Phosor
Browse documentationModel API

Model API

Choose a model guide

ModelIntegration guide
MiniMax H3 TurboTurbo model IDs, examples, settings and limits
MiniMax H3H3 model IDs, examples, settings and limits
Other video, image and audio modelsShared submit reference — use only the section for your selected model

All generation modes share the status, result and history endpoints. The tables below are the full catalog reference.

API version: v1.2.4

Build an AI video generator API integration with MiniMax H3: choose the text-to-video API, image-to-video API, or reference-to-video API with Standard and Turbo models through Phosor. Start from text, a starting image, or image and video references. Image generation is also available.

Public base URL:

  • https://phosor.ai

Reference common contract: common-api.md

MiniMax H3 Standard and Turbo

Choose an input mode and send its exact model ID. Set model explicitly to select H3; omitting it does not select H3.

Input modeStandard model IDTurbo model IDRequired inputs
Text to videominimax/h3/text-to-videominimax/h3/text-to-video/turboprompt
Image to videominimax/h3/image-to-videominimax/h3/image-to-video/turboprompt and starting-frame image_url; optional end_image_url
Reference to videominimax/h3/reference-to-videominimax/h3/reference-to-video/turboprompt and at least one image in reference_image_urls or video in reference_video_urls

Reference-to-video also accepts optional reference_audio_urls. Audio alone is not sufficient: include at least one reference image or video. Starting-frame and ending-frame inputs belong to image-to-video; use the reference arrays for reference-to-video.

For all six H3 models, use duration (4–15 seconds), resolution_tier (480p or 768p) and aspect_ratio. Output uses fixed 24fps. See video generation limits for dimensions and current reference requirements, and current pricing for Standard and Turbo rates.

Each model link above includes a complete request for that mode and variant. The Quick Start walks through submission, status and result retrieval; the submit reference documents all request parameters. Keep the returned request_id to retrieve your video.

Endpoint Summary

MethodPathPurpose
POST/api/v1/inference/submitSubmit video or image generation job
GET/api/v1/inference/status/{request_id}Poll job status
GET/api/v1/inference/result/{request_id}Fetch generated result
GET/api/v1/inference/historyList job history
GET/api/v1/inference/details/{request_id}Fetch task details
GET/api/v1/inference/download/{request_id}Download result media
POST/api/v1/storage/image/uploadUpload image
POST/api/v1/storage/image/importImport image from URL
GET/api/v1/lorasList LoRA models
GET/api/v1/loras/{lora_id}Get LoRA details
GET/api/v1/loras/{lora_id}/statusGet LoRA status
POST/api/v1/loras/{lora_id}/saveActivate a LoRA (extend expiry to 7 days)
DELETE/api/v1/loras/{lora_id}Delete LoRA

Authentication

All calls require:

cURL
X-API-Key: your_api_key_here

Inference Submit

POST /api/v1/inference/submit

Accepted response

JSON
{ "request_id": "uuid", "status": "queued" }

Common request fields

Fields depend on the selected model. For H3, use the input mode and duration controls above; consult the submit reference for model-specific parameters.

  • prompt
  • model
  • width
  • height
  • resolution_tier
  • aspect_ratio
  • num_frames
  • duration
  • frames_per_second
  • image_url
  • end_image_url
  • reference_image_urls
  • reference_video_urls
  • reference_audio_urls
  • audio_url
  • video_url
  • lora_id
  • lora_scale
  • seed

Other public model ID examples

The following are additional examples. See the model catalog and submit reference for available models and their inputs.

  • openai/gpt-image-2/text-to-image
  • qwen-image/v2512/text-to-image
  • qwen-image/v2512/text-to-image/lora
  • qwen-image/v2511/image-edit
  • z-image/turbo/text-to-image
  • z-image/turbo/text-to-image/lora
  • z-image/turbo/image-to-image
  • z-image/turbo/image-to-image/lora
  • wan/v2.2-a14b/text-to-video
  • wan/v2.2-a14b/image-to-video
  • wan/v2.2-a14b/speech-to-video
  • wan/v2.2-a14b/animate

Status / Result

GET /api/v1/inference/status/{request_id}

Returns task state, logs, progress, and request_id.

GET /api/v1/inference/result/{request_id}

Returns result payload with media URL(s) and request id.

GET /api/v1/inference/download/{request_id}

Downloads the final media file.

Notes

  • Use request_id to poll status and fetch results.
  • Error schema follows common-api.md.