minimax/h3-singularity-lora/reference-to-videoFrom 45 credits for a 5-second clip· Reference to Video

MiniMax H3 Singularity LoRA reference-to-video mixes up to nine images, three clips and three audio clips with up to three LoRAs on a community H3 fine-tune. Shots run 3 to 15 seconds at 480p to 1080p with stereo sound, in seven aspect ratios.

Call it with the model name minimax/h3-singularity-lora/reference-to-video on POST /jobs/createTask, from 45 credits for a 5-second clip ($0.45). New accounts get 30 free credits to try it in the Playground first.

  • Image to Video
  • Reference to Video
  • Extend Video
  • Up to 1080p
  • 3–15s clips
  • Start and end frames
  • Up to 9 reference images
  • 7 aspect ratios
  • Loads your own LoRA

Introduction

MiniMax H3 Singularity LoRA Reference to video builds one shot from your reference material on Singularity, the community's HDR-trained fine-tune of the open-weights MiniMax H3, with up to three of your own H3 LoRAs loaded. Send up to nine images, three clips and three audio clips, name them by number in the prompt, and get 3 to 15 seconds at 480p to 1080p with a stereo soundtrack in one of seven aspect ratios.

Choose a mode by starting material

What you haveMode
an opening imageMiniMax H3 Singularity LoRA Image to video
reference images, clips or audioMiniMax H3 Singularity LoRA Reference to video
a written scene onlyMiniMax H3 LoRA Text to video

Key features

Choreography from a clip, identity from stills

A reference clip hands over the movement and camera, and still images pin the face and the costume. Singularity's training on fast, high-dynamic-range action keeps that borrowed motion sharp instead of smeared.

Standard H3 LoRAs on a sharper base

Any LoRA trained for MiniMax H3 loads here unchanged. Pair a character LoRA with stills of the same person for the firmest identity hold.

Request parameters

ParameterRequiredDescription
aspect_ratioNoThe shape of the frame: wide, square or tall. 16:9 · 9:16 · 1:1 · 4:3 · 3:4 · 21:9 · 9:21 · Default: 16:9
duration_secondsNoHow long the clip runs. You pay per second. ≥ 3 · ≤ 15 · Default: 5
lorasNoLoad your own LoRA files to lock in a style or character. Default: []
promptYesRefer to the references by number and describe the shot they should produce, including the sound.
reference_audio_urlsNoAudio to match: a voice, music or a rhythm.
reference_image_urlsNoPictures of the people, outfits or places the result should keep.
reference_video_urlsNoClips whose motion, camera or style the result should follow.
resolutionNoOutput size. Bigger sizes look sharper and cost more. 480p · 540p · 768p · 1080p · Default: 480p
seedNoReuse the same number to get a repeatable result. ≥ 0 · ≤ 2147483647

Pricing

Prices are in credits, the same credits you use on the website (1 credit = $0.01). The hold is taken when the task is accepted and settled to the actual charge when it finishes; failed tasks are refunded.

OptionCredits per second≈ USD
resolution 480p9$0.09
resolution 540p13.5$0.14
resolution 768p20.25$0.20
resolution 1080p40.5$0.41

For an exact number before you run, send the same body to POST /jobs/quote.

Use cases

Restage a routine with a new performer

Send the routine as Video 1 and the new performer as Picture 1, then describe the setting. The move carries over; the person and the place are yours.

Consistent fighters across a sequence

Keep each fighter's stills and LoRA fixed and change only the beat you describe. Wide shots keep their faces better on this fine-tune than on the regular model.

Music-driven cuts

Add the track as Audio 1 and ask for movement on the beat. Combined with a reference clip of the dance, the result follows both the rhythm and the choreography.

Prompt guide & pro tips

Keep reference clips short and clean

Trim away the run-up and the landing you do not need, and prefer a locked-off camera in the reference so the model reads the body movement rather than the operator's.

