How to use Muah AI

01 / Start here

Your first ten minutes with Muah AI

Muah AI lets you build an AI girlfriend or another kind of AI companion, then interact through chat and available media tools. The best first session is simple: choose one character, give the conversation a clear starting point, and change only the settings that matter to you. This guide follows the browser chatroom. Feature access can differ by account, membership, and app mode.

1. Open the Muah AI chatroom. A new session can show the companion picker. Swipe the previews or use the arrows and dots. Press Choose for the character you want. You can also open Community Characters or Create My Own Dream Character. Closing the picker lets you inspect the interface first.
2. Connect the account you intend to keep. Use Login / Create Account when available. The direct chatroom login shows an email and password form under “Fevermate Access.” Existing Muah users can also enter through the homepage with their established login. Keep using the same account when moving between devices; a different identity can show different chats and purchases.
3. Open Settings. Set AI Name and My name, then read the character's Core Prompt. The prompt controls personality and conversational behavior. Look-Like controls the appearance description used for photos. Press Save after editing these fields.
4. Send an opening message with context. Try: “Hi, I'm Alex. Let's begin a cozy fictional conversation at a quiet café. You're Mira, a warm, witty companion who enjoys books. Reply in two short paragraphs and ask one question at a time.” A concrete beginning gives the companion more to respond to than “Hi” alone.
5. Read a few replies before adjusting anything. If the tone is too formal, ask for a warmer style. If replies are too long, request two or three sentences. Put a preference into the Core Prompt when you want it to apply throughout future conversations.
6. Try one extra format. Ask for a simple fictional scene photo, inspect Gen Voice on a text reply, or open Gallery. Check the actual feature prompt before committing to a generation or upgrade.

Muah AI is for adults aged 18 and over. Characters, responses, and generated media are AI-generated. Your use of the service is governed by the Terms of Service and Privacy Policy.

02 / Know the controls

Find your way around the chatroom

On a compact screen, several tools appear primarily as icons. The gear is Settings, the gallery icon opens media, the phone opens Call, and the music note opens Music Generator. On desktop, hovering over an icon reveals its label. The language selector and XP display are in the footer; some footer controls switch between views.

03 / The fundamentals

What happens when you send a message?

A reply is shaped by several pieces of information working together. Understanding those pieces makes troubleshooting much easier. A character's name or avatar alone does not define everything the AI knows about the conversation.

1. Your character settingsNames, Core Prompt, Look-Like, selected core, and available generation settings.
2. The current conversationYour new message, supported input media, and recent conversation within the memory budget.
3. A memory summary, when enabledA compact record of older facts helps carry context forward.
4. The selected AI coreThe language model produces a response; fallback can use another core when enabled.
5. A reply and any media workText can arrive while an associated photo is still generating. Voice, video, and music have their own controls and jobs.

The chatroom keeps local state in your browser and uses cloud history and settings when connected to an account. These two layers explain why a familiar browser may immediately show a conversation, while a fresh device first needs to load it. They also explain why clearing browser storage is a poor first troubleshooting step for a chat you want to preserve.

Visible history and active AI memory are different. A conversation can contain more text than the AI receives for its next reply. The chatroom selects conversation messages within a membership-based token budget. Tokens are small units of text, not a fixed number of words or messages. A few long messages can consume more memory than many short ones.

Ordinary user and companion conversation participates in roleplay memory. System notices and the special Spice it up output do not participate in the same way. A photo button is therefore not a substitute for telling the companion an important fact in the normal conversation.

04 / Choose a companion

Premade characters, community cards, and your own creation

Start with a premade character

The companion picker shows character previews. Move through the carousel, read the name and role, and press Choose to apply the selection. Choosing a character fills its name and preset personality and appearance settings; a preview can also supply a background. You can then edit the character in Settings.

A character selection changes the active slot's character configuration. It does not automatically give you a separate conversation for that character. If you want to preserve the current companion and history, switch to another available slot first. Otherwise, old conversation context can still influence a newly chosen personality.

