Prologue
Start here — how to generate images with your bot
How to Generate Images
Your trigger command is the word that starts every image generation. Type it followed by a description of what you want to see.
Every Memescale bot has a trigger command — a single word that tells the bot to generate an image. You chose this word during setup. If your character is named "Hedge", your trigger is probably "hedge".
To generate an image, type the trigger followed by a description of what you want. The bot splits your message on the first space — everything before it is the trigger, everything after it is your prompt.
/hedge a puppy in a space suit
hedge a puppy in a space suitBoth formats work identically. The bot matches triggers with or without a leading slash, and matching is case-insensitive. "/Hedge", "hedge", and "/HEDGE" all do the same thing.
When the bot receives your prompt, it sends a brief flash message — something like "Processing prompt. Hedge Inbound." — that auto-deletes after a few seconds. Then the generated image appears as a reply to your original message.
The text after the trigger is your prompt. More detail produces better results. Instead of "hedge standing", try "hedge standing on a rooftop at sunset overlooking a city skyline". Section 0.3 below covers prompt writing in detail.
If you type the trigger and nothing happens — no flash message, no image, no error — your group is not authorized yet. Read Section 0.2 below.
Unlocking Groups for Generation
All groups are locked by default. Your bot will not generate images in any chat until you explicitly unlock it.
Your bot ships with an empty allowed list. This means zero groups can generate images — not your DMs, not any group chat, nothing. This is intentional. It prevents unauthorized usage and spam. You need to unlock at least one group before your bot will generate anything.
This applies to every bot, including free trials. There are no exceptions — every bot owner needs to unlock groups themselves.
If You Are the Owner (Fastest Method)
The owner is whichever Telegram username you entered during setup. That specific account has a shortcut: send /getid in any chat where the bot is present, and the bot will immediately authorize that chat for generations. The bot replies with "Authorized ({chat_id})" — that group is now unlocked. You're done.
If the authorization API call fails for any reason, the bot falls back to just showing you the numeric chat ID. If that happens, use the manual method below.
The owner's Telegram username is the ONLY account that can auto-authorize via /getid. When anyone else sends /getid, the bot discovers the group and replies with the numeric chat ID, but does NOT authorize it. That group still needs manual authorization through the dashboard.
Manual Authorization (For Non-Owners or Dashboard Preference)
- Send /getid in the group where the bot is added. The bot replies with the numeric chat ID and records the group internally.
- Open your Memescale dashboard and go to the Groups tab.
- Find the group under "Discovered Groups" — it will show the group name, type, and member count.
- Click Authorize. The dashboard writes the change to the bot's config and restarts the bot automatically.
- The group is now unlocked for image generation.
Verifying It Worked
Send /whoisactive in the group. The bot replies with "1" if generation is active in that group, or "-" if it is not. The slash is optional and matching is case-insensitive.
If you get "-", go to the Groups tab in your dashboard and confirm the group actually appears in the authorized list. If it does not, authorize it again.
Text commands (like /ca or /website) work in all groups regardless of authorization. Only image generation requires an unlocked group.
For the full authorization workflow including bulk authorization and revoking access, see Section 3.1 Group Authorization.
Writing Better Prompts
The more descriptive your prompt, the better the result. Here is how to write prompts that produce good images.
The bot processes your prompt text exactly as written — there is no special syntax beyond the trigger word. The quality of your output depends almost entirely on how much detail you provide.
| Quality | Example Prompt | What You Get |
|---|---|---|
| Basic | hedge standing | A generic standing pose with a default background. The model fills in everything you left out, and the results are unpredictable. |
| Better | hedge wearing a suit in an office | A recognizable scene with a specific outfit and setting. The model has enough to work with. |
| Best | hedge wearing a tailored navy suit sitting in a leather chair in a dimly lit executive office, cinematic lighting | A detailed composition with specific clothing, furniture, environment, and lighting. This is where the model shines. |
What to Include in Your Prompts
- Setting and background — where is the character? A rooftop, a beach, a neon-lit alley, a medieval castle.
