Help

• Be specific.
  ❌ "Write something about marketing"
  ✅ "Write a 200-word LinkedIn post about email marketing trends for B2B SaaS, professional tone"
• Provide context.
  ❌ "Is this a good price?"
  ✅ "I'm comparing cloud hosting in Europe for a small SaaS app with ~1000 users. Is €49/month for a 4-core VPS a good price?"
• Assign a role.
  ❌ "Help me with this code"
  ✅ "You are a senior Python developer. Review this function for performance issues and suggest improvements"
• One chat per subject. Start a new conversation when you change topics — it keeps responses focused and avoids confusing the model with unrelated history.
• Break down complex tasks. Instead of "Build me a full website," start with "Design the navigation structure," then "Write the homepage copy," then "Create the CSS layout." Step by step gets better results.
• Iterate. If the first answer isn't right, refine your prompt rather than starting over. Tell the AI what to change: "Make it shorter," "More formal tone," "Focus on the cost comparison."
• Context limits. Every AI model has a context limit. Very long conversations may lose earlier messages — start fresh when things get long.

Auto lets AiFuse pick the best model for each message. Select Auto from the mode strip and just type — AiFuse analyzes your prompt, classifies it (coding, creative, factual, analysis, or image), and routes it to the highest-scored model for that category. A purple "Response by [Model]" banner shows which model was selected. Each message is classified independently, so the model can change as the conversation shifts topics.

The Tier selector controls quality vs. cost: Lite filters to budget models, Balanced filters to mid-priced models for the best quality-per-dollar (default), and Pro unlocks the premium flagships like Claude Opus and GPT-5.4 for top-tier quality. If you ask for an image, it generates inline in the chat — no need to switch to Art Studio. You can switch to Single mode anytime for manual control.

Auto is the default for new accounts, but Single gives you manual control. You pick one model, have a conversation, and switch to a different model anytime — even mid-conversation. Your chat history stays intact when you switch. Different models have different strengths — some are better at coding, others at writing or analysis. Experiment and use what works best for each task. Simple, fast, and the cheapest way to work.

Keep in mind that different models have different pricing. Switching from a budget model to a premium one mid-conversation means subsequent messages will cost more. The cost is always shown after each response.

Duo sends your prompt to two models at the same time. You see both responses side by side, pick the one you prefer, and that response becomes the conversation context for your next message. It's useful when you want a second opinion, when you're not sure which model handles a topic best, or when quality matters more than speed.

Example: You ask both models to write a function that validates an email address. Each model takes a different approach — select the one you prefer and continue building on it.

When you use Duo with a context level other than None, it enters interactive mode. After both models respond, a "Pick this" button appears below each reply. Selecting one makes that response the shared context for the next message — both models will see it as the conversation history going forward. This lets you build a conversation where you always keep the best answer. With context set to None, there's no Pick this — each message is a fresh standalone comparison.

Fusion pairs two models on every message. You choose two AIs — an orchestrator and a consultant. Both respond to your prompt in parallel, then the orchestrator combines both perspectives into one strong, refined answer. Use it when quality matters most and you don't mind the extra cost.

Example: You ask a question about a legal contract. Both models analyze it independently, then the orchestrator combines insights from both into one comprehensive answer — catching things either model might have missed alone.

The input bar includes several tools to help you work with AI more effectively.

• Attachments — Attach files for the AI to read and analyze. Drag and drop, paste from clipboard, or use the paperclip icon. Useful for reviewing documents, extracting data from images, summarizing reports, or debugging code. Supported file types depend on the model — if a model doesn't support your file type, you'll see a clear error message. You can attach up to 5 files per message, each up to 10MB. Files are never stored on AiFuse servers — they are sent directly to the AI provider and immediately discarded.
• Prompt Builder — Opens a guided form to structure your message. Choose from suggestion cards (Compare, Draft, Voice, Analyze) or set a custom role, instructions, and output format. The builder combines everything into a well-crafted prompt. Great for complex requests or when you're not sure how to phrase something. Make sure you've selected the right model before using the Prompt Builder — your prompt is sent automatically when you confirm.
• Incognito — Toggle Incognito mode on or off. See the Incognito mode section above for full details.
• Send — Send your message to the selected model. While the AI is responding, the Send button becomes a Stop button — tap it to cancel the response at any time.

Context controls how much conversation history is sent to the AI with each message. More context means the AI remembers more of your conversation, but costs more since you're sending more tokens.

AiFuse offers four levels:

Each mode has a different default:

You can change the context level at any time from the dropdown next to the model selector. The cost difference between Economy and Quality can be 3–4x on longer conversations. If the AI starts forgetting earlier details, try a higher context level — or start a fresh conversation.

Duo has two sub-modes depending on context: with Economy, Balanced (default), or Quality selected, Duo runs in interactive mode — you pick your preferred reply each turn with the 'Pick this' button, and that response becomes the shared context for the next message. Select None and Duo switches to compare mode — each message becomes a standalone, independent comparison with no conversation memory.