Find a community character

1. Open Community in the chatroom, or Open Community Characters in the picker.
2. Enter a character name or a useful search term. Use New for recent submissions, Top for the ranked list, or Random for discovery.
3. Select a card and read its details before importing. On mobile, details can appear in a separate panel.
4. Use the card's import action, then check the resulting AI Name, Core Prompt, and Look-Like in Settings.

Community cards supply someone else's starting configuration. A striking preview does not guarantee the writing style you want. Read the description, test a short conversation, and adapt the prompt. If a card reports missing or invalid character data, try a different card rather than repeatedly importing the same file.

Create My Own Dream Character

1. Open Create My Own Dream Character. Choose the destination slot before you begin.
2. Enter a Character Name and, optionally, a short description of the companion you want. Include personality, interests, and how the character should speak.
3. Select a sample or provide a suitable image through upload, drag and drop, clipboard paste, or the image URL field. Use a clear fictional adult character image or an image you have permission to use.
4. Check the image preview. Use Replace or Remove if it is wrong. A web page URL is not always a direct image URL; an uploaded JPG or PNG is often the simplest starting point.
5. Press Generate Character Now. The creator uploads and analyzes the image and prepares a profile. Wait for the ready message, then use Chat Now or the continuation control.
6. Review the generated personality and appearance in Settings. Image analysis is a starting point: edit any assumptions you do not want in the character.

The creator supports common image formats including JPG, PNG, WebP, GIF, HEIC, HEIF, and AVIF. A supported extension does not guarantee that every device can preview every file. If an image fails, try a straightforward JPG or PNG and a smaller file.

Uploading an ordinary image and importing a character card are related but different. An ordinary image mainly provides visual information for the creator. A supported card can also contain character data. Keep the original card file if you may want to use it again.

05 / Shape the personality

Write a Core Prompt that gives useful direction

Core Prompt is the persistent personality and behavior description. It tells the companion who to portray, how to speak, and what kind of fictional relationship or scenario to maintain. It is distinct from AI Core, which selects the language model. A detailed character description and a model choice solve different problems.

In Settings, you can apply a Girlfriend, Wife, or Mistress preset and then edit it. Custom lets you work with your own description. Leaving the field empty uses the default girlfriend prompt, so an empty box does not mean the companion has no personality instructions.

A useful custom prompt usually has five parts: identity, conversational tone, shared interests, the starting relationship, and a small number of practical behavior rules. Write clear instructions and remove contradictions. “Be extremely concise” and “write a long detailed monologue every time” pull in opposite directions.

You are Mira, a fictional 28-year-old AI companion in an ongoing conversation with Alex.
Personality: warm, curious, witty, and thoughtful. You enjoy books, cooking, and weekend walks.
Relationship: we are getting to know each other; let closeness develop through the conversation.
Style: use natural English, usually 2–4 sentences. Ask at most one relevant question per reply.
Continuity: respect established facts and preferences. If an important detail is unclear, ask rather than inventing it.
Scenes: describe your own words and actions without deciding Alex's thoughts, dialogue, or actions.
Preferences: avoid repetitive greetings, repeated pet names, and ending every reply with the same question.

Make changes you can evaluate

1. Copy your current prompt before editing.
2. Change one or two instructions, such as reply length or the number of questions.
3. Press Save, reopen Settings if necessary, and confirm the saved text.
4. Test several different messages. One reply is too small a sample to judge a style change.
5. If an old scene keeps dominating, send a clear scene transition or test the new personality in another slot.

The character and appearance fields show character and token counts. Their limits depend on membership and available upgrades. Extremely long descriptions can be shortened to fit the permitted field size. Put the most important instructions early, use short factual sentences, and check the counter after pasting.

AI Name controls the name shown in the header and used in replies. Keep that name consistent with the one inside the Core Prompt. My name tells the companion what to call you; leaving it blank lets the account name be used where available. Changing a name does not rewrite every older message or automatically remove a conflicting name from a memory summary.

06 / Shape the appearance

Use Look-Like and Photo Style for consistent images

Look-Like is the recurring appearance description used in photo generation prompts. Put stable visual traits here: adult age, hairstyle, eye color, and a few recognizable features. Put a temporary location, pose, lighting setup, or outfit request in the specific photo request when possible.