- What the character is doing — standing, sitting, running, holding something, looking at the camera.
- Clothing and accessories — a leather jacket, sunglasses, a crown, a lab coat.
- Art style — cinematic, anime, watercolor, pixel art, photorealistic.
- Mood and lighting — golden hour, dramatic shadows, soft glow, noir, overexposed.
Build your prompt from the inside out: start with the character and what they are doing, then add the environment, then add the style and mood. Each layer gives the model more to work with.
For generation timing details, see Section 15.1 Generation Times. For common prompting mistakes, see Section 18.3 Common Mistakes.
Quick Reference
Essential commands and where to find more detail in this documentation.
Processing Order
When someone sends a trigger command, the bot checks in this order: blocked users first, then group authorization, then per-user rate limits, then generation. If any check fails, the request is silently dropped — no error message is sent.
Where to Go Next
Overview
Introduction to the Memescale platform
What is Memescale?
Memescale provides custom-trained, autonomous AI image generation bots for Telegram and X (Twitter) with truly unlimited generations and zero credit systems.
When you purchase a Memescale bot, you're getting a fully autonomous AI agent trained specifically on your character that operates independently on Telegram or X. We handle the entire technical pipeline—from fine-tuning LoRA weights on your training data to deploying the inference system to configuring webhook integrations.
Your bot comes with a dedicated profile including custom username, profile picture, display name, and bio that you control in real-time through the Memescale dashboard.
Models run locally on our dedicated GPU infrastructure with pre-compiled weights, eliminating the per-generation costs that plague API-based alternatives. One flat monthly fee covers everything.
| Metric (monthly) | Shared | Dedicated |
|---|---|---|
| Image Capacity | Unlimited | Unlimited |
| Per-Image Cost | $0 | $0 |
| Credit System | None | None |
| Usage Caps | None | None |
*Theoretical maximum if prompts were sent continuously every second of every day.
Platform Architecture
Memescale runs on isolated, DDoS-resistant GPU servers where all ports are closed except for cryptographically protected SSH tunnels.
Our security model is built on complete port closure with the sole exception of cryptographically protected SSH tunnels that connect our VPS to GPU servers.
- Telegram webhooks run on entirely separate infrastructure
- Communication occurs exclusively through encrypted SSH tunnels
- Air-gapped architecture prevents unauthorized access and DDoS attacks
- Zero open HTTP endpoints or API dependencies
Even if external services like AWS experience outages, your bot continues operating because the entire generation pipeline exists on infrastructure we own and control directly.
Local Model Execution
Local model execution means every component of the inference chain lives on our hardware in remote locations. Pre-compiled LoRA weights are stored directly on GPU-equipped servers alongside the base diffusion models, eliminating network latency and external API calls.
- Prompt parsing happens locally
- Image generation runs on dedicated GPUs
- Delivery occurs without third-party services
- No throttling, rate limits, or unexpected failures
Key Differentiators
Memescale eliminates friction through transparent pricing and immediate deployment capabilities.
Competitors charge $0.04-$0.10 per image or impose monthly generation caps of 1,000-5,000 images, creating financial anxiety for community-driven projects.
| Feature | Memescale | Competitors |
|---|---|---|
| Pricing Model | Monthly fee | Monthly + credits |
| Training | $0 | $199-$899+ |
| Limits | Unlimited | Capped |
| Hidden Fees | None | Overage charges |
| Profile Changes | Instant via dashboard | Support tickets required |
| Infrastructure | Owned hardware | Shared cloud |
Operational Friction Eliminated
- Profile customization happens instantly through the dashboard
- Update username, display name, or bio with real-time changes
- No support tickets required for standard modifications
- Telegram bots deploy within 1 day, X agents within 2 days
Fair Queue Distribution
The round-robin queue system on shared compute rotates one image at a time between projects, regardless of individual queue sizes. Maximum four projects per GPU ensures fair distribution.
