Integrated timeline
Write every visible action, shot, camera move, speaker, line, and synchronized scene sound.
MiniMax H3 prompt guide
Use MiniMax’s official audiovisual fields for text, first-frame, first-and-last-frame, last-frame, and full-reference generation.
Official base format
Keep the order exact. Put dialogue and sound heard inside the scene in the timeline. Separate ambience from music only the audience hears.
Write every visible action, shot, camera move, speaker, line, and synchronized scene sound.
Summarize ambience, physical action sounds, and non-verbal human sounds across the clip.
Describe instruments, tempo, rhythm, and dynamics—or write N/A when there is no score.
Pick the mode first
Text builds the full audiovisual timeline
Picture 1 is the exact first frame
Pictures 1 and 2 lock both endpoints
Picture 1 is the exact final frame
Images, video, and audio define reference relationships
Shot 1 has no timestamp · later cuts increase strictly · the final event must fit inside 4–15 seconds
Copy, then rebuild
T2VA · 8 seconds · stereo sound
integrated_multimodal_description: [Shot 1] Live-action premium product film. A cobalt running shoe rests on rain-dark stone at blue hour, framed in a low medium close-up. The camera pushes in slowly as one warm rim light travels across the mesh. The shoe lifts, lands once in a shallow puddle, and sends a clean ring of water outward with one synchronized impact. [Shot 2] At 00:05.000, the camera cuts to an extreme close-up of water beads rolling over the laces as the shoe settles into its final hero angle. overall_soundscape: Fine rain taps the pavement. One firm footfall and crisp water splash sit above distant city air. non_diegetic_music: A restrained electronic bass pulse enters at the landing, followed by one short glassy synth tone at the final frame.
FL2VA · exact endpoints · 10 seconds
How the reference pictures align with the target video — Picture 1 (from Shot 1) aligns with the 0.00-second mark of the target video; Picture 2 (from Shot 2) aligns with the 10.00-second mark of the target video. integrated_multimodal_description: [Shot 1] Fully preserve Picture 1 as the opening composition. A paper fox stands on a white studio floor. The camera begins static as a warm breeze lifts the paper ears and loose paper leaves spiral around the subject. The fox steps forward while the camera arcs right with small amplitude at slow speed. [Shot 2] At 00:06.000, the leaves align into the circular pattern shown in Picture 2. The fox turns toward camera, the light warms, and the camera settles into the exact final framing, pose, object positions, and shadows of Picture 2 at 10.00 seconds. overall_soundscape: Dry paper rustles softly with each step and leaf movement. A light studio breeze continues throughout. non_diegetic_music: Sparse wooden percussion marks the steps, then stops exactly on the final-frame landing.
Ref2VA · image + video + audio
subject_definitions: <Subject 1> is the singer shown in <Picture 1>, including the same face, silver jacket, short dark hair, and body proportions. <Video 1> provides only the slow circular camera path and stage movement. <Audio 1> provides the voice timbre for <Subject 1>. summary: [reference generation + audio reference] Create a new performance using the singer from Picture 1, the camera language from Video 1, and the voice timbre from Audio 1. retention_analysis: Preserve Subject 1 identity and wardrobe from Picture 1. Transfer only the camera path and performance pacing from Video 1; do not copy its performer, set, or colors. Reference Audio 1 only for voice timbre. detailed_description: [Shot 1] Live-action night performance on a dark cobalt stage. Subject 1 stands beneath one coral spotlight. The camera follows the slow circular path from Video 1 as she turns and sings: “I can see the morning now.” Her lip movement matches the line and the voice uses the timbre from Audio 1. Fine haze moves through the light while the silver jacket remains consistent. overall_soundscape: Soft stage-room ambience, controlled breathing, and light fabric movement remain audible. non_diegetic_music: Low analog synth chords sustain beneath the vocal with no percussion.
MiniMax H3 questions
Use the official three-part order: integrated_multimodal_description, overall_soundscape, and non_diegetic_music. Put all timed visual action, dialogue, and scene sound inside the first field.
Read the full answer →Use T2VA for text only, I2VA for an exact first frame, FL2VA to connect exact first and last frames, L2VA to land on an exact final frame, and Ref2VA for full multimodal references.
Read the full answer →Place dialogue and sounds heard inside the scene at the exact shot where they happen. Use overall_soundscape for ambience and physical sound across the clip, then use non_diegetic_music only for the audience score.
Read the full answer →Label every input consistently and define what it controls. A picture can control identity or style, a video can control movement or editing, and audio can control voice, music, rhythm, or reusable sound.
Read the full answer →Use the official alignment line, assign Picture 1 to 0.00 seconds, assign Picture 2 to the exact final time, and describe the continuous physical path that reaches the last frame without adding action afterward.
Read the full answer →Build your H3 prompt
Official sources