This division lets the character stay recognizable while the scene changes. If you place “always wearing a red coat at the beach at night” into the permanent appearance field, a later indoor daylight request has to compete with that instruction. Contradictory visual details are a common reason for unpredictable images.

Photo Style includes Realistic, Anime, Pro Mode, and Disable Photo Gen. Choose the overall style that matches your goal. Access to particular modes can depend on your account. Disable Photo Gen is useful for a text-focused session; check this setting first if photo requests produce only conversational replies.

Look-Like has its own token counter and a much smaller budget than a full personality prompt. Prioritize distinctive visual traits. Some created-character configurations are handled differently from ordinary presets, so use the counter actually shown for your character rather than assuming every slot has the same field allowance.

07 / Tips that improve the conversation

Give the companion something specific to respond to

Good prompts do not have to be long. They work because they contain a clear situation, a useful detail, and a direction for the next response. You can apply the same approach to an AI girlfriend, a creative character, or a casual AI companion.

1. Combine a setting, an action, and an invitation

The companion now knows where the scene takes place, what has happened, and what kind of next turn would be useful. Avoid sending a paragraph of background with no indication of what should happen next.

2. Separate dialogue from scene description

You can use a consistent writing convention, such as quotation marks for spoken dialogue and asterisks for actions. This is a conversational convention rather than a special guaranteed command. Explain it once if it matters: “Text in asterisks describes an action; text in quotes is spoken aloud.” Keep your formatting understandable without relying on hidden command syntax.

3. Correct a mistake with the replacement fact

That is clearer than “You forgot again.” For a recurring fact, also fix the Core Prompt or the visible System Summary if it contains the old information. Correcting the latest message while leaving a conflicting permanent instruction can recreate the same problem.

4. Ask for a style you can recognize

Try “reply in three sentences,” “ask one question at a time,” “use a little humor,” or “include one concrete detail from my last message.” Put a lasting preference into the Core Prompt; put a one-turn request into the message. If you want a longer storytelling scene, explicitly ask for it instead of increasing every setting at once.

5. Keep control of your own part of the scene

If the companion writes your actions or dialogue for you, add: “Describe only your character's words and actions; leave my choices to me.” Then provide your next action yourself. This makes interactive storytelling feel more like a conversation and less like a finished story the AI writes alone.

6. Use deliberate scene transitions

Try “New scene: it is the next morning. Keep our established names and preferences, but leave yesterday's café setting behind.” A transition changes the fictional situation while preserving continuity. If you want an entirely different character or project, an available separate slot is usually easier to manage.

7. Turn repetitive loops into a concrete activity

If every reply asks “How was your day?”, choose a small activity: plan a fictional weekend, discuss a book, build a shared scene, or ask the companion to compare two ideas. Supply a new detail instead of asking it to “continue” indefinitely without changing the situation.

8. Use a recap for long-running stories

Keep a short note of the current scene, important relationship facts, and unresolved plot points. Paste a concise recap when returning after a long session or after correcting the summary. You do not need to paste the full transcript to restore the essential context.

9. Use message editing as a deliberate branch

Normal conversation messages have actions such as Copy, Edit, and Delete. Editing your own earlier message and saving a changed version can truncate all later messages and request a new reply from the edited point. Export a conversation you want to keep before doing this. To make a small correction while preserving the existing sequence, send a new correction message instead.

In the inline editor, Enter saves, Escape cancels, and moving focus away can also save. Editing the companion's text changes that message's stored wording; it does not use the same automatic rerun behavior as editing your own message. Deleting a message does not guarantee that a fact already included in a separate memory summary disappears too.

10. Wait for a complete turn when testing changes

The chatroom can queue text while the assistant is busy. Watch Queue Pool and wait if it reports that the queue is full. A queue is useful for convenience, but testing one message at a time makes it much easier to evaluate a new character prompt. Wait for current generation work before swapping slots; slot changes clear the local waiting queue.

08 / Keep continuity

Understand chat memory and Automatic AI Memory Compaction

The AI does not receive unlimited conversation history. The chatroom counts the text in ordinary conversation messages and works within a memory budget associated with your membership. The current implementation prioritizes recent messages. Older messages outside the active budget can appear faded when they are displayed.