Getting Started
Account setup and first steps
Account Creation & Payment
Payment automatically creates your account—no manual signup required.
Simply enter your information during checkout and your dashboard becomes accessible immediately. You'll receive a Telegram notification when your bot training completes.
Onboarding Process
- Complete the contact form with your Telegram username or X handle
- Upload 8-20 images of your character through the interface
- System validates image quality and minimum file size requirements
- Proceed to cryptocurrency payment
- Receive Telegram notification when training completes
- Test your bot in a private evaluation group
Payment Methods
Memescale accepts only cryptocurrency payments including major stablecoins and crypto assets across multiple networks.
Test the bot's output quality in a private group. If you're not satisfied, request a full refund. Once you approve the bot, your subscription begins and next payment is due in exactly 31 days.
Choosing Your Compute Option
Choose between shared and dedicated compute based on your community size, budget, and performance requirements.
| Feature | Shared Compute | Dedicated Compute |
|---|---|---|
| Price | $199/month | $449/month |
| Projects per GPU | Up to 4 | 1 (exclusive) |
| Queue System | Round-robin | No queue |
| Generation Time | 11 seconds | 5 seconds guaranteed |
| VRAM Allocation | Shared | 100% |
| Multi-Character | No | Up to 8 characters |
Shared Compute Details
Shared compute at $199 monthly distributes one GPU across a maximum of four projects using a round-robin queue system.
If one project has ten prompts queued and your community member sends a single prompt, once the current generation completes, yours executes next regardless of how many prompts other projects have waiting.
Dedicated Compute Details
- 100% VRAM allocation with zero sharing
- No throttling under any circumstances
- Sub 5-second generation times guaranteed for Telegram
- Support for up to 8 characters with dynamic LoRA hotswapping
- Higher resolutions available upon request at no additional cost
Smaller communities or those testing viability often start with shared compute, while established projects with high engagement opt for dedicated.
Image Upload & Training
Upload 8-20 clear, consistent images of your character from multiple angles through the dashboard.
Training Data Requirements
- 8-20 images showing your character from multiple angles
- Front views, side profiles, three-quarter angles, different expressions
- High quality without blur or heavy occlusion
- Consistent styling across all images
- Minimum 1MB file size per image
Training data quality directly determines model output accuracy. The system validates that each image meets requirements before proceeding to payment.
Professional dataset creation is available as an option for users struggling with dataset quality. Model quality is Memescale's primary marketing mechanism, so delivering stellar results is at the forefront of the entire business operation.
First Generation Test
Test your bot immediately using your custom trigger command once training completes.
| Platform | Typical Response Time |
|---|---|
| Telegram (Shared) | 11 seconds |
| Telegram (Dedicated) | 5 seconds |
| X (Twitter) | 7-37 seconds (includes polling) |
Evaluation Process
- Receive Telegram notification when training completes
- Join the private evaluation group as an admin
- Use your custom trigger command followed by a prompt
- Generate multiple test images with varied prompts
- Assess character consistency and style accuracy
- Approve the bot or request a refund
If the output quality doesn't meet your expectations, request a full refund—your payment will be returned in full. Alternatively, proceed with one of your two free revision rounds.
Once you approve the bot, your subscription officially begins and the next payment becomes due in exactly 31 days from your initial payment date. Most users identify potential issues within 10-15 test generations.
Bot Configuration & Controls
Self-service management for Telegram bots
Notification Preferences
Receive Telegram notifications when configuration changes occur, rate limits trigger, or commands are used.
Owner Configuration
After your bot is provisioned, you must send /start to the bot in a private Telegram DM. This registers your personal chat ID with the bot so it knows where to deliver owner notifications. Until you complete this step, the bot cannot send you direct messages.
