> ## Documentation Index
> Fetch the complete documentation index at: https://docs.runable.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Save Credits Tips
> Practical tips to get more out of your credits on every task.
Every action in Runable costs credits. Some actions cost a lot more than others. This page covers the most effective ways to stretch your credits further without sacrificing quality.
## Use canvas tools instead of chat prompts
This is the single biggest credit saver. When you need to edit an image or video that is already on the canvas, use the canvas toolbar instead of typing a request in chat.
### Direct canvas edit (cheaper)
Select an image on the canvas, click **Remove BG**, **Text Edit**, **Mark & Edit**, or **Upscale** from the toolbar. The operation runs directly on the image without going through the agent.
### Chat prompt (more expensive)
Type "remove the background from this image" in the chat. The agent processes the message, calls the tool, and applies the result. You pay for the agent's reasoning plus the operation.
**Direct canvas edits use up to 10x fewer credits** than sending the same request as a chat prompt. This applies to:
| Canvas tool | What it does | Why it saves credits |
| ------------------- | --------------------------------------------- | ---------------------------------------------------- |
| **Remove BG** | Removes background from an image | Runs the operation directly, no agent reasoning cost |
| **Text Edit** | Detects and changes text in an image | Same operation, no chat message cost |
| **Mark & Edit** | Draw regions and give per-region instructions | Targeted edit, no agent overhead |
| **Upscale** | Increase image resolution (1x-4x) | Direct operation, no prompt processing |
| **Edit Image** | Refine an image with a prompt | Runs image model directly, skips the agent loop |
| **Edit Video** | Edit a video with instructions | Direct video model call |
| **Remove Video BG** | Remove video background | Direct operation, no agent overhead |
| **Motion Control** | Apply motion from a reference video | Direct canvas operation |
Whenever you see a button for it on the canvas toolbar, use the button. Save chat prompts for tasks that need the agent's reasoning, like building something new or making complex multi-step changes.
## Mute videos you do not need audio for
Videos with AI-generated audio cost roughly **2x the credits** of muted videos. If you are creating a product teaser, a social media clip, or any video where you plan to add your own music or voiceover later, turn off audio during generation.
Toggle the **Mute** button in the video metadata options before generating. The UI shows a tooltip: "It takes half the tokens to generate a muted video."
Create a 10-second product showcase video for my bakery's new croissant. Close-up on a marble surface, warm lighting, slow rotation. 9:16 for Instagram Reels. Generate muted. I will add background music separately.
## Choose the right model tier
Not every task needs the most powerful model. Using Lite for simple tasks saves significant credits compared to Max.
| Task | Recommended tier | Why |
| ----------------------------------------------------- | ------------------------------------------- | ---------------------------------------------- |
| Fix a typo on your website | Lite | Simple edit, no deep reasoning needed |
| Change a color or font | Lite | Straightforward CSS change |
| Generate a standard landing page | Pro | Good balance for most builds |
| Create a presentation deck | Pro | Handles layout and content well |
| Build a complex app with auth, payments, and database | Max | Needs deep reasoning to get architecture right |
| Write a detailed research report | Max | Needs thorough analysis |
Lite costs the least per message. Pro is the sweet spot for most work. Save Max for projects where getting it wrong would cost you more in credits than the extra cost of Max.
## Use lower resolution for drafts
Higher resolution costs more credits for both images and videos.
* **Video:** 720p costs less than 1080p, which costs less than 4K. Generate at 720p while iterating on the concept. Once you are happy with the result, regenerate the final version at higher resolution.
* **Images:** Start with the default resolution. Only upscale when you need the final version for print or large displays. Upscaling from 2x to 4x can cost up to 17x more.
## Use Plan Mode to avoid wasted work
Plan Mode costs a small amount of credits for the planning step, but it can save you a lot by preventing the agent from building something wrong. If a complex project goes sideways 3 steps in, you have already spent credits on work you need to redo.
Click **Plan** before sending complex requests. The agent writes a plan, you approve it, and it builds correctly the first time.
## Save your brand to memory
Every time you describe your brand colors, fonts, and preferences in a prompt, the agent processes those extra tokens. Save your brand identity to memory once, and every future conversation starts with those defaults already loaded.
Remember this for all future work: My business is Flour & Fold bakery. Brand colors are cream (#FFF8F0) and dark brown (#3D2B1F) with coral (#E07A5F) accents. Font is DM Sans. Tone is warm and inviting. Address: 218 Court St, Brooklyn NY 11201. Instagram: @flourandfold.
This saves tokens on every future prompt, which saves credits.
## Write detailed prompts upfront
A vague prompt like "make me a website" leads to back-and-forth. Every message costs credits. A detailed first prompt that includes your requirements, content, style preferences, and structure gives the agent everything it needs to get it right in fewer messages.
### Vague (more rounds, more credits)
"Make me a website for my bakery."
Agent asks 4-5 questions. You answer. It builds. You ask for changes. 6-8 messages total.
### Detailed (fewer rounds, fewer credits)
"Build a website for Flour & Fold bakery with a hero, menu section, about us, and contact form. Cream and brown colors, DM Sans font, warm tone."
Agent builds on the first try. 2-3 messages total.
## Quick reference
| Tip | Estimated savings |
| -------------------------------------------------- | --------------------------------------------------- |
| Use canvas tools instead of chat prompts | Up to 10x per operation |
| Mute videos when you do not need AI audio | 2x on video generation |
| Use Lite for simple edits | 3-5x vs Max |
| Generate at 720p while iterating, upscale later | 2-4x on video |
| Use Plan Mode for complex projects | Prevents entire rebuild costs |
| Save brand to memory | Saves tokens on every prompt |
| Write detailed prompts | Fewer message rounds |
## Next steps
See how credits work and what each plan includes.
Learn about the canvas workspace where direct editing saves the most credits.