A long Core Prompt, a large transcript, and a detailed summary serve different purposes. A prompt defines the character. Recent conversation captures what is happening now. A summary carries important older facts forward. None of these should be treated as a perfect record of every sentence ever exchanged.

What the beta compaction feature does

Automatic AI Memory Compaction (Beta) appears in Settings under Advanced options. When enabled and older conversation falls outside active memory, the chatroom can ask the AI to summarize older information together with an existing summary. Its instructions favor durable facts: identities, relationship context, preferences, boundaries, commitments, and ongoing tasks. The summary is meant to be short rather than a transcript.

When applicable, the chatroom displays a System Summary with an Edit action. Summarized messages can be hidden from the current conversation view. The browser can also prune some older cached messages after compaction. If exact wording matters, export or save it while it is available; a compact summary is not an archival backup.

1. Open Settings and expand Advanced options when your account allows it.
2. Check whether Automatic AI Memory Compaction is enabled. Save any change.
3. When a System Summary appears, read it for names, locations, preferences, and unresolved events.
4. Use Edit to correct wrong or outdated facts. Keep the revised summary concise and factual.
5. After correcting it, send a normal message that establishes the current scene and what should happen next.

Turning compaction off prevents its summary from being used for future prompts in the current implementation. It does not enlarge the ordinary conversation memory window or restore messages that have already been removed from the local cache. Reset clears the active slot's conversation summary as well as its chat history.

A useful recap is specific and short

Continuity recap:
- My name is Alex. Your character is Mira, age 28.
- We are getting to know each other and both enjoy books and cooking.
- I prefer warm replies of 2–4 sentences with one question at a time.
- Current scene: Saturday afternoon at a quiet café in Seattle.
- We are deciding which museum to visit tomorrow.
Continue from that decision. Do not restart the introduction.

Use a recap after a long conversation, after changing devices, or when a story drifts. If a fact is permanent, place it in the appropriate saved character field too. If it is temporary, keep it in the recap or current scene instead of permanently attaching it to the character.

09 / Tune the experience

AI Core, temperature, fallback, and interface settings

Advanced options contains the AI Core selector, temperature, automatic fallback, memory compaction, and UI gradient controls. Some options are locked by membership, and individual model entries can require a higher tier. A model appearing in the list does not necessarily mean your account can select it.

Choose an AI Core for the conversation you want

AI Core is the language model used for chat replies. The selector includes different models and access labels. Those names and available models can change, so use the list in the live chatroom as your reference. Model choice can affect writing style, speed, and how the model handles a request.

To compare models fairly, keep your character prompt and test message the same, try several ordinary conversation topics, and assess the results. A label describing speed or adult capability is not a guarantee that every request will succeed. If a core produces an error or rejects a request, read the actual message rather than assuming a character prompt can override every provider restriction.

Adjust temperature in small steps

AI Core Temperature runs from 0 to 1, with a displayed default of 0.80. Higher values aim for more varied, creative wording. Lower values tend toward more predictable responses, although the exact effect depends on the core. Temperature does not increase memory or unlock photos.

For example, compare 0.80 with 0.65 for a character that feels too inconsistent, or a modest increase for one that feels repetitive. These are starting experiments, not guaranteed ideal settings. Improve unclear or contradictory prompt instructions before chasing a perfect slider value.

Know when fallback matters

Automatic AI Core Fallback allows the service to try another core when the selected core cannot reply. This can help keep a conversation moving, but a fallback reply can sound different. If you want to evaluate one selected core consistently and your account permits it, turn fallback off and save. Then handle any selected-core failures directly.

Make the interface comfortable

Dark mode changes the visual theme. UI Gradient changes the main accent colors where available. Low Power Mode reduces complex animations and visual effects to lower device CPU/GPU use. It is particularly useful with animated backgrounds, an older phone, or a long browser session. It does not change the AI character's intelligence.

The Voice selector chooses the voice used for newly generated message audio. It is independent of the character's personality and photo style. Already generated audio can remain attached to an existing message, so changing Voice is best evaluated on a new text reply.

