 1 is the female protagonist. Image 2 is her small dog. Keep the woman and dog consistent with their respective reference images while they walk together through a sunlit garden." - **`reference_image_urls`** (`list`, _optional_): URLs of subject/style reference images, referenced in the prompt as Image 1, Image 2, and so on. Reference images, videos, and audio clips must add up to at most 12 files. - Array of string - Examples: ["https://storage.googleapis.com/falserverless/example_inputs/hailuo23/pro_i2v_in.jpg"] - **`reference_video_urls`** (`list`, _optional_): URLs of motion/reference video clips (2-15 seconds each, combined duration at most 15 seconds), referenced in the prompt as Video 1, Video 2, and so on. Reference images, videos, and audio clips must add up to at most 12 files. - Array of string - **`reference_audio_urls`** (`list`, _optional_): URLs of reference audio clips (2-15 seconds each, combined duration at most 15 seconds), referenced in the prompt as Audio 1, Audio 2, and so on. Images, videos, and audio can be provided individually or together. Reference images, videos, and audio clips must add up to at most 12 files. - Array of string - **`duration`** (`float`, _optional_): Video length in seconds. The output can run up to about 0.7 s longer than requested. Default value: `5` - Default: `5` - Range: `0.92` to `15` - **`resolution`** (`ResolutionEnum`, _optional_): The native generation resolution, or 1080P latent refinement from a native 768P source. Default value: `"768P"` - Default: `"768P"` - Options: `"480P"`, `"768P"`, `"1080P"` - **`seed`** (`integer`, _optional_): Random seed. A random seed is selected when omitted. - **`enable_safety_checker`** (`boolean`, _optional_): If set to true, the safety checker will be enabled. Default value: `true` - Default: `true` - **`sync_mode`** (`boolean`, _optional_): Return the generated video as base64 instead of a CDN URL. - Default: `false` - **`prompt_expansion_mode`** (`string`, _required_): How much effort to spend rewriting the prompt before generation. 'disabled' skips prompt expansion. 'balanced' returns in about a second. 'quality' spends up to ~30s on a richer prompt. Default value: `"balanced"` - Default: `"balanced"` - Examples: "disabled", "balanced", "quality" - **`aspect_ratio`** (`AspectRatioEnum`, _optional_): The aspect ratio of the generated video. Default value: `"adaptive"` - Default: `"adaptive"` - Options: `"adaptive"`, `"21:9"`, `"16:9"`, `"4:3"`, `"1:1"`, `"3:4"`, `"9:16"` - **`reference_conditions`** (`list`, _optional_): Ordered reference list; cross-modality order is semantic and is preserved end-to-end. Do not combine with reference_*_urls. - Array of ReferenceCondition **Required Parameters Example**: ```json { "prompt": "Image 1 is the female protagonist. Image 2 is her small dog. Keep the woman and dog consistent with their respective reference images while they walk together through a sunlit garden.", "prompt_expansion_mode": "disabled" } ``` **Full Example**: ```json { "prompt": "Image 1 is the female protagonist. Image 2 is her small dog. Keep the woman and dog consistent with their respective reference images while they walk together through a sunlit garden.", "reference_image_urls": [ "https://storage.googleapis.com/falserverless/example_inputs/hailuo23/pro_i2v_in.jpg" ], "duration": 5, "resolution": "768P", "enable_safety_checker": true, "prompt_expansion_mode": "disabled", "aspect_ratio": "adaptive", "reference_conditions": [ { "type": "image", "url": "" } ] } ``` ### Output Schema The API returns the following output format: - **`video`** (`File`, _required_): The generated video - **`expanded_prompt`** (`string`, _optional_): The prompt after expansion, as sent to the model. Null when prompt expansion was disabled, left the prompt unchanged, or was performed internally by MiniMax's hosted API. - **`seed`** (`integer`, _required_): Base seed for reproducing the generation. - **`reference_tokens`** (`integer`, _optional_): Packed reference conditioner rows encoded for this request (images, videos, keyframes, and audio). The first 4096 rows are included in the clip price. - **`timings`** (`object`, _optional_): Timing breakdown in seconds. 'inference' is the DiT denoising time on the GPU backend. Null on routes that do not report backend timings. **Example Response**: ```json { "video": { "url": "", "content_type": "image/png", "file_name": "z9RV14K95DvU.png", "file_size": 4404019 } } ``` ## Usage Examples ### cURL ```bash curl --request POST \ --url https://fal.run/minimax/h3-max-turbo/reference-to-video \ --header "Authorization: Key $FAL_KEY" \ --header "Content-Type: application/json" \ --data '{ "prompt": "Image 1 is the female protagonist. Image 2 is her small dog. Keep the woman and dog consistent with their respective reference images while they walk together through a sunlit garden.", "prompt_expansion_mode": "disabled" }' ``` ### Python Ensure you have the Python client installed: ```bash pip install fal-client ``` Then use the API client to make requests: ```python import fal_client def on_queue_update(update): if isinstance(update, fal_client.InProgress): for log in update.logs: print(log["message"]) result = fal_client.subscribe( "minimax/h3-max-turbo/reference-to-video", arguments={ "prompt": "Image 1 is the female protagonist. Image 2 is her small dog. Keep the woman and dog consistent with their respective reference images while they walk together through a sunlit garden.", "prompt_expansion_mode": "disabled" }, with_logs=True, on_queue_update=on_queue_update, ) print(result) ``` ### JavaScript Ensure you have the JavaScript client installed: ```bash npm install --save @fal-ai/client ``` Then use the API client to make requests: ```javascript import { fal } from "@fal-ai/client"; const result = await fal.subscribe("minimax/h3-max-turbo/reference-to-video", { input: { prompt: "Image 1 is the female protagonist. Image 2 is her small dog. Keep the woman and dog consistent with their respective reference images while they walk together through a sunlit garden.", prompt_expansion_mode: "disabled" }, logs: true, onQueueUpdate: (update) => { if (update.status === "IN_PROGRESS") { update.logs.map((log) => log.message).forEach(console.log); } }, }); console.log(result.data); console.log(result.requestId); ``` ## Additional Resources ### Documentation - [Model Playground](https://fal.ai/models/minimax/h3-max-turbo/reference-to-video) - [API Documentation](https://fal.ai/models/minimax/h3-max-turbo/reference-to-video/api) - [OpenAPI Schema](https://fal.ai/api/openapi/queue/openapi.json?endpoint_id=minimax/h3-max-turbo/reference-to-video) ### fal.ai Platform - [Platform Documentation](https://fal.ai/docs/documentation) - [Python Client](https://fal.ai/docs/api-reference/client-libraries/python) - [JavaScript Client](https://fal.ai/docs/api-reference/client-libraries/javascript) ### Other agent-readable surfaces This file covers one model. To find anything else: - [Platform overview](https://fal.ai/llms.txt): Entry points and representative endpoint IDs - [Documentation index](https://fal.ai/docs/llms.txt): Every documentation page - [Full documentation text](https://fal.ai/docs/llms-full.txt): The whole documentation inlined - Any other model: `https://fal.ai/models//llms.txt`