- Open Telegram and search for your bot's username
- Send /start in the private DM
- The bot confirms your chat ID has been registered
- You will now receive notifications based on your selected preferences
Notification Types
| Type | Triggers When | Default |
|---|---|---|
| All | Any configuration change, rate limit event, or command usage | Enabled |
| Rate Limits Only | A user hits their daily generation cap | Off |
| Commands Only | A custom text command is triggered by a user | Off |
| Profile Changes | Bot profile fields are modified through the dashboard | Off |
| Off | No notifications sent | Off |
Notification preferences are configured per bot from the Management tab in the dashboard. Changes take effect immediately — the bot reads the updated configuration on the next incoming request.
When you modify groups, rate limits, commands, or characters through the dashboard, the bot sends a Telegram notification confirming the change was applied and the service was restarted.
Custom Text Commands
Define custom trigger-response text pairs that your bot responds to in any Telegram chat — no group authorization required.
Custom text commands function in every group the bot is added to, regardless of whether that group is authorized for image generation. Authorization only gates the image generation trigger — text commands are always active.
How Text Commands Work
Each command consists of a trigger word and a text response. When a user sends the trigger word (with or without a forward slash), the bot replies with the associated text. Commands are managed entirely from the Commands tab in the dashboard.
Common Use Cases
| Trigger | Response Example | Purpose |
|---|---|---|
| ca | 0x7a250d5630B4cF539739dF2C5dAcb4c659F2488D | Share contract address |
| website | https://yourproject.com | Link to project website |
| socials | Twitter: @project | Telegram: @projectchat | Social media links |
| buy | https://app.uniswap.org/#/swap?outputCurrency=0x... | Direct swap link |
| roadmap | Phase 1: Launch. Phase 2: CEX listings. Phase 3: ... | Project roadmap summary |
Managing Commands
- Add commands by specifying a trigger word and response text in the dashboard
- Edit existing commands inline — modify the trigger or response at any time
- Toggle commands on or off without deleting them
- Delete commands permanently with a single click
- Changes sync to the live bot immediately after saving
Users can invoke commands with or without the forward slash. Entering 'ca' or '/ca' both trigger the same response. The trigger word is case-insensitive.
User Rate Limits & Blocks
Control individual user access to image generation through daily generation caps and permanent blocks.
Restriction Types
| Type | Behavior | Use Case |
|---|---|---|
| Daily Limit | User can generate up to N images per day, then receives a rate-limited message | Prevent excessive usage by specific users |
| Block | User's trigger commands are silently ignored with no response | Permanently remove a user's ability to generate |
Adding Restrictions
- Open the User Limits tab in the dashboard
- Select the bot you want to manage
- Click Add Restriction and enter the Telegram username
- Choose between Daily Limit (specify the number) or Block
- Save — the restriction syncs to the live bot immediately
Editing & Removing
Existing restrictions can be edited inline. Switch a user from a daily limit to a block, adjust the daily cap, or remove the restriction entirely. Every change maintains an edit history showing what was changed and when, visible in the restriction details.
The bot checks group authorization first, then user restrictions. A blocked user in an authorized group will be silently ignored. A rate-limited user who has exceeded their daily cap receives an informational message.
Per-Bot Restrictions
Restrictions are scoped per bot. If you operate multiple bots, blocking a user on one bot does not affect their access to your other bots. Each bot maintains its own independent restriction list.
Multi-Character & Showcase System
Advanced deployment and automated content generation
Multi-Character Deployment
Dedicated compute plans and above support up to eight total characters per bot, each with a unique trigger word and independently trained LoRA model.
Multi-character deployment requires Dedicated Compute ($449/mo), X Agent ($899/mo), or Combo ($1,199/mo). Shared compute bots operate with a single character only.
