MiniMax Hailuo
What is MiniMax Hailuo?
MiniMax Hailuo is MiniMax's video generation lineup. Three variants share the same /v1/tasks shape: minimax-2.0 with start-and-end frame control, minimax-2.3 with auto routing between text-to-video and image-to-video, and minimax-2.3-fast for image-forward speed. Output is 768p (default) or 1080p in 6 or 10 second clips; minimax-2.0 also supports 512p for image-to-video. The 1080p mode is locked to 6-second duration; 10-second runs return at 768p. Prompts can run up to 2000 characters across all three variants, which leaves room for narrative beats and detailed scene descriptions. Pick the variant that matches the controls your workflow actually needs.
Key features of MiniMax Hailuo
Five features define how Hailuo fits into a production pipeline.
Three variants on one endpoint
minimax-2.0 carries the start-and-end frame control. minimax-2.3 auto-routes between text-to-video and image-to-video based on whether you pass a start_image_url. minimax-2.3-fast runs the image-led path with shorter turnaround.
Optional end frame on minimax-2.0
Pass end_image_url alongside the start frame and minimax-2.0 animates the connective camera move between them. Useful for ads where you have both the open and the close locked.
768p or 1080p output, 512p on minimax-2.0 I2V
768p is the default and works at 6 or 10 seconds. 1080p is locked to 6 seconds. minimax-2.0 also exposes 512p for image-to-video. Pick the resolution at the placement, not at the prompt.
6 or 10 second clips
6 seconds covers feed loops and short ads where the cut needs to play in any context. 10 seconds gives room for narrative pacing and longer camera moves, at 768p.
2000-character prompts
Long prompt budget across all three variants. Useful when the brief includes detailed subject, environment, lighting, and motion description in one string.
Best for
Start-and-end frame ads
minimax-2.0 with matched hero frames when you already have key art. The model handles the connective camera move.
Text-first storytelling
All three variants accept prompts up to 2000 characters for longer narrative beats.
Quick image-led shoots
minimax-2.3-fast when you have a plate and need motion quickly with a tight latency budget.
Resolution-gated delivery
768p for 10-second cuts and 1080p when shorter runs are acceptable.
Mixed-input upstream forms
minimax-2.3 auto-routes based on whether a start image is present, which keeps the call site simple.
Thumbnail-to-clip teasers
Single start_image_url inputs for social teasers built from existing key art.
Variants
Hailuo has three variants on the same API. Each one exposes different controls, so the choice is really about which controls fit the call.
minimax-2.0
The start-and-end frame variant. Pass a starting frame and, optionally, an ending frame, and the model animates the connective motion. Resolution options are 512p (I2V only), 768p, and 1080p; the end frame works in any resolution.
minimax-2.3
The auto-routing variant. Pass a prompt, a start image, or both, and minimax-2.3 picks between text-to-video (T2V-2.3) and image-to-video (I2V-2.3) automatically. Useful when the upstream form lets users mix inputs and you do not want branching call sites.
minimax-2.3-fast
The image-led speed variant. start_image_url is required, the prompt is optional, and turnaround is shorter than 2.3. Use it when the workflow is image-led and the latency budget is tight.
Use cases
Build a 6-second 1080p ad with a fixed open and close by routing the brief through minimax-2.0, passing the opening hero as start_image_url, and the closing logo card as end_image_url. Run a mixed-input upstream form (text alone, image alone, or both) against minimax-2.3 so a single call site handles every case. Generate fast image-led teasers from product photography by sending start_image_url to minimax-2.3-fast and skipping the prompt entirely. For longer 10-second narrative cuts, drop down to 768p and let the 2000-character prompt budget carry the scene description.
API examples
Call MiniMax Hailuo from any language by POSTing to /v1/tasks. Full parameter docs live at docs.unifically.com/models/video/minimax/hailuo.
curl -X POST https://api.unifically.com/v1/tasks \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "hailuo/minimax-2.0",
"input": {
"prompt": "A beautiful sunset animation",
"start_image_url": "https://example.com/sunset.jpg",
"duration": 6,
"resolution": "1080p"
}
}'
Successful submission returns a task_id. Poll GET /v1/tasks/<task_id> or set a callback_url on the request to receive the finished result.
FAQs
People also ask
Use minimax-2.0 when you need an end frame to anchor the closing composition. Use minimax-2.3 when the upstream form mixes prompts and start images and you want one model that auto-routes between text-to-video and image-to-video. Use minimax-2.3-fast when uploads are mandatory and you want shorter turnaround on image-led jobs.
Duration is 6 or 10 seconds across all three variants. On minimax-2.3 and minimax-2.3-fast, resolution is 768p (default) or 1080p. On minimax-2.0, resolution is 512p (I2V only), 768p (default), or 1080p. 1080p only supports 6-second clips; 10-second clips run at 768p.
minimax-2.0 and minimax-2.3 expect at least one of prompt or start_image_url. minimax-2.3-fast requires a start image and treats the prompt as optional.
Hailuo 2.0's start-and-end frame control is its standout feature in this segment. Kling 2.6 leans on prompt-to-video with native audio. Wan 2.6 leans on multi-shot and reference-to-video. Choose by which of those features the workflow actually needs.
Up to 2000 characters across the three variants.
