tolstoy-studio

Tolstoy Studio — create product content

Tolstoy Studio turns a plain-language brief into brand-ready marketing media. The generate_studio_content tool hands the brief to Tolstoy's Studio Agent, which already owns scriptwriting, model casting, scene direction, pacing, and brand-safety. Your job is to convey the user's intent and pick the output format — not to engineer the creative.

What you can make

• PDP product video — short, product-forward video for a product detail page.
• Social / UGC video — TikTok / Reels / Shorts style, hook-first.
• Ad creative — video or image sized for paid placements.
• Product / lifestyle image — on-model or in-scene stills.
• Animate a product photo — turn a static product image into a short motion clip.

How to drive it well

1. Pass the brief verbatim. Relay the user's product/brand intent as written. Do NOT rewrite named brands/products into generic descriptions, and do NOT append lighting/camera/scene boilerplate — the Studio Agent adds all that. A tight, honest brief beats a long engineered prompt.

2. Choose assetType: "video" for clips, "image" for stills.

3. Choose aspectRatio by destination:

   Destination	aspectRatio
   TikTok / Reels / Shorts / mobile PDP	9:16 (video default)
   Instagram / Facebook feed	4:5 or 1:1
   YouTube / desktop / on-site widget	16:9
   Square / generic	1:1 (image default)
   If the user didn't say where it's going, use the default, mention it, and ask only if genuinely ambiguous.

4. Reference images = public URL only. If the user gives a public HTTPS image URL of the product, pass it in referenceImageUrls (up to 4) so the output features the real product. To animate a photo: assetType: "video" + the photo URL in referenceImageUrls + a brief like "subtle motion, keep the product centered." If they paste/attach an image with no URL, describe what you see in the prompt instead — the tool does not accept uploads.

5. Refine with iterate_studio_content, don't regenerate. For "punchier hook", "make a 16:9 too", "warmer tone", call iterate_studio_content with the chatId from the prior result. It keeps the full creative context; a fresh generate loses it.

6. Report back. Generation is asynchronous — share the returned chatId and tell the user it's rendering in the Tolstoy canvas.

Examples

"Make a TikTok for my matcha whisk — photo: https://cdn.example/whisk.jpg" → generate_studio_content { assetType: "video", aspectRatio: "9:16", prompt: "TikTok-style video for my matcha whisk", referenceImageUrls: ["https://cdn.example/whisk.jpg"] }

"Animate this product shot: https://cdn.example/serum.jpg" → generate_studio_content { assetType: "video", aspectRatio: "9:16", prompt: "Animate this product with subtle motion, product stays centered", referenceImageUrls: ["https://cdn.example/serum.jpg"] }

"Make it 16:9 with a stronger hook" → iterate_studio_content { chatId, message: "16:9 version with a stronger opening hook" }

Notes

• Insufficient-credits error → tell the user to add credits in Tolstoy billing.
• This skill creates content; it does not publish to a store or ad platform.