AI creation, your own references, and files prepared for each channel
Spark brings together decisions that are often split across briefing, generation, and file preparation. Model and integration availability can change, so the options shown in the product at creation time are the source of truth.
A focused brief and comparable variations
The content title starts the workflow. You can add a CTA as the main text, choose one to three variations, and upload up to two additional images of your own to guide the subject, product, setting, or visual language.
Variations are meant to explore different directions, not promise a specific outcome. Checking text, anatomy, brands, likeness, and alignment with the content is part of final approval.
Formats for video, social media, and music
The workflow supports the primary formats for YouTube, Shorts, square Instagram posts, Twitch, and Facebook. Album and single covers can be exported at up to 3000 × 3000 pixels, depending on the selected output.
PNG and JPG are the safest download choices across browsers. WebP and AVIF depend on browser support. True TIFF export is not currently available and should be treated as a future feature.
Selectable models with dynamic availability
Selectable options may include Nano Banana Pro/Gemini, GPT Image 2, Grok Imagine, and Ideogram 4. The actual set can vary with configuration, plan, provider capacity, or maintenance.
Models behave differently with composition, reference fidelity, and rendered text. Comparing each output with the brief is more reliable than assuming one model is ideal for every asset.
Connected publishing and editing: current status
When YouTube credentials and permissions are configured, Spark can set the thumbnail on an existing video. It does not create or upload the video, and it cannot bypass account or channel permissions.
Refinement is available within the creation workflow. The advanced manual editor remains behind a feature flag that is off by default and is In development. New integrations and scheduling are also future possibilities, not currently contracted features.
Frequently asked questions
How many images can Spark generate in one request?
You can choose one to three variations. The active option and its associated usage are shown in the interface before generation.
Can I use my own photos?
Yes. You may add up to two of your own images as references, provided you have the rights and permission to process them.
Which AI models are always available?
No model should be assumed to be permanent. Nano Banana Pro/Gemini, GPT Image 2, Grok Imagine, and Ideogram 4 may be offered, but availability is dynamic.
Is the advanced manual editor available now?
Not by default. It is In development and protected by a feature flag that is disabled in the standard workflow.
Keep exploring
Choose a format and compare visual directions
Use the options currently shown in the product to build a brief for your publishing destination.