How to Use Render Mode - Cliptude Docs
Black Friday: Cliptude Unlimited LIFETIME DEAL is 70% off — $299.70 instead of $999. Use code BLACKFRIDAY at checkout.Get 70% off →
Cliptude Cliptude Docs

How to Use Render Mode

Render mode is for when the voice and the pictures are handled outside Cliptude. Your agent plans the video shot by shot, and Cliptude only aligns, builds and renders it. Cliptude searches for nothing, writes nothing, voices nothing and generates nothing, so the price is flat: 15 credits per minute. You don't have to supply everything yourself: if you don't have the narration or the footage, your agent can make or find it.

At a glance

  • The material: a voiceover recording and a picture or clip for every shot. You can supply them, or your agent can make and find them (see step 2).
  • Your agent brings: the plan, meaning the look, which file goes under which sentence, each graphic and its text, the cuts and the sound. It can also write the script, produce the narration and source the footage.
  • Cliptude does: aligns the narration to the word, builds the locked plan and renders it.
  • Price: 15 credits per minute, so a 10-minute video is 150 credits. Re-rendering a finished video costs 10% of the job's credits.
  • Works with: Faceless Documentary, Top 10 and Vector Explainer. Avatar presenters aren't available in render mode.

1. Connect your agent and install the skills

Render mode runs through an agent such as Claude, Codex or Cursor. Connect it to the Cliptude MCP server and sign in, or use the CLI if your agent has no MCP.

https://mcp.cliptude.com/mcp

npm i -g @cliptude/cli
cliptude auth login

npx skills add cliptude/cliptude

Step-by-step connection instructions for each agent are in the API docs. The cliptude-plan skill holds the render mode rules, so the agent already knows them.

2. Get the material: yours or your agent's

Render mode needs a voiceover and a picture or clip for every shot. Where they come from is up to you. If you have them, hand them over. If you don't, your agent can get them.

The voiceover

  • You have a recording: give your agent the file or a public link. It is the whole soundtrack.
  • You don't: your agent can write the script and turn it into narration with a text-to-speech tool or key you have, such as ElevenLabs or OpenAI. Or it can use Cliptude's own voices (create_voiceover), which are charged at the normal voice rate on top of the render. Either way it then uploads that audio as the voiceover. Render mode needs at least 6 minutes of audio.

The pictures and clips

Plan on about one clip for every 4 to 6 seconds of narration. The same few clips over and over read as cheap. Your agent can source all of it. It picks the places from your topic and each shot, from the list the cliptude-plan skill gives it. A history documentary leans on archives, a product story on stock or your own footage, a news topic on fair-use excerpts.

  • Free stock: Pexels and Pixabay, photos and video, with an API key you create. Both allow commercial use without credit.
  • Stock you subscribe to: Storyblocks, Vecteezy, Shutterstock, Getty Images and iStock, Adobe Stock, Envato Elements, Pond5, Unsplash+ and Artgrid, through their API or downloads with your own key or account.
  • Public domain archives: NASA, US government works, the Library of Congress, the US National Archives, Wikimedia Commons and the Internet Archive. Check each file's own licence.
  • Your own browser: your agent can also work in sites you are signed in to. See Using your own browser below.
  • Short fair-use excerpts of a few seconds from YouTube or other video sites, when the narration is about them. Credit them on screen or in the description. You decide and are responsible.
  • Your own material: your footage, photos, screen recordings and documents.
  • Pictures it generates with an image or video model you have a key for.

Keys for these sources go in your agent's environment, never in the plan. The skill ships a sourcing.env.example that lists them. Cliptude never sees them.

Rules that apply to everything

  • Permission to use each file. Cliptude stores and renders what is sent, so every file must be one you may publish. Read stock licences for identifiable people and brands. A download service gives you no rights to what it fetches.
  • Credit your sources. Ask the agent to note where each picture came from.
  • Clean pictures under graphics. A graphic that sits on a picture shows through to it, so don't put one over a screen capture or any picture with text in it.

Using your own browser

If your agent can drive your Chrome, it can source the way you would. It can only do this when its host provides browser control: the Claude in Chrome extension, or Chrome DevTools or remote debugging through an MCP server such as chrome-devtools-mcp.

