Marketers
Brief, drafts, visuals, and the reporting that follows: one workspace, every channel.
Go from a product brief to a scripted, shot, and assembled ad in every format you need.
What is being advertised.
Who it is for.
Where it will run.
Target seconds.
Lines, claims, or offers that must appear.
Yours, so the ad shows the real thing.
Make an ad. - Product: [your product] - Audience: [your audience] - Platforms: instagram,tiktok - Length: 20s - Must include: [your must include] Work in this order, and stop for approval after step 2. Everything after it costs real money. 1. **Script**: a hook in the first 2 seconds, one idea, one call to action. Write for the ear, not the page. 2. **Storyboard**: a `media.timeline` with `mode: "storyboard"`: one clip per shot with its prompt and duration, plus the voiceover and music lanes. WAIT. 3. **Assets**: generate the shots; use `animate-image` over full video generation wherever a still with a camera move will do. When product photos were supplied, COMPOSITE them, never re-generate the product: a model asked to picture a product it has never seen invents a different one, and the ad then advertises something that does not exist. 4. **Voice**: `inference speech` for the VO, then mix with a ducked bed. 5. **Assemble**: cut to the plan, burn captions (most feed video is watched muted, so uncaptioned is unwatched), and export EVERY requested platform aspect from the same master. 6. **Deliver**: one Drive folder, one file per platform, plus the script. Never state a claim the user did not give you. Inventing a statistic or a superlative for an ad is not creative licence; it is a legal problem for them. ## Delivery Write the finished artifact into Drive. Never upload a user's content to a third-party file host (`0x0.st`, `catbox.moe`, `transfer.sh`, pastebins, any off-idapt URL). On a computer, render to `/tmp/output/` and `computer file-download` it into Drive. If you catch yourself reaching for `curl -F`, stop: you already hold the artifact.
Copy, visuals, and reporting live in three different tools.
Draft copy, generate the images, and file everything in the same workspace the reporting agent reads.
One model's voice does not fit every channel.
Run the brief through two or three models in parallel and keep the best take per channel.
Drop the brief and brand assets in Drive; every chat and agent works from the same source.
More about DriveOne prompt, several models: pick the strongest angle per channel instead of settling for one voice.
More about Multi-Model ChatProduce channel-sized images beside the copy they belong to; iterate in the same thread.
More about Image GenerationGPT, Claude, Gemini, Grok, Llama: pick mid-conversation, cheapest-first.
DALL·E, Imagen, FLUX, Stable Diffusion: same composer, same files.
Folders, previews, and lightweight versioning for AI work.
Delegated runs with schedules, bounded execution, and full progress traces.
Skills, scoped permissions, persistent memory: any model you pick.
Real-time info, cited sources, every model: no per-provider tax.
Try 200+ models free for 7 days, then pick the plan that fits. Cancel anytime.
Try nowMonday reporting eats Monday.
An automation compiles the week's numbers and drafts the update before you sit down.
A scheduled agent assembles the weekly update and files it in the campaign folder.
More about Automations