Character References
Use clear references to guide faces, outfits and proportions, then inspect consistency in every result.
Reference to Video AI (Multimodal)
Guide a compatible model with reference images, clips or audio. This independent platform does not currently offer Pika H3 Flex itself.
Reference to Video AI
Guide a compatible model such as MiniMax H3 with reference images, clips or audio. Input support varies by model. Pika H3 Flex itself is not available on this independent platform.
Reference
Prompt
High-end cinematic racing ad. Feature the race car from the reference image as the main subject. Begin with a low-angle tracking shot over wet asphalt at sunrise, then accelerate through Barcelona and shift into a frozen London drift where splashing water becomes a glowing number-eight ice sculpture. Ultra-realistic materials, consistent vehicle identity and a warm-gold to icy-blue color transition.
Why Use
Reference to Video AI in Pika H3 Flex
References give the selected model visual and audio guidance for faces, wardrobe, products and environments. They can help reduce drift, but exact consistency is not guaranteed.
See the References and the Result Together
Illustrative examples pair prompts with clips, and include source images where supplied. They are not verified Pika H3 Flex outputs or a guarantee for the model currently selected. Review the recorded model when available and adapt the prompt to your own inputs.
All-in-One Reference Video Tool
From Reference Set to Finished Scene
Multimodal Inputs
Mix images, short clips and audio in one prompt and call each reference exactly where you need it.
Product References
Guide product shape, materials and packaging with your own images. Inspect logos and small details before publishing.
Faster Iteration
Reuse the same reference set for new scenes instead of rebuilding a shoot for every variation.
How to Create Video from References
Build a reference video in three steps: add your references, write the brief, then generate and compare.
Creative walkthroughs
Existing library media is shown as inspiration. Original full settings, timing and attempt costs were not recorded. The settings and revisions below are suggested starting points, not a claimed reproduction or new Flex benchmark.
Give each character reference one role
Library inspiration: Character campaign. Original model and full reference set are not recorded; this is not a continuity benchmark.
Existing inspiration clip
- Suggested starting settings
- Suggested new attempt: MiniMax H3 reference-to-video, 5 seconds, 768p, 16:9. Two-image estimate below excludes video/audio inputs; adding a setting image can change the quote.
- Input roles
- Picture 1: licensed front view defining identity and outfit. Picture 2: the same character from a compatible side view. Optional Picture 3: setting only. Use your own files; the inspiration poster is not an original reference set.
- What to check
- Compare face, outfit and body proportions across the clip; ensure the setting image does not introduce another person.
- If the result drifts
- If identities mix, remove conflicting references and test one short, fixed-camera action before adding another shot. Use the selected workflow’s exact reference syntax.
Starting prompt
Picture 1 defines the character’s face and outfit; Picture 2 shows the same character from another angle. The character walks to a table, places a notebook on it and stops. Fixed medium-wide camera, consistent soft light. Keep face and outfit recognizable. Footsteps and notebook contact only; no dialogue or music.
Room for every kind of creator.
One credit balance for images, video and audio. Pick a monthly or yearly plan, and top up with a credit pack anytime.
Lite
Perfect for beginners exploring AI
$9.99/ month$12.99
$119.88 billed yearly
3,000 Credits / month
- Videos
- ~61
- Images
- ~750
- Tracks
- ~136
- Access to video generation models
- Access to image generation models
- Music and audio creation
- AI Agent chat
- Commercial use under the Terms
- AI Workflow Infinite Canvas included
Secure checkout · cancel renewal anytime
Pro
PopularIdeal for high-frequency creation
$24.99/ month$34.99
$299.88 billed yearly
10,000 Credits / month
- Videos
- ~204
- Images
- ~2,500
- Tracks
- ~454
- Access to video generation models
- Access to image generation models
- Music and audio creation
- AI Agent chat
- Commercial use under the Terms
- AI Workflow Infinite Canvas included
Secure checkout · cancel renewal anytime
Ultra
Designed for extreme production
$79.99/ month$129.99
$959.88 billed yearly
30,000 Credits / month
- Videos
- ~612
- Images
- ~7,500
- Tracks
- ~1,363
- Access to video generation models
- Access to image generation models
- Music and audio creation
- AI Agent chat
- Commercial use under the Terms
- AI Workflow Infinite Canvas included
Secure checkout · cancel renewal anytime
Reference to Video AI FAQs
Not yet. This is an independent platform, not the official Pika service. MiniMax H3 is a separate model, not Pika H3 Flex. Choose one of the listed models and check its supported inputs, settings and Credit cost before generating. For the official Flex service, start at pika.art/pika-h3-flex.
Reference to Video uses supported images, clips or audio to guide a new video. The selected model determines supported reference types and limits; character and product details can still vary between runs.
It depends on the selected model. The upload area shows the image, video and audio limits for the current model before you generate.
Insert each uploaded reference into the prompt where it matters, for example the character in one shot and the product in another.
Clear, well-lit references and consistent instructions give the best results. Results can still vary between models and runs.
Waiting time depends on the queue, provider, selected model, resolution and clip duration. We have not measured a typical completion time for this integration. Follow the existing task’s status in My Work; waiting alone does not mean it failed.
The Credit estimate updates with the model, duration and resolution you choose and is shown before every generation.
Paid outputs generated with a Lite, Pro or Ultra subscription, or a one-time Credit Pack, may be used commercially under our Terms of Service, the selected model's terms and applicable intellectual property law. You must have the rights to your inputs and remain responsible for how you use the output.
Completed videos appear in My Work and My Creations. Download copies you need to keep; storage availability and media-link lifetimes can vary, and the service is not a permanent archive.
Create Stunning AI Videos
All in One Stop
Choose a compatible model, add your references and review the Credit cost before generating.
Free Credits and example video costs
What does the signup offer cover?
Eligible new accounts can receive up to 10 signup Credits. Granted signup Credits expire after 3 days.
Grants are subject to eligibility and the daily allocation limit, so registration does not guarantee a grant. Your Credit history shows the amount actually received and its expiry. Signup Credits are not unlimited and may be insufficient for a video.
These are estimates for one new attempt with the listed models. The draft uses text only; the image example uses one first frame; the reference example uses two still images. No video or audio references are included.
| Workflow | Model and settings | Estimated Credits | Pack-price equivalent (USD) | Signup allowance comparison |
|---|---|---|---|---|
| Short text-only draft | seedance-2.0-mini 4s · 480p · 16:9 | 30 | $0.18 | Full signup grant is insufficient |
| Single-image animation | minimax-h3 5s · 768p · 16:9 | 122 | $0.73 | Full signup grant is insufficient |
| Two-image reference shot | minimax-h3 5s · 768p · 16:9 | 122 | $0.73 | Full signup grant is insufficient |
USD equivalents use the current 5000-Credit Pack at $29.99. They allocate part of a pack purchase, not a separate per-video cash payment. Subscription rates and other packs differ; actual remaining allowance may be lower.
A successful result uses Credits even if you dislike it. Budget for several attempts: three successful attempts at the same settings use about three times the listed Credits. Failed generation tasks return their Credits. Video/audio reference duration and other inputs can affect the final quote.