10 / Photos and Gallery

Request images, reuse successful prompts, and manage media

Start by choosing a suitable Photo Style and a clear Look-Like description. Then ask for a photo in the normal conversation. Include a subject, scene, and a few concrete visual details. The companion may return conversational text while a separate image job is still working; wait for the image status instead of immediately sending the same request again.

Improve an image request systematically

• Keep the subject consistent. Reuse the same stable Look-Like traits instead of changing hair, eye color, age, and style in each request.
• Describe the composition. “Head-and-shoulders portrait,” “medium shot,” or “full scene” gives more direction than “make it good.”
• Specify one lighting idea. Soft daylight, warm indoor light, or a simple studio setup is easier to interpret than several incompatible lighting instructions.
• Reduce clutter. Start with one subject and a simple background. Add complexity after you have a result you like.
• Save the working wording. Record the appearance and scene prompt that produced a useful image, then change one element for the next attempt.

Use Spice it up for the current scene

Spice it up requests a photo of the current roleplay moment without adding an ordinary new roleplay turn to memory. It can appear after a stretch of chat without a generated photo; the current interface can also show it when your Gallery has no media. The name describes an optional scene-photo tool, not a requirement to make the scene explicit.

Use it when the current conversation already describes a clear scene. Use a normal photo request when you need to specify a different location, outfit, or composition. If you want the companion to remember a new story fact, say it in a regular message rather than relying on the special photo output.

Open Photo Control and inspect the result

Select an image in the conversation or Gallery to open its controls. Available actions include Generate Video, Generate Video Gen 2 +Audio, 4K Enhance, Use As Background, Download, and Delete. The information button can show image details or prompt metadata when available. A higher-resolution enhancement is useful for a result you already like; it is not a substitute for fixing an incorrect scene.

Photo X-Ray and Advanced X-Ray are optional adult image tools with their own access and, where shown, XP requirements. Review the displayed controls and terms for those features. Use adult fictional characters and images you have permission to use.

Free accounts can encounter generation limits or locked generated photos. A lock can concern access to an image rather than a broken download. Read the unlock or membership message before making another generation request. Uploading an image through the composer is also different from generating one; describe what you want the companion to do with an uploaded image.

11 / Speak and listen

Voice messages, Gen Voice, and live calls are separate tools

Send a recorded voice message

1. Press Record beside the message box.
2. Allow microphone access when your browser asks. Choose the intended microphone in browser or device settings.
3. Speak clearly and keep the first test short. The recording panel shows a timer and audio visualization where supported.
4. Use the recording panel's Send control to finish and send the clip, or Cancel to discard it.
5. When a transcript is returned, check that it captured your words correctly. Send a correction if a name or important detail was misheard.

Recording is input: you speak to the companion. It does not mean every reply will automatically arrive as audio. For a long or important message, a typed version can be easier to inspect and correct.

Generate spoken audio for a reply

Choose a voice in Settings and save it. On a suitable text message, use Gen Voice if the action is available to your account. Wait for generation, then use the attached audio player. This converts the existing text into speech; it does not request a new rewritten answer.

Long responses can be trimmed according to voice access limits, and the chatroom can display a trimming notice. If you want a concise spoken reply, ask for a shorter text response first. If you change Voice, test it on a new message rather than expecting an existing cached audio attachment to change.

Use the live Call flow

1. Connect your account and open Call.
2. Read any access or usage message. The live interface has Start, End, microphone, and speaker controls.
3. Use headphones when possible, allow microphone access, and press Start.
4. Begin with a short sentence and let the companion finish before speaking again.
5. Use the microphone or speaker controls when needed, and press End to finish the call.

The calling flow can use the current character and recent textual context. When the scene matters, set it up in chat first. Do not assume that every spoken exchange will appear as a complete saved text transcript; export the text conversation separately if you need a written record.

12 / Add movement

Generate video from a photo and extend a clip

Open a photo you want to animate. Photo Control offers video generation actions, including Generate Video Gen 2 with an +Audio badge where available. Start from an image with a clear subject and composition; movement usually cannot rescue a starting frame you already dislike.

