H3 Max
H3 Max creates synchronized-audio video from text or frames for 1–15 seconds at 480P/768P/1080P. It supports first/middle/last frames and up to 9 image, 3 video and 3 audio references, with 12 references total. Middle frames require first/last frames, an interior middleFrameTimeSec and native 480P/768P. A soundtrack replaces output audio and remains distinct from the separate LipSync model. Extend adds 1–15 seconds at up to 2K to a 1.625–60 second source; Insert replaces a selected interval; Recast replaces people from 1–4 photos while preserving source motion, cuts and audio.
Cost
from 3 tokens/s
Try H3 Max in Givon AI
Open a working project for regular creation, or use Playground to inspect the model parameters.
Use through API
Request fields, constraints, and ready-to-use examples for developers. The user workflow starts above.
IDEMPOTENCY_KEY="${IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.givon.ai/api/v1/generations \
-H "Authorization: Bearer $GIVON_API_KEY" \
-H "Idempotency-Key: $IDEMPOTENCY_KEY" \
-H "Content-Type: application/json" \
-d '{"type":"video","model":"minimax-h3-max","input":{"prompt":"cinematic drone shot over a city at night","mode":"create","aspectRatio":"16:9","resolution":"768p","duration":5,"audioEnabled":true,"referenceAssetUris":["asset://asset_..."]}}'Input fields
* required| Field | Type | Default | Allowed |
|---|---|---|---|
mode mode | string | create | create, extend, insert, recast |
prompt* promptGeneration prompt describing the scene, action and direction. | string | — | up to 7000 chars |
aspectRatio aspectRatio | string | 16:9 | 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, auto |
resolution resolution | string | 768p | 480p, 768p, 1080p, 2k |
duration duration | number | 5 | 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 |
audioEnabled audioEnabledAudio is created automatically and cannot be turned off separately. | boolean | true | boolean |
Start image startImageAsset input for the startImage slot. | string | — | image · asset, https, data |
Middle image middleImageAsset input for the middleImage slot. | string | — | image · asset, https, data |
End image endImageAsset input for the endImage slot. | string | — | image · asset, https, data |
Reference video referenceVideoAsset input for the referenceVideo slot. | string | — | video · asset, https, data |
Reference video 2 referenceVideo2Asset input for the referenceVideo2 slot. | string | — | video · asset, https, data |
Reference video 3 referenceVideo3Asset input for the referenceVideo3 slot. | string | — | video · asset, https, data |
Speech audio speechAudioAsset input for the speechAudio slot. | string | — | voiceover · asset, https, data |
Speech audio 2 speechAudio2Asset input for the speechAudio2 slot. | string | — | voiceover · asset, https, data |
Speech audio 3 speechAudio3Asset input for the speechAudio3 slot. | string | — | voiceover · asset, https, data |
Soundtrack audio soundtrackAudioAsset input for the soundtrackAudio slot. | string | — | voiceover · asset, https, data |
Reference images referenceAssetUrisOptional reference assets as asset:// URIs, HTTPS URLs, or data URIs. The item limit applies to physical references after repeated owned asset identities and identities already bound to video slots are excluded. URL entries remain distinct; every supplied item is validated. | array | — | image · up to 9 · asset, https, data |
providerOptions providerOptions | object | — | object |
mode- Type
- string
- Default
- create
- Allowed
- create, extend, insert, recast
promptGeneration prompt describing the scene, action and direction.
- Type
- string
- Default
- —
- Allowed
- up to 7000 chars
aspectRatio- Type
- string
- Default
- 16:9
- Allowed
- 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, auto
resolution- Type
- string
- Default
- 768p
- Allowed
- 480p, 768p, 1080p, 2k
duration- Type
- number
- Default
- 5
- Allowed
- 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
audioEnabledAudio is created automatically and cannot be turned off separately.
- Type
- boolean
- Default
- true
- Allowed
- boolean
startImageAsset input for the startImage slot.
- Type
- string
- Default
- —
- Allowed
- image · asset, https, data
middleImageAsset input for the middleImage slot.
- Type
- string
- Default
- —
- Allowed
- image · asset, https, data
endImageAsset input for the endImage slot.
- Type
- string
- Default
- —
- Allowed
- image · asset, https, data
referenceVideoAsset input for the referenceVideo slot.
- Type
- string
- Default
- —
- Allowed
- video · asset, https, data
referenceVideo2Asset input for the referenceVideo2 slot.
- Type
- string
- Default
- —
- Allowed
- video · asset, https, data
referenceVideo3Asset input for the referenceVideo3 slot.
- Type
- string
- Default
- —
- Allowed
- video · asset, https, data
speechAudioAsset input for the speechAudio slot.
- Type
- string
- Default
- —
- Allowed
- voiceover · asset, https, data
speechAudio2Asset input for the speechAudio2 slot.
- Type
- string
- Default
- —
- Allowed
- voiceover · asset, https, data
speechAudio3Asset input for the speechAudio3 slot.
- Type
- string
- Default
- —
- Allowed
- voiceover · asset, https, data
soundtrackAudioAsset input for the soundtrackAudio slot.
- Type
- string
- Default
- —
- Allowed
- voiceover · asset, https, data
referenceAssetUrisOptional reference assets as asset:// URIs, HTTPS URLs, or data URIs. The item limit applies to physical references after repeated owned asset identities and identities already bound to video slots are excluded. URL entries remain distinct; every supplied item is validated.
- Type
- array
- Default
- —
- Allowed
- image · up to 9 · asset, https, data
providerOptions- Type
- object
- Default
- —
- Allowed
- object
Cost
from 3 tokens/sThe variant is selected automatically from request fields, so you do not need to send it.
Capabilities
text_to_videoimage_to_videovideo_referencevideo_extendvideo_insertvideo_editstartImage:imagemiddleImage:imageendImage:imagereferenceVideo:videoreferenceVideo2:videoreferenceVideo3:videospeechAudio:voiceoverspeechAudio2:voiceoverspeechAudio3:voiceoversoundtrackAudio:voiceoverreferenceAssetUris:image[]<=12