Skip to content

MiniMax H3 / Reference to video

Build a multi-shot MiniMax H3 story with consistent references

A longer story is easier to manage as separately generated shots with a shared visual brief. Reference-to-video can help carry subject cues between them, but it does not guarantee perfect continuity. Plan the edits so that small differences are less disruptive and inspect every selected shot.

By JollyAIReviewed 2 min read

Create a continuity sheet

Record the subject's defining features, clothing or product details, environment and light direction. Choose a small set of approved reference images. Do not replace those references casually midway through production; a new image can change the appearance the model is being asked to follow.

Give each shot one purpose: establish a location, show an action or reveal a detail. Avoid asking every shot to be a complete story. A close-up insert can bridge two wider shots more gracefully than a generated transformation between unrelated compositions.

Generate in a reviewable order

  1. Approve an establishing shot and save the matching reference set.
  2. Generate a second shot with a different framing but the same subject description.
  3. Compare distinctive details side by side before generating more shots.
  4. Assemble a rough edit and replace continuity failures that remain visible in context.

Example shot brief

PROMPT RECIPE / ADAPT TO YOUR SOURCE

Use the reference images for the same green travel bag with black handles and one front pocket. Show a close view of the bag resting on a wooden bench at sunrise. A gentle camera push highlights the fabric. Preserve its shape and colours, with quiet station ambience.

Manage continuity failures

If a detail keeps changing, simplify the shot or avoid making that detail the focus of a cut. Do not rely on a seed alone to lock identity across different prompts and inputs. Keep the approved original references, generation settings and final filenames together so a replacement shot can be produced without reconstructing the entire creative brief.

REVIEWED JOLLYAI IMPLEMENTATION

The workflow behind the guide

One or two image references + optional video + written brief → reference conditioning → separate Ref2VA checkpoint → video/audio decoding → combined video file.

Inputs, presets & access
PRO. One or two images and optionally one video; the reviewed backend uses about the first three seconds of the reference video. Current ordinary pixel budget: 0.3 MP.
Implementation note
Local H3 turbo workflow. Six sampling steps; simple scheduler; turbo adapter; video/audio shifts 12/3. Current environment: pruned W4A8 FL2VA weights with a 4B encoder and matching projection. R2V uses a separate reference checkpoint. These settings describe the reviewed JollyAI deployment, not every H3 implementation.

Configuration reviewed September 20, 2026. Your current playground controls and plans page determine availability. Prompt recipes above are suggestions, not guarantees or claimed test results.

Open Reference to video ↗

Sources and useful next steps

The implementation notes describe JollyAI's reviewed local integration. Upstream capabilities and licences may differ from the settings available here.

Found an outdated setting or a reproducible problem? Contact JollyAI with the workflow and job reference. Keep account tokens and private input files out of public reports.

KEEP BUILDING

Continue the workflow.