# The best way to turn one blog post into a week of content with AI

_By Apoorv Khanna, November 16, 2026_

Give your agent the post (a launch note or a transcript works too) and have it write the LinkedIn, X and newsletter drafts, using its own model or a second one through Vaaya's `llm` tool. Then add one matching image with `fal/generate` and, if you want audio, a short narration with `deepgram/speak`. Everything comes back as drafts and files for you to review. Nothing in this workflow posts to LinkedIn or X, so publishing stays your step.

## What "best" means for this job

- **Faithful to the source.** A rewrite must never add a number or a promise the post did not make.
- **Shaped for each channel.** A LinkedIn post, an X thread and a newsletter section are different formats. Each draft should fit its own.
- **Cost per asset you keep.** Text costs little. The image and the audio are the paid generations, so the number of attempts matters.
- **You release it.** The agent stops at drafts.

## The options in the catalog

Prices are as listed on 16 November 2026. The [catalog](/catalog) has current prices.

| Tool | What it returns | Price per call | When to pick it |
| --- | --- | --- | --- |
| `llm` tool | Text from a second model, with `model_used` and `charged_usd` | Per token, at the model's own rate, with no markup | A draft or review from a second model |
| `vaaya/onescrape` | The page as markdown, one row per URL | 2¢ per URL | The source is a URL and you lack the text |
| `deepgram/transcribe` | A transcript of an audio or video file | 1¢ per declared audio minute, minimum 2¢ | The source is a recording |
| `fal/generate` with `nano-banana-pro` | One image as a hosted URL | 33¢ | The default hero image |
| `fal/generate` with `gpt-image-2-5--flare--text-to-image` | One image with readable text in it | 10¢ | A card that shows the headline |
| `fal/generate` with `seedream--v4-5--text-to-image` | One image as a hosted URL | 4¢ | Bulk or iteration, when its quality already meets the brief |
| `deepgram/speak` | A hosted MP3 | 1¢ per 250 characters, minimum 2¢ | A draft take of the narration |
| `fal/generate` with `elevenlabs--tts--turbo-v2-5` | A speech file in a preset voice such as `Liam` | 5¢ per 1,000 characters, rounded up | The take you publish |

## The default path

**Text.** Let the agent you already run write the drafts. It has the post in context, and Vaaya adds no charge. Use the `llm` tool when you want a second model: Vaaya's guidance lists "draft with a cheap model, review with a good one" as a two-call pattern. Pass `cheap`, `mid` or `best`, or an exact slug from 340+ models.

**Image.** One `fal/generate` call. Vaaya's media guidance picks on quality first and uses price to break ties, so its default for marketing images is `nano-banana-pro` at 33¢. Step to `gpt-image-2-5--flare--text-to-image` when the image has to show the headline.

**Audio.** Optional. The same guidance treats `deepgram/speak` as the budget option, fine for drafts and utility narration, and prefers the ElevenLabs model for a polished brand voiceover. Check the script by ear with a draft take, then pay for the take you will publish.

Step-by-step guides: [how to generate and revise an image](/blog/how-to-generate-images-with-vaaya) and [how to create a voiceover](/blog/how-to-create-a-voiceover-with-vaaya).

## One post, five drafts, priced

Arithmetic at listed prices on 16 November 2026. The source is a launch post saved as a file, so no read call is needed. The week is one piece a day: two LinkedIn posts with different angles, one X thread, one newsletter section and one narration.

- Four text drafts and a 400-character narration script, written by your own agent: 0¢ through Vaaya.
- One hero image with `nano-banana-pro`: 33¢.
- A draft narration with `deepgram/speak`: 400 characters is two blocks of 250, so 2¢.
- The final narration with `elevenlabs--tts--turbo-v2-5`: up to 1,000 characters is 5¢.