How Multi-Character Works
Each character is a separately trained LoRA model with its own trigger word. When a user sends a trigger command, the bot matches it against the primary trigger first, then checks additional characters. The matching character's LoRA weights load onto the GPU, and the image generates using that character's model and suffix configuration.
| Specification | Value |
|---|---|
| Base Character | 1 included with every dedicated plan |
| Additional Characters | Up to 7 more (free, on request) |
| Maximum Total | 8 characters per bot |
| Trigger Words | Unique per character, case-insensitive |
| Generation Time | 5 seconds (single character) / 11 seconds (multi-character) |
| LoRA Hotswapping | Automatic per trigger — no manual switching |
Generation Time Impact
Loading multiple LoRA models into VRAM requires additional compute overhead. With a single character loaded, generation completes in 5 seconds. With two or more characters, the dynamic hotswapping process increases generation time to 11 seconds. Removing all additional characters restores 5-second generation immediately.
Requesting Additional Characters
- Submit a request through the Payments tab in your dashboard
- We reach out via Telegram to collect training images for the new character
- The new character is trained and deployed — typically live by the next day
- The character appears in your Multi-Character tab for management
Managing Characters from the Dashboard
- Enable or disable individual characters without removing them
- Edit trigger words at any time — changes sync to the live bot immediately
- Disabled characters do not respond to their trigger word
- The primary character (base model) cannot be disabled
The Multi-Character tab only appears in the dashboard when at least one of your bots has additional characters configured. If no characters exist beyond the primary, the tab is hidden.
Demo Showcase Groups
Each bot can have one demo group where the bot sends prompts to itself continuously, generating thousands of images per day that are automatically collected and hosted on a free subdomain.
How Demo Groups Work
A demo group is a dedicated Telegram supergroup where your bot acts as its own prompter. The bot sends image generation prompts to itself on a timed interval, producing a continuous stream of images without any human input. With a 2-minute interval between generations, a single demo group produces over 700 images per day. At maximum throughput, this scales to over 4,000 images per day.
Each bot supports exactly one demo group. Creating a new demo group replaces the existing one. This is a dedicated showcase channel, not a general-purpose feature.
Prompt Packs
The prompter draws from curated prompt packs — themed collections of prompts designed to produce specific categories of images. You select a pack from the dashboard, and the bot cycles through its prompts in order, generating themed content automatically.
| Pack Type | Description |
|---|---|
| Nature | Outdoor scenes, landscapes, wildlife, and natural environments |
| Luxurious Lifestyle | High-end fashion, exotic cars, yachts, penthouses, jewelry |
| Crypto Trading | Charts, trading desks, bull runs, diamond hands, market themes |
| Meme Culture | Internet culture references, viral formats, community humor |
| Shuffle (All Packs) | Randomized selection across every available prompt pack |
Additional prompt packs are added regularly. The Shuffle option randomizes across all available packs, providing varied content without manual pack switching.
Free Hosted Website
When you create a demo group, a free website is automatically provisioned at your character's subdomain. Every image generated in the demo group is uploaded to this site and hosted indefinitely at no cost.
Website Generation
Beyond hosting images, website-based image generation can be implemented on your subdomain upon request at no additional cost. This allows visitors to generate images directly from the website using a text prompt, with results delivered in real time via WebSocket.
Managing the Prompter
- Start and stop the prompter from the Demo tab in the dashboard
- Switch prompt packs at any time — the new pack loads immediately
- Monitor prompter status including total prompts sent, current pack, and last generation timestamp
- View the Telegram group directly via the invite link in the dashboard
- Edit the group title or delete the demo group entirely from the same interface
Real-Time Configuration & Deployment
Every configuration change made through the dashboard propagates to the live bot within seconds through a unified runtime file and automatic service restart.
How Configuration Reaches Your Bot
The dashboard writes all bot configuration — authorized groups, user restrictions, custom commands, character settings, and rate limits — into a single unified runtime file on the bot's server. After writing the file, the bot's system service restarts automatically. On the next incoming Telegram message, the bot reads the updated configuration from disk and applies it immediately.
- You make a change in the dashboard (authorize a group, add a command, block a user, etc.)