What it can do there:

  • Search Storyblocks, Shutterstock, Getty or any other site you are signed in to, and download the licensed file.
  • Find a public-domain photo on an archive page.
  • Capture a frame or screenshot of something the narration is about.

The ground rules:

  • It asks you first, before it starts.
  • Only your accounts and sites you have access to.
  • Files come the site's own way: its download button or API, within that licence.
  • It stops at a paywall, CAPTCHA or sign-in it doesn't have, rather than getting around it.
  • It sends each file to Cliptude by uploading it (upload_plan_media or cliptude upload-media), or on a remote host with a public link.
  • It notes where each picture came from in its summary, so you can credit it.

To use it, add a line to your prompt such as: "You may drive my Chrome to source footage from the sites I'm signed in to. Ask me before you start."

3. Brief your agent

Copy this prompt, fill in the parts in [brackets] and paste it into your agent.

Connect the Cliptude MCP server at https://mcp.cliptude.com/mcp and sign in. If you can't use MCP, install the CLI with `npm i -g @cliptude/cli` and run `cliptude auth login`. Then install the Cliptude skills: `npx skills add cliptude/cliptude`.

Make me a [10]-minute Faceless Documentary video about [topic] in render mode only (render_only=true, 15 credits/min flat).

Cliptude only builds and renders. It makes no paid calls of its own: no TTS, no stock search, no AI images. So you get the material.

Voiceover: [path or URL to my narration audio, OR "none: write the script and make the narration for me"]. If I have none, research the topic, write the script, and produce the audio with [a TTS tool or key you have, or Cliptude's create_voiceover]. It must be at least 6 minutes. Upload it as the voiceover.

Footage and pictures: [what I'm supplying, OR "source it all yourself"]. Source what's missing following the sourcing section of the cliptude-plan skill. Decide where to look from the topic and each shot: stock sites, public-domain archives, fair-use excerpts, or my signed-in Chrome if you can drive it (ask me first). Use only files I may publish, and keep a list of where each came from so I can credit them. Ask me for any API key you need.

Plan every shot following the cliptude-plan skill's render-only rules:
- Every picture slot is one of my URLs or `keep`. Use no `fallback_query`.
- Every plate is `{"beat": true}` or one of my URLs.
- Use no presenter windows.

Run `estimate_credits(render_only=true, output="plan", voiceover_seconds=…)` and `check_plan` until the plan is valid. Show me the plan and its cost before you lock or build anything. After the build, review every picture and graphic before rendering.

The agent stops to show you the plan and its cost before anything is built.

4. What happens next

  1. Upload. The files, whether yours or sourced by the agent, go to Cliptude, and nothing has to be public. The agent uploads the narration first, then the pictures and clips onto the job. A public https link also works for any file.
  2. Narration. Cliptude transcribes your voiceover and returns every sentence with its start and end time, down to the word.
  3. The plan. The agent assigns one of your files to every shot and writes the graphics, cuts and sound around them. Library music, graphics, overlays, transitions and sound effects cost nothing extra.
  4. Check and lock. The free check rejects anything that would need a search: a picture slot with no file, a fallback_query or a presenter window. Fix, then lock.
  5. Build. Cliptude builds exactly the locked plan. If a file can't be loaded, the job fails and names it rather than quietly searching for a replacement.
  6. Review, then render. The agent looks at every picture and graphic as built, you approve, and the video is rendered.

5. Good to know

  • Your clips play muted. Your voiceover is the soundtrack.
  • Clips hold still unless the plan sets a camera move.
  • Short clips freeze on their last frame. If a shot is longer than the clip you give it, split the sentence into several shots or use longer clips. Trim long clips with a start and end time.
  • A graphic's picture slot shows one image. A clip placed there shows as a single frame.
  • A graphic without a picture shows the footage under it. Nothing is generated.
  • Upload links last an hour. Start the job with your narration within the hour.

6. Render mode or the normal workflow?

Choose render mode when you already have the narration and the footage. If you want Cliptude to find the footage and record the voice, use the normal agent-planned workflow instead: see How to Make Documentaries on Claude. Credit prices for both are on the pricing page.