docs(scenario): the game pilot, large-art and audio guidance

[gfargo]

What a real game's art showed, recorded in the Scenario docs. About 100 CU across scratch projects, every quote matched its bill, one sample per subject. The game's art is not included; only settings and measurements.

• New benchmark README (`benchmarks/provider-scenario-game-pilot`): verdicts per job. Poor for 32 px native props (different trees, holes from the cutout). Strong for 240 px landmarks (a ski lodge came back as a 240x208 native grid at high confidence with 63 colours, more detailed than the game's sprite, 18 CU, or 8 CU with the 1 CU Klein render) and for covers (GPT Image 2, then Pixelate, legible title text intact).
• Pixelate's grid is not literal on large inputs: 240 asked on a 960 px image gave 2 px blocks (480 cells) and a medium-confidence refine; 120 gave 4 px blocks and the exact 240 grid. Documented as an empirical note to test one image first.
• SCENARIO.md: a "Large and illustrative art" section with the working settings, a "Sound effects and other non-image models" section (MM Audio 1 CU, ElevenLabs 30 CU, both flat at any length; adapter is images only), flat GPT Image 2 pricing (11 medium, 44 high), and the chaining section now uses `revision.fromStyle`.
• PROVIDERS.md and the skill's `scenario.md` updated to match.

Docs only. Audio is not judged here; the shorter basket takes await the owner's ear.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
pixelkiln	Ready	Preview	Oct 6, 2026 2:32am UTC

[github-actions]

🎉 This PR is included in version 0.86.1 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