- The dashboard saves the change to the central configuration database
- A unified runtime file is written to the bot's server containing all current settings
- The bot's system service restarts automatically
- On the next incoming message, the bot loads the updated runtime file and operates with the new configuration
The service restart completes in under a second. Messages received during the restart are queued by Telegram and processed immediately after the bot comes back online. No messages are lost.
Group Authorization Process in Detail
The full process from adding a bot to a group through to live image generation follows a specific sequence:
- Add the bot to a Telegram group using the bot's username
- Send /getid in the group — this allows the server to discover the chat and register its metadata
- Open the Groups tab in the dashboard — the group appears in the discovered list
- Select the group and click Authorize — the runtime file updates and the service restarts
- The bot now listens for the trigger word in that group
- When a user sends the trigger followed by a prompt, the bot queues the request and the GPU begins generating the image as soon as the model is ready
Custom text commands operate independently from group authorization. Commands respond in any group the bot is added to, even if that group is not authorized for image generation. Only the image generation trigger requires explicit authorization.
What the Runtime File Contains
| Section | Controls |
|---|---|
| Authorized Chat IDs | Which groups can trigger image generation |
| Characters | Multi-character LoRA configurations, trigger words, enabled/disabled state |
| Text Commands | Custom trigger-response pairs |
| Blocked Users | Usernames permanently blocked from generation |
| Rate Limited Users | Usernames with daily generation caps |
All five sections are written atomically in a single file update. If the write fails for any reason, the dashboard rolls back the change to keep the configuration database and the bot's runtime file in sync.
Transaction Safety
Every configuration write follows a rollback pattern. The dashboard first updates the central database, then writes the runtime file to the bot server. If the runtime file write fails, the dashboard reverts the database change. This guarantees that the dashboard and the live bot always reflect the same configuration state.
Platform Features
Core functionality and capabilities
Telegram Bots
Telegram bots receive a custom username, profile picture, and bio that you control instantly through the dashboard.
Telegram bots operate as fully autonomous agents with dedicated profiles including custom username, profile picture, display name, and bio.
How Users Interact
/gary wearing a suit and holding a briefcase
OR
gary wearing a suit and holding a briefcaseCommunity members interact with your bot using your custom trigger command followed by their prompt. The forward slash is optional—we support both formats for convenience.
Performance by Tier
| Tier | Generation Time | Multi-Character | Watermark |
|---|---|---|---|
| Shared | 11 seconds | No | Yes (removable) |
| Dedicated | 5 seconds | Up to 8 | No |
Changes to your trigger command or authorized Telegram Group IDs can be requested through support for same-day processing at no charge.
X (Twitter) Agents
X agents include a verified Twitter account, X API plan, and dedicated compute for $899/month.
The $899 monthly subscription includes a verified Twitter account, full X API plan, 30-second polling, and dedicated compute resources.
How X Agents Work
- Bot polls X every 30 seconds for mentions
- Detects tweets containing both @username and trigger word
- Processes all pending generation requests
- Tweets resulting images as replies from your character's account
X Agent Specifications
| Feature | Included |
|---|---|
| Verified Account | Yes |
| X API Plan | Full access |
| Polling Interval | 30 seconds |
| Compute Type | Dedicated |
| Image Limits | Unlimited |
| Characters per Account | 1 |
Deployment takes up to 2 business days because X API approval typically requires 12-24 hours after application submission.
Dashboard Management
The Memescale dashboard provides real-time bot customization and centralized management.
Real-Time Profile Customization
- Edit bot username instantly
- Update display name
- Modify bio text
- Change profile picture
Dashboard Features
| Feature | Description |
|---|---|
| Order Tracking | Visibility into training status from payment through deployment |
| Billing Management | Upgrade, downgrade, or cancel subscriptions |
| Profile Editing | Real-time changes to bot appearance |
| Support Messaging | Integrated ticketing system |
| User Controls | Whitelist, blacklist, and rate limiting |
Compute Infrastructure
Technical architecture and performance
Dedicated Compute
Dedicated compute provides 100% VRAM allocation with zero sharing, no throttling, and 5-second generation times.
| Specification | Value |
|---|---|
| VRAM Allocation | 100% |
| Sharing | None |
| Generation Time | 5 seconds (single character) / 11 seconds (multi-character) |
| Monthly Capacity | Unlimited |
| Multi-Character | Up to 8 total |
Why 5 Seconds is Guaranteed
The 5-second generation time is mathematically verifiable because of the isolated pipeline architecture with zero external dependencies—every component operates on infrastructure we own and control directly.