Give each still a single job

One image for the face, one for the outfit, one for the room. Then write the prompt in beats: Picture 1 enters, copies the spin from Video 1, lands facing the camera, with the drum hit from Audio 1 on the landing.

Plan for the render time

Singularity renders more slowly than the regular MiniMax H3 LoRA family; our five-second 480p reference shot took about 80 seconds. Settle the choreography at 480p before stepping up to 1080p.

API quick start

Call the MiniMax H3 Singularity LoRA API in three steps

  1. Create an API key. Sign in to PopNow and create a key in the API console. Keys start with sk-popnow- and use the same credits as the website.
  2. Send the request. POST /jobs/createTask with model minimax/h3-singularity-lora/reference-to-video and an input object that follows the parameter table. The response returns a task ID straight away.
  3. Get the result. Poll GET /jobs/recordInfo with the task ID, or pass a callBackUrl and we post the result to it. Download the output links when the task succeeds.
# 1. Create the task. Keep the Idempotency-Key if you retry after a network error.
export IDEMPOTENCY_KEY="$(uuidgen)"
curl -X POST https://api.popnow.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $POPNOW_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $IDEMPOTENCY_KEY" \
  -d '{
  "model": "minimax/h3-singularity-lora/reference-to-video",
  "input": {
    "prompt": "A slow aerial shot drifting over a misty pine forest at sunrise",
    "reference_image_urls": [
      "https://media.popnow.ai/influencer/examples/ava.webp"
    ],
    "aspect_ratio": "16:9",
    "resolution": "480p",
    "duration_seconds": 5
  }
}'

# 2. Copy data.taskId from the response, then poll until state is succeeded, failed, canceled or expired.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.popnow.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
  -H "Authorization: Bearer $POPNOW_API_KEY"

Replace the prompt and any links with your own. The API tab has the response, polling and callback reference.

MiniMax H3 Singularity LoRA API questions

How much does the MiniMax H3 Singularity LoRA API cost?

MiniMax H3 Singularity LoRA starts at 45 credits ($0.45) for a 5-second clip with Reference to Video, its cheapest task. Prices by option: 480p 9 credits per second, 540p 13.5 credits per second, 768p 20.25 credits per second, 1080p 40.5 credits per second. Credits are the same as on the website (1 credit = $0.01), new accounts get 30 free credits, and failed tasks are refunded.

What is the model name for MiniMax H3 Singularity LoRA in the API?

Put minimax/h3-singularity-lora/image-to-video for Image to Video; minimax/h3-singularity-lora/reference-to-video for Reference to Video; minimax/h3-singularity-lora/video-extend for Extend Video in the model field of POST /jobs/createTask.

Do my MiniMax H3 LoRAs work on it?

Yes. Singularity keeps the H3 architecture, so LoRAs trained for H3 load as they are. We ran the same LoRA file on both families.

Is it slower?

Yes. A five-second 480p clip took about 80 seconds in our testing, roughly twice the regular family.

Why is there no text-to-video?

This family offers the three Singularity modes that start from your material: an opening frame, references, or a clip to extend. For text-only prompts use MiniMax H3 LoRA Text to video.

What are the limits of the MiniMax H3 Singularity LoRA API?

Clip length: 3–15 s; Resolutions: 480p · 540p · 768p · 1080p; Aspect ratios: 16:9 · 9:16 · 1:1 · 4:3 · 3:4 · 21:9 · 9:21; Reference audio: 3; Reference images: 9. Requests outside these limits are rejected before anything is charged.

Can I try MiniMax H3 Singularity LoRA without writing code?

Yes. The Playground runs MiniMax H3 Singularity LoRA on your PopNow credits, needs no API key, and shows the matching code for every run.

All MiniMax models

The regular MiniMax H3 LoRA is quicker and cheaper for calm, slow shots and adds text-to-video. Switch to Singularity when a fight, a dance or a whip pan starts to smear.

More about MiniMax H3 Singularity LoRA