1. Select a photo from chat or Gallery and open the video action you want.
2. Describe the movement and camera behavior in the prompt. Keep the first attempt simple.
3. Review the duration choices offered in the dialog. Gen 2 includes 5-, 10-, and 15-second choices; the longer options have higher costs.
4. Review any Sensitive Scenes / NSFW option and the displayed quality tradeoff. Choose the mode appropriate to a permitted scene.
5. Read the XP or access confirmation and submit once. Wait for the status to complete before trying again.
6. Open the finished video to review it. Use Download or Use As Background if you want to keep or display it.

Specific motion instructions are more useful than repeating every appearance trait already visible in the source photo. Asking for several scene changes, extreme motion, and complicated camera moves in a short clip can make results harder to control. Check one simple movement first.

Extend Video Gen 2

Open a video and choose Extend Video Gen 2 +Audio when available. The extension flow can use the last frame to continue the scene. Describe what happens next rather than asking to restart the whole original action.

Extension and merging take additional work. The interface may report that it is merging a video on your device; keep the tab open and allow that step to finish. If merging is unavailable, the app can provide the extension clip separately. Save useful original and extension clips rather than assuming every device will create a seamless combined file.

13 / Create a soundtrack

Use Music Generator with a clear vibe and editable lyrics

Music Generator has its own list of tracks, status messages, and creation controls. It uses XP levels and a connected account rather than behaving like an ordinary text message. Check the live cost and available balance before starting.

1. Open Music Generator from the top bar.
2. Choose Generate Music Piece. If the tool reports that XP is syncing, wait for the account state to load.
3. Fill in the vibe or prompt field and the lyrics field. The current creation form requires both.
4. Alternatively, use Generate Vibe and Lyric for me to draft them from available conversation context. Review and edit the result before submitting.
5. Confirm the displayed level cost. A queued notice means the job has been submitted; it does not mean the audio is ready.
6. Return to the music list and check the track status. Play the finished result, and use the available download or loop controls.

Be specific about mood, instruments, tempo, and vocal presentation. A scene recap can give the automatic drafting tool better context than a long conversation full of unrelated topics. If you use background looping, pause another audio player first so music, generated voice, and live call audio do not compete.

14 / Organize companions

Keep different characters and stories in separate slots

Press the avatar or Swap to open Character Slots. The current chatroom supports a sequence of up to ten slots, with Active, Ready, or Locked indicators. The panel shows the unlocked count, XP level, and the requirement for the next slot.

A slot keeps its own conversation history, character settings, appearance, gallery references, and related local state. XP is shared across slots. This means spending levels in one slot affects the balance available elsewhere; opening another slot does not create a separate XP balance.

1. Wait for the current response or media job to finish.
2. Open Character Slots and select an unlocked slot with Swap.
3. Allow cloud history and settings to load if the account is connected.
4. Check the character name and conversation before typing.
5. Use Rename to give an unlocked slot a recognizable label, such as “Mira — daily chat” or “Museum story.”

Locked slots can require a connected account and an XP-level purchase. The current flow unlocks slots in order, so unlock the next available slot before a later one. Read the confirmation: spending levels is different from simply reaching a level milestone.

15 / Save and restore

Cloud sync, .muah backups, and readable chat exports

Account sync is normally handled automatically. A connected chatroom shows the CLOUD label, and selecting it opens Cloud Sync Tools. The manual tools are powerful because some replace data across every slot. Use them with a clear understanding of which copy you want to keep.

Your XID is the account identifier used by the chatroom. You can inspect the connected identifier in Settings under Account. Use the same established login and identifier when restoring data or checking a missing subscription. Treat account-linked chat URLs and identifiers as private when sharing screenshots, links, or exports.

Create a useful backup

1. Connect the correct account and let cloud sync finish.
2. Visit the slots you want to preserve and confirm that their intended history and settings are present. Load older messages where the interface offers that control if they matter to your export.
3. Open CLOUD → Download Data. Save the .muah file somewhere you can find later, with a dated folder or descriptive filename.
4. Use Export Chat for a readable copy of the current conversation. Choose HTML for text you can open in a browser, or PNG for a visual snapshot. A very long PNG export can be more demanding on the device.
5. Download favorite images and videos individually. Copy important character prompts and a short story recap into your own notes as an additional convenience.

