Master prompting techniques to generate professional AI videos. This guide contains best practices for multi-modal references.
🌐The sample videos below are generated with Chinese prompts. When you write your prompt in English (or any other language), the AI-generated speech and subtitles will match your language.
Muse Video deeply follows natural language logic, allowing you to flexibly combine the following elements based on your needs.
Beyond text descriptions, you can also upload assets to lock in your ideal visual standards. Muse Video supports deep referencing of images, audio, and video.
Clearly specify reference objects in your prompt, e.g. "reference the composition of Image 1" or "reference the action from Video 2".
The model automatically extracts core features from reference objects and combines them with text for creation, maintaining high fidelity while preserving creativity.
Image 1
@Image 1, friendly and joyful atmosphere, then the screen gradually blurs, displaying the text "Joy is in Muse Video" in the center.Image 1
Image 1
Image 1
@Image 1 are jogging on a school track in sportswear. The girl looks at the boy and says confidently: "We can definitely do it!" Cut to a close-up of the boy, who hesitantly replies: "Are you sure?" Cut back to a medium close-up of the girl, who says cheerfully: "Yes!" The mood is bright and determined. Speech bubbles appear around the speaking character with the dialogue.Image 1
@Image 1 and @Image 2. The girl is in a strawberry garden, picks one, takes a bite, and says with a smile: "This is the real deal!" A speech bubble appears around her containing the dialogue.Image 1,2,3
@Image 1, @Image 2, @Image 3, change the background to white. The camera is on a white table, the lens focuses on the camera in close-up, then slowly rotates around it, clearly showing the front, side, and back.Image 1,2,3
Image 1,2,3
@Image 1, @Image 2, @Image 3, generate a scene of her eating cake in a café.Logo & Reference
@Image 2, first show a mid-shot of her releasing silver floating lanterns with holographic projections, then pull back to reveal floating lanterns filling the sky. The image gradually blurs, then the Logo from @Image 1 appears. Overall style is 3D cyberpunk sci-fi animation.Multi-image Assets
Five-image Combo
@Image 4, bustling with customers. The girl from @Image 1 is wearing the outfit from @Image 2, organizing items on the counter. The boy from @Image 3 is a customer who walks up, wanting to ask for her contact information. The logo from @Image 5 is always displayed in the bottom-right corner.Storyboard
Character Storyboard
@Image 3. The girl is waiting for dad to finish cooking. She says: "아빠, 배고파요! 밥 다 됐어요?" Her appearance references @Image 1. Then the camera pans right to @Image 4's composition. Dad's appearance references @Image 2. Dad replies: "거의 다 됐어, 조금만 기다려!" Then cut back to a close-up of the daughter's slightly disappointed expression: "아직 멀었어요? 맛있는 냄새 나는데..." Then cut to dad's close-up: "이제 진짜 금방이야. “빨리빨리” 하지 말고 손부터 씬고 와!"Video 1
Character Image
@Video 1, generate a fight scene with @Image 2 and @Image 1. @Image 2 is the left character, @Image 1 is the right character. With intense background music.Video 1
@Video 1, generate a golden stallion galloping on a grassland, then freeze its magnificent running pose, transforming into a horse-shaped gold pendant.Video 1
Image 1
@Video 1, create a concept video of a tech park with the high-rise from @Image 1 as the visual center, also using a first-person diving perspective, highlighting the tech feel of the park in @Image 1.Video 1
Image 1
@Video 1, have the character in @Image 2 playing the flute while surrounded by the same particle effects.Video 1
Image 1
@Video 1, have the girl in @Image 1 grow the same wings, with the wing generation trajectory matching.Original Video 1
@Video 1.Original Video 1
@Video 1, keep the desktop clean and tidy, leaving only what they're holding.Original Video 1
Replacement Asset 1
@Video 1 with the face cream from @Image 1, keeping the motion and camera movement unchanged.Video 1
@Video 1: two late-arriving men run towards them, the five finally meet and chat happily.Video 1
@Video 1 forward, give the man in white an over-the-shoulder shot. He says: "It's not that bad. You're just stressed. Everyone goes through this, you just need to keep going."Video 1
Video 2
@Video 1, the moment the leaf touches the ground, golden particle effects burst out, a breeze blows through, cut to @Video 2.1. Be Specific: The clearer and more precise your prompt, the less likely you'll get unpredictable or bizarre results.
2. Less is More: Don't overload subjects or actions with too many modifiers — it blurs the focus.
3. Leverage References: When text alone can't describe a complex composition, camera movement, or effect, find suitable image and video assets to assist.
4. Physical Plausibility: Avoid describing physically impossible scenarios — the model relies on real-world physics to some extent.
[Scene type] + [Subject 1][State 1] + [Subject 2][State 2] + [Environment] + [Lighting & Mood] + [Camera Movement]