Skip to main content

Generation quality

Use auto for first passes, then lock in the model that worked for that creative style.

Brand pack efficiency

Keep a hero master at the highest resolution you need, then resize down — never upscale.

Project organization

Separate Studio exploration projects from finals so finished assets stay searchable.

Credit management

Start with short video durations to validate motion before paying for long clips.

​
How do I get the best results with the least credit spend?

Pass model: "auto" on first image passes. The router picks the best available model for your prompt. Once you find a variant you like, note the model ID from get_generation_history and use it explicitly for subsequent runs. This avoids re-exploring the model space on every generation.
Video credits scale with duration. Test your motion concept at the shortest available duration (typically 3–5 seconds) before committing to a longer clip. Only extend once the camera move and lighting read correctly.
Run score_creative on a new ad variant before increasing paid distribution. The predictive score surfaces likely performers so you allocate budget to variants most likely to convert, not just the ones that look good internally.

​
How do I keep brand assets organized?

Maintain a single master creative at the highest resolution your placements require — typically 4K or 2K. Use Brand Pack to resize down into smaller formats. Re-exporting from a low-resolution source degrades output quality.
Create two projects per campaign: one for experiments and AI drafts, one for approved finals. This keeps your asset library clean and makes it easy to hand off finished files without sifting through rejected variants.
ElarisLabs offers several video surfaces — URL to Video, Campaign video, Studio video nodes, and hyperlapse. Pick the right one early to avoid rework.

​
How do I keep agentic workflows safe and cost-controlled?

Create a separate brand-bound API key for each MCP client — one for Claude Desktop, one for Cursor, one for CI pipelines. Set a spend cap on each. If an experiment runs away, only that key’s budget is at risk; the others are unaffected.
Free tools like list_models, list_brands, and get_credits let you confirm connectivity and current balances before spending credits. Start every agentic session with a discovery call.