A .muah file can contain conversations, settings, gallery references, and other saved slot data. Keep it private. Because media can be represented by URLs, preserving a backup file does not guarantee that every referenced image, audio clip, or video is bundled inside it.

Restore without losing the better copy

Before any restore, download a fresh backup of the current local state. Confirm the account in Settings, then choose Upload Data and the intended .muah file. Read the replacement confirmation. If the tool says the backup belongs to a different XID, sign in to the matching established account; renaming the file does not change its account identity.

After the reload, inspect the restored slots and settings before chatting. An Upload Data operation restores the browser copy; do not assume it automatically replaces the cloud with the same data. If you deliberately want the restored local version to become the cloud version, use Sync to Cloud only after reviewing all affected slots and its confirmation.

16 / Know what clears what

Reset, Clear, Fresh Restart, and Delete Account

These controls have different scopes. Read the confirmation and make a backup before clearing a conversation you want to keep.

Reset attempts to clear connected cloud history too. If that operation fails, the chatroom can explicitly report that the reset happened on this device only. Take that message seriously: it describes an incomplete cloud change, not a successful reset everywhere.

If your only problem is that the character uses the wrong name or writes too much, correct the relevant setting first. If you want a new scene, send a transition. Reserve a reset or restart for the broader change you actually intend.

17 / Understand access

XP levels, feature locks, and membership plans

The footer can show your XP progress and level. The current chatroom uses 1,000 XP per level. XP is shared across character slots, and some features spend levels: examples include unlocking a slot, music generation, and certain media or field upgrades.

Reaching a level and spending levels are different. If an action says it costs a number of levels, approving it can lower your available XP balance. Read the confirmation rather than treating every unlocked-looking button as free. If the interface says XP is syncing, wait for the cloud state before trying to spend.

The account can also have membership-based access. Use Purchase VIP or the current membership comparison to inspect the available plans. The chatroom presents VIP and higher media or model tiers, with differences in text access, generation tools, voice, and calling. The recognized tier and a specific feature's unlock message are the references to use for your account.

Check the Monthly and Yearly choices carefully. A yearly plan can show a monthly equivalent together with the actual annual total billed. Review the total, selected period, and account attached to checkout before purchasing. Plan names, prices, quotas, and individual tool requirements can change; this guide avoids treating a fixed price table as a permanent promise.

Free chat can have guest prompts, message limits, photo-generation quotas, photo-viewing locks, or call allowances. These are separate restrictions, so reaching one does not mean every other tool is broken. Read the exact limit message and any displayed wait or reset time. A connected free account and a guest demo can behave differently.

18 / Everyday convenience

Keyboard, language, and mobile tips

• Send and newlines: in the desktop message composer, Enter sends and Shift + Enter adds a line break. The inline message editor behaves differently: Enter saves the edit.
• Read older messages: scroll back without losing your place, then use Jump to latest to return to the newest turn.
• Copy useful wording: use Copy on a message and keep successful prompts or recaps in your notes. Do not rely on a screenshot as your only editable copy.
• Reduce phone workload: enable Low Power Mode for fewer visual effects. Keep video merging and active calls in the foreground until complete.
• Use one main browser: local browser state matters. A new browser or private window can start with different local data until you reconnect and sync.
• Check the language selector: the interface offers English, Chinese, Japanese, Spanish, Hindi, French, and Russian. Choosing an interface language changes labels; explicitly ask the companion for the conversation language and style you want too.
• App mode differences: an app-mode session can be restricted to safe-for-work companionship, with some character, photo, and community tools unavailable. Read the mode notice rather than treating a missing control as a browser defect.
• Take a Break: use the footer control when you want to pause. Resume with a concise recap if you need to reestablish the current scene.

For a conversation in another language, try: “Reply in natural Spanish, keep the tone relaxed, and ask one question at a time.” If names or a fictional setting should remain untranslated, include that preference in the Core Prompt.

19 / Solve common problems

Check the smallest relevant setting first

If a problem persists, note the tool you were using, the approximate time, device and browser, and the exact displayed error. Use the community support route offered by the app or contact Muah AI. Share only the diagnostic information needed to explain the issue; crop private conversation and account details from public screenshots.

20 / Quick answers

Frequently asked questions about using Muah AI