Multi-Character Support
- Base model included
- Up to 7 additional characters for free — request via the Payments tab
- Different trigger words for each character
- Dynamic LoRA hotswapping between characters
- Loading multiple custom models requires additional compute, increasing generation time from 5 to 11 seconds — all additional characters are free and can be removed at any time to restore 5-second generations
Available Plans
| Plan | Price | Includes |
|---|---|---|
| Telegram Dedicated | $449/mo | Dedicated GPU, multi-character, priority support |
| X Agent | $899/mo | Verified account, X API, dedicated compute |
| Combo (Telegram + X) | $1,199/mo | Everything + VIP support + optional web embed |
Higher resolutions beyond 1024x1024 are available upon request at no additional cost on all dedicated plans.
Server Architecture
Memescale bots run on isolated GPU processes with zero external dependencies and DDoS-resistant protection.
Security Model
- Complete port closure except SSH tunnels
- Cryptographically protected connections
- Air-gapped server design
- Telegram webhooks on separate infrastructure
- Zero external API dependencies
Even if AWS experiences outages, your bot continues operating because the entire generation pipeline exists on infrastructure we own and manage directly.
Local Model Execution
Pre-compiled LoRA weights stored directly on GPU-equipped servers means generation speed and availability never depend on external services, rate limits, or network conditions beyond our control.
User-Level Controls
| Control | Description |
|---|---|
| Block User | Prevent specific users from generating |
| Rate Limit | Restrict generations per hour/day |
| Whitelist | Allow only approved users |
| Per-Platform | Different rules for Telegram vs X |
For combo package customers who opt for website embedding, we include complimentary proof-of-work bot blocker protection to prevent automated abuse.
Training & Model Quality
Dataset preparation and model optimization
Dataset Requirements
Submit 8-20 clear, consistent images showing your character from multiple angles with good lighting and varied poses.
Training data quality is the single most important factor determining model output accuracy and character consistency.
Image Requirements
- 8-20 images total
- Multiple angles: front, side, three-quarter
- Consistent lighting without heavy shadows
- Clear visibility without significant occlusion
- No blur or extreme brightness
- Consistent art style across all images
File Specifications
While we recommend images be approximately 1MB or larger, this is a guideline rather than a hard requirement—the validation process checks for adequate resolution and quality.
Training Process
After payment and upload, we handle fine-tuning using your dataset and deploy your bot the same day.
What We Handle
- Fine-tune LoRA weights on your specific dataset
- Deploy inference system to GPU infrastructure
- Configure webhook integrations
- Set up bot profile with your specifications
- Add you to private evaluation group
- Send Telegram notification when complete
| Platform | Delivery Time | Reason |
|---|---|---|
| Telegram | Within 12 hours | Direct API integration |
| X (Twitter) | Within 12-24 hours | X API approval required |
Evaluation Period
When training completes, you're added as admin to a private group containing your bot. Test the generation quality before community deployment.
If the output quality doesn't meet expectations, request a full refund with no questions asked. Your subscription only begins after you approve the bot.
Model Revisions
All models include two free revision rounds where you can submit additional training data to improve accuracy.
When to Request Revisions
- Character consistency needs improvement
- Facial features aren't accurate
- Model struggles with certain angles
- Specific poses or expressions need refinement
Professional Dataset Creation
We generate a large batch of images using your existing model, curate the highest-quality outputs through manual selection, and retrain using this curated dataset.
This service is particularly valuable for commercial operations where generation quality