Projects let you group conversations with shared context. Create a project from the Projects button in the sidebar — give it a name, an optional system prompt (instructions that apply to every conversation in the project), and pin reference files that all conversations can access.

After each conversation, AiFuse auto-generates a summary so the project "remembers" what was discussed. When you start a new conversation in a project, the AI sees the system prompt, pinned files, and previous summaries — building continuity across sessions. You can also move existing conversations into a project from the history panel. Auto-summaries cost approximately $0.0005 each; conversations use standard model pricing.

Talk to AI in real time using your microphone. Two modes are available:

• Hands-Free — The AI automatically detects when you stop speaking using voice activity detection (VAD). Just talk naturally — the AI responds when it senses a pause. Best for flowing conversations.
• Tap-to-Talk — Tap the microphone button once to start speaking, tap again to stop. You control exactly when the AI should listen. Best when you need to think between turns or are in a noisy environment.

On desktop, hold the spacebar as a shortcut to toggle the microphone in Tap-to-Talk mode.

In Hands-Free mode, you can adjust how long the AI waits after you stop speaking before it responds:

• Fast (0.8s) — Responds almost immediately after you pause. Best for quick back-and-forth.
• Normal (1.5s) — Waits a natural pause before responding. The default — works well for most conversations.
• Relaxed (2.5s) — Gives you more time to think or gather your thoughts between sentences.

Each voice provider offers a selection of AI voices. Pick the one you prefer from the voice selector before or during a session:

• Grok — 5 voices — Ara, Eve, Leo, Rex, Sal. You can switch voices mid-session.
• OpenAI — 8 voices — Alloy, Ash, Ballad, Coral, Echo, Sage, Shimmer, Verse.

The voice provider is determined by the model you select — Grok models use Grok voices, OpenAI models use OpenAI voices.

Art Studio lets you generate images and videos from text prompts. Switch between Image and Video mode using the toggle at the top. Each mode shows only the models and parameters relevant to that type.

Type a description of what you want, pick a model, and hit Generate. The estimated cost is shown before you generate. Each model has different strengths — experiment to see which fits your style.

Available parameters depend on the model you select. Not every model supports every option:

• Aspect Ratio — Choose the shape of your output — square, landscape, portrait, or widescreen. Available ratios vary by model.
• Quality — Some models offer quality tiers (Standard, HD, etc.). Higher quality produces more detailed results but costs more.
• Background — Toggle between opaque and transparent backgrounds. Only available for image generation on supported models.
• Duration — For video generation — set the length in seconds. Available range depends on the model.

Resolution, output format, and image count options may also appear depending on the model's capabilities.

The Exclude field lets you specify what you don't want in your image. For example: "blurry, text, watermark, low quality."

Not all models support negative prompts — the Exclude toggle only appears when the selected model has this capability.

Some models accept a reference image to guide the style or composition of the output. Upload an image and the AI will use it as visual inspiration.

The reference image option only appears when the selected model supports it. This is different from img2img — it influences style, not direct editing.

The Settings page (accessible from the user menu) lets you customize your experience:

• Default mode: Choose which chat mode (Single, Duo, or Fusion) opens by default.
• Context presets: Set the default context level for each mode independently.
• Default model: Pick a preferred model that's pre-selected when you start a new conversation.
• Appearance: Switch between Light, Dark, and Auto (follows your system) themes.
• Language: Switch the interface between English and Español.
• Change password: Update your account password.
• Export data: Download all your account data as a JSON file.
• Delete account: Permanently delete your account and all associated data.

• Your prompts are sent only to the AI provider you select — nothing is shared across providers or used for advertising.
• Payment is handled entirely by Stripe — AiFuse never sees or stores your card details or address information. The less we know about you, the better you're protected.
• Incognito mode lets you chat without saving history — conversations only live in memory and are gone when you leave or refresh.
• Don't paste sensitive information like passwords, financial details, or personal identifiers into prompts — while AiFuse doesn't store them, the AI provider will process them.

Context vs Incognito: These are independent settings. Context controls how much history the AI sees. Incognito controls whether your conversation is saved. You can use both together — for example, Balanced context with Incognito on means the AI sees recent history during the session, but nothing is saved when you leave.

For full details, see our Privacy Policy.

Quick fixes for the most common problems:

• Message didn't send: Your balance may be too low for the estimated cost. An "Add Funds" link appears in the error — click it to go to the Usage page and top up. In Art Studio, the Generate button is disabled when your balance is below the estimated cost.
• AI forgot earlier details: Your context level might be too low. Try switching to a higher level or start a fresh conversation.
• Response seems wrong or incomplete: Try a different model — each has different strengths. Or rephrase your prompt with more detail.
• Costs seem higher than expected: Long conversations on Quality context cost more. Switch to Economy for simple tasks, or start a new conversation to reset the context.
• Voice won't start: Make sure your browser has microphone permission. Check that no other app is using the mic. Try refreshing the page.
• Voice says insufficient balance: Voice requires a funded account and is not available during the free trial. Add funds to enable voice sessions.

Found a bug, want to request a new model, or have ideas for improvement?
Contact us and we'll help sort it out.