audiosarray
Audio requires at least one reference image or video.
MiniMax H3 Reference to Video
minimax/h3/reference-to-videoGenerate native-stereo 2K video guided by image, video, and audio references with MiniMax H3.
Try the model
Playground
Input
Allowed values: 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15.
Additional Settings More
Customize your input with more control.
PNG, JPEG, WebP, or GIF · 20 MiB maximum each · up to 9
Enter a JSON array.
Each clip must be 2–15 seconds; combined duration must not exceed 15 seconds.
Audio requires at least one reference image or video.
Output aspect ratio. Omit to let the model choose from the prompt and reference inputs. Allowed values: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16.
Output resolution. 768P is generated natively; 2K is upscaled via an in-context regeneration pass. Allowed values: 768P, 2K.
Whether to add an AIGC label watermark to the generated video.
ResultExample
What would you like to do next?
Examples
Select an example to load its inputs and result into the playground.
- Open result ↗
Input
{ "aigc_watermark": false, "audios": [], "duration": 7, "images": [ "https://static.sandbase.ai/examples/bff323df-f593-467c-b6a7-96a7bf473983/input/4c91f51e-f6d9-426a-8896-4678c7a115c7.jpg" ], "prompt": "anime scene, two-shot on a campus path: the reference character (male student) stops and faces a middle-aged female dormitory counselor in her 40s, counselor has chestnut bob hair, glasses, navy blazer, holding a small key, slightly serious expression, noon lighting, calm atmosphere, cinematic composition, Hyouka-style", "resolution": "768P", "videos": [] } - Open result ↗
Input
{ "aigc_watermark": false, "audios": [], "duration": 4, "images": [ "https://static.sandbase.ai/examples/f37db918-73c5-4c0b-8801-ad0dc81e3e4f/input/b05d0c19-55b3-4e16-bf09-a50b297dcde6.jpg" ], "prompt": "anime close-up, the reference character's hand reaching out to take a small key, then a quick facial close-up showing an embarrassed awkward expression, soft natural lighting, Hyouka-style reaction shot, comedic timing", "resolution": "768P", "videos": [] }
MiniMax H3 Reference to Video
MiniMax H3 reference-to-video: up to 9 images, 3 videos, and 3 audio clips keep characters, motion, and voice consistent in a new 2K clip.
Fields
| Field | Type | Default | Notes |
|---|---|---|---|
prompt | string, required | — | Up to 7,000 characters. |
images | image URL[] (≤9) | — | Reference images, Image 1, Image 2, … |
videos | video URL[] (≤3) | — | Reference clips, Video 1, … |
audios | audio URL[] (≤3) | — | Reference audio, Audio 1, … |
duration | 4–15, required | 5 | Output length in seconds. |
aspect_ratio | 21:9 / 16:9 / 4:3 / 1:1 / 3:4 / 9:16 | — | Output frame shape. |
resolution | 768P / 2K | 2K | Sets the per-second rate. |
aigc_watermark | boolean | false | Add MiniMax's AI-generated watermark. |
Billing
You pay for output seconds: $0.13 at 2K (the default) and $0.06 at 768P. Each reference image beyond the fifth adds $0.04.
| Output | 2K | 768P |
|---|---|---|
| 4s | $0.52 | $0.24 |
| 5s (default) | $0.65 | $0.30 |
| 10s | $1.30 | $0.60 |
| 15s | $1.95 | $0.90 |
| 5s with 9 images | $0.81 | $0.46 |
Failed runs are not charged.
Prompting
- Give each reference a role: “Image 1 is the singer, Audio 1 is her voice, Video 1 sets the dance”.
Limits
durationis required.
Also on SandBase
Other MiniMax H3 endpoints: minimax/h3/text-to-video, minimax/h3/image-to-video, minimax/h3-max/reference-to-video (faster, cheaper).
Start building
Send your first request
Submit this model through the SandBase Unified Run API.
# The quoted heredoc keeps Unicode and shell metacharacters unchanged.
result=$(curl --fail-with-body --silent \
-X POST "https://api.sandbase.ai/v1/run" \
-H "Authorization: Bearer $SANDBASE_API_KEY" \
-H "Content-Type: application/json" \
--data-binary @- <<'SANDBASE_JSON'
{
"model": "minimax/h3/reference-to-video",
"prompt": "Keep the referenced character and motion consistent while creating a cinematic garden scene with synchronized ambient sound.",
"duration": 5,
"resolution": "2K",
"aigc_watermark": false
}
SANDBASE_JSON
)
run_id=$(printf '%s' "$result" | jq -r .id)
for attempt in $(seq 1 120); do
status=$(printf '%s' "$result" | jq -r .status)
case "$status" in completed|failed|timeout) break ;; esac
sleep 2
result=$(curl --fail-with-body --silent \
-H "Authorization: Bearer $SANDBASE_API_KEY" \
"https://api.sandbase.ai/v1/run/$run_id")
done
status=$(printf '%s' "$result" | jq -r .status)
[ "$status" = completed ] || { echo "Generation ended: $status" >&2; exit 1; }
printf '%s\n' "$result"2. Authentication
Create a SandBase API key and set it in your server environment. All requests use Bearer authentication.
export SANDBASE_API_KEY="YOUR_API_KEY"Keep the key on your server. Do not expose it in browser code or commit it to a repository.
3. Queue
Submit a request
POST to /v1/run with this model ID and its inputs. The request example above includes submission and polling.
Fetch request status and result
Use the returned run ID to poll until the status is completed, failed or timeout. Read output only after completed.
curl "https://api.sandbase.ai/v1/run/$RUN_ID" \
-H "Authorization: Bearer $SANDBASE_API_KEY"4. Files
Hosted files
Pass a publicly accessible file URL in the model’s image, video or audio field. Supported formats and limits are shown beside each Playground input.
Uploading images
Use the upload or replace control on supported image fields in the Playground. Video URL fields accept a hosted driving video; they do not imply video-upload support.
5. Schema
Input
imagesarray
promptstringrequired
Default: "Keep the referenced character and motion consistent while creating a cinematic garden scene with synchronized ambient sound."
videosarray
Each clip must be 2–15 seconds; combined duration must not exceed 15 seconds.
durationintegerrequired
Default: 5
Possible values: 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
resolutionstring
Output resolution. 768P is generated natively; 2K is upscaled via an in-context regeneration pass.
Default: "2K"
Possible values: "768P", "2K"
aspect_ratiostring
Output aspect ratio. Omit to let the model choose from the prompt and reference inputs.
Possible values: "21:9", "16:9", "4:3", "1:1", "3:4", "9:16"
aigc_watermarkboolean
Whether to add an AIGC label watermark to the generated video.
Default: false
{
"model": "minimax/h3/reference-to-video",
"prompt": "Keep the referenced character and motion consistent while creating a cinematic garden scene with synchronized ambient sound.",
"duration": 5,
"resolution": "2K",
"aigc_watermark": false
}Output
A completed run exposes its result in output. The following is the model’s saved example, not a new generation.
[
{
"poster": "https://static.sandbase.ai/examples/posters/ffaeca365afefbb47353.jpg",
"source": "Completed user prediction (human visual safety screened)",
"source_url": "https://www.sandbase.ai",
"url": "https://static.sandbase.ai/examples/bff323df-f593-467c-b6a7-96a7bf473983/output/43ca4f16-10f2-42ff-ab1c-6f829f514451.mp4"
}
]