Show the model what matters
Guide a character, setting, or movement with images, video, and audio. Pika supports up to 12 references; describe the role of each instead of expecting every detail to carry over perfectly.
Try Pika H3 FlexPika H3 Flex AI Video Generator
Your idea. Directed in motion.
Explore video with synchronized sound. Start with a scene, shape the camera move, and use references to guide the next take.
Native audio · Reference-guided direction · Draft-to-finish workflow
Interface preview · No generation or charges on this page · Illustrative template footage
Pika H3 Flex Image-to-Video: Direct the Motion
A useful prompt describes what moves, how the camera follows, and what you hear. These template examples illustrate shot ideas; they are not verified Pika H3 Flex outputs.
PromptTurn a quiet American roadside still into a cinematic story with desert light, natural character motion, and smooth camera continuity.
Pika H3 Flex: References, Audio & Creative Control
These are documented Pika H3 Flex capabilities. Availability here depends on the generation service connected to this preview. Media below is illustrative.
Give the scene its own sound
Generate synchronized stereo audio with the picture. Pika’s hosted API also offers separate dialogue, score, SFX, foley, and ambience stems for further mixing.
Try Pika H3 FlexDirect the shot, then the sequence
Describe a tracking shot, a close-up, or a cut to the next scene. Multi-shot prompts help explore a short narrative; complex action and continuity still need review.
Try Pika H3 FlexDraft first. Finish the chosen take.
Explore at native 480p or 768p. Pika’s hosted workflow can upscale the selected take to 1080p, 2K, or 4K; high-resolution finishing is not native 4K generation.
Try Pika H3 FlexA clearer brief. A better first take.
Use this preview to plan a shot. In a connected Pika H3 Flex workflow, iterate on a draft before spending time on the final export.
Step 1: Define the subject
Describe the scene or add an image. Say which details matter: the product silhouette, the outfit, or the setting.
Step 2: Write the action and sound
Specify one camera move and a clear action. Add dialogue, ambience, or music cues to give the scene a rhythm.
Step 3: Compare, then refine
Change one part of the brief at a time. Check motion, reference fidelity, and sound before choosing a take to finish.
What will you make first?
Build a scene around one moment
Plan an opening, an action, and a final beat. Use Pika H3 Flex to explore how the scene feels in motion, then review the transitions.
Credits
Buy once, create when you need. Purchased credits never expire.
Buy once, create when you need. Purchased credits never expire.
Trial
$9.90 USD
200 credits
Up to 20 basic 5s videos at 480p, without prompt enhancement.
Buy creditsStarter
$19.90 USD
500 credits
Up to 50 basic 5s videos at 480p, without prompt enhancement.
Buy creditsCreator
$49.90 USD
1400 credits
Up to 140 basic 5s videos at 480p, without prompt enhancement.
Buy creditsStudio
$99.90 USD
3200 credits
Up to 320 basic 5s videos at 480p, without prompt enhancement.
Buy creditsNo subscription. Verified accounts receive 10 welcome credits once. Generation reserves credits; confirmed failures return them. Refunds revoke the corresponding purchased credits.
Three details worth getting right
A specific creative brief gives you something useful to compare between takes.
What must stay the same?
Name the subject and its defining details. Review those details in every take.
What should move?
Separate the subject’s action from the camera move. Start with a simple shot.
What should we hear?
Describe the ambience, dialogue, and music as deliberately as the image.
FAQ
Model capabilities, practical limits, and what this preview actually does.
Yours isn't here? Email the teamPika H3 Flex is Pika’s post-trained MiniMax H3 model for video with synchronized audio. This independent preview introduces the model and lets you explore a prompt interface; it is not the official Pika service and does not generate video.
Pika’s seven-second figure measures transformer computation for a 15-second 768p clip in a controlled benchmark. It is not total request time: queueing, input processing, finishing, and service overhead can add time.
The model supports text, endpoint frames, and mixed references. Reference mode requires an image or video; audio alone is not enough. This homepage currently lets you try a prompt and one local image, without sending either to a generation service.
Open weights support native 480p and 768p video with mixed audio. Higher-resolution upscaling and separate audio stems are Pika-hosted features. Local inference needs suitable hardware and is subject to the model license.
The plans and credit amounts on this page are placeholders, not Pika API prices. No payment or credit deduction happens here. Check the connected provider’s current prices, terms, and rights to your inputs before using generated work commercially.
Your next shot starts
with a clearer idea.
Describe the subject, movement, and sound. Explore the interface before generation becomes available here.
