From title to finished visual in six clear decisions
Spark organizes editorial context and visual choices without requiring you to describe every pixel. AI proposes options; you select, review, and decide what is ready to publish.
1. Add the title and context
Start with the title or central idea. If a short phrase needs to appear prominently in the image, add the optional CTA and verify its exact spelling before generation.
2. Choose a platform, style, and references
Select the destination, such as YouTube, Shorts, square Instagram, Twitch, or Facebook, and choose from the available visual options. You may also use up to two additional images of your own, subject to rights and consent.
Choose from the models offered at that moment. Nano Banana Pro/Gemini, GPT Image 2, Grok Imagine, and Ideogram 4 may be selectable, but availability varies.
3 and 4. Generate and compare directions
Request one to three variations. Compare hierarchy, the relationship between text and image, reference fidelity, and fit with the topic instead of choosing only the most dramatic option.
Inspect details at full size and also view the asset small. Generated text, hands, objects, logos, and likenesses require human review.
5. Refine the selected direction
Use the available refinement workflow to adjust the result or explore another direction. The advanced manual editor remains In development and off by default; do not plan a delivery around controls that are not visible in your account.
6. Export or publish
Download the file in an available format and resolution. PNG and JPG are safe choices; WebP and AVIF depend on browser support. Free downloads include a watermark.
If the YouTube integration has valid credentials and permissions, Spark can set the thumbnail on an existing video. Otherwise, download the file and publish it through the destination channel.
Frequently asked questions
Do I need to write a technical prompt?
No. A clear title, context, platform, and coherent options can structure the brief. References and a CTA are optional additions.
Can I compare more than one idea?
Yes. The workflow offers one to three variations, based on your choice before generation.
Does Spark automatically post to every social network?
No. The confirmed connected action is setting a thumbnail on an existing YouTube video when credentials and permissions are configured. Other integrations and scheduling are future work.
Keep exploring
Turn context into visual alternatives
Prepare the title, choose the destination, and generate only the variations you intend to compare.