The image call:

```json
{
  "service": "fal",
  "action": "generate",
  "params": {
    "model": "nano-banana-pro",
    "prompt": "Editorial photo of a paper calendar on a wooden desk, morning light, empty space on the left for a headline",
    "aspect_ratio": "16:9",
    "resolution": "4K"
  },
  "max_cost_cents": 33
}
```

Total: 33¢ + 2¢ + 5¢ = 40¢. Without audio it is 33¢. With the headline card in place of the hero, the image line is 10¢ and the total is 17¢.

A second-model review is the one line without a fixed price: the `llm` tool meters tokens at the model's own rate, with no markup. Vaaya's guidance puts a typical call under 0.1¢ on the cheap tier, at 0.1¢ to 1¢ on mid and at 1¢ to 3¢ on best. A long post on a top model can cost more, and each call returns `charged_usd`.

## How to run it

Connect your agent with the [install guide](/install), then give it a request like this:

> Use Vaaya to turn `content/launch.md` into a week of drafts: two LinkedIn posts with different angles, one X thread and one newsletter section. Match the voice of the two samples in `samples/` and use only facts from the post. Make one 16:9 hero image to go with them. Write a 400-character narration script and voice it. Save everything in `drafts/`, then stop. Do not publish or send anything. Ask `consult` for the calls and tell me each price before you spend.

The agent can ask Vaaya's `consult` tool to plan the paid steps. It returns the exact calls, each `use` call carries a `max_cost_cents` ceiling, and a call priced above its ceiling is refused before it runs. The `llm` tool has no per-call ceiling. It is bounded by `max_tokens`, the limit on the agent's key and a $10 per day inference cap on the account.

## What to check, and where this falls short

Before anything goes out:

- **Facts.** Compare every number, name and claim in each draft with the source post.
- **The image.** Open it at the size it will be shown. `nano-banana-pro` takes an aspect ratio and a resolution tier and does not accept pixel sizes, so generate 16:9 and crop. Vaaya's size table lists 1200 by 627 for a LinkedIn shared image and 1200 by 630 for an OpenGraph card. Read any text in an image letter by letter.
- **The file URL.** Images and audio come back in the same response. Save the URL then, because running the call again is a new paid generation.
- **The audio.** Listen to the whole take. The `text` you send must be only the words to be spoken, and `deepgram/speak` accepts up to 2,000 characters per call.

The limits:

- **It does not post.** The catalog has no action that publishes to LinkedIn or X. Held for review means the drafts sit in your files or your chat until you paste them in.
- **No assembled video.** The one workflow `consult` plans that edits a narrated video together is a product demo built from a screen recording. Here you get an image and an audio file to combine in your own editor.
- **Chat connectors.** The `llm` tool is not listed in the claude.ai and ChatGPT connectors. There the agent writes the drafts itself.

## Why this suits a pay-per-call wallet

Repurposing comes in bursts: a few calls on the day a post ships, then nothing until the next one. The job also crosses text, image and speech models, and one prepaid balance covers all of them without a separate account and key for each. Give the agent a key limited to the Media and LLM router categories with a weekly limit in [spending controls](/spending-controls), and a regeneration loop stops at a number you chose. Pack sizes are on the [pricing page](/pricing).

## Questions

**Can an AI agent post to LinkedIn or X for me through Vaaya?**

No. The Vaaya catalog has no action that publishes to LinkedIn or X. Your agent returns drafts, an image and optional audio, and you post them yourself.

**What does it cost to repurpose one blog post?**

At prices listed on 16 November 2026, one nano-banana-pro image is 33¢ and a 400-character narration is 2¢ with deepgram/speak or 5¢ with the ElevenLabs model. Drafts your own agent writes carry no Vaaya charge. Text from the llm tool is billed per token at the model's own rate, with no markup.

**Which model should write the drafts?**

Start with the agent you already use, because it has the post in context. For a second model, Vaaya's llm tool takes cheap, mid or best, or an exact model slug. Each call is one-shot, so include the post in every prompt.

**Can the image include my headline?**

Yes. Vaaya's routing guidance sends images that need readable text to the gpt-image-2-5--flare--text-to-image model on fal/generate, listed at 10¢ per image. Read the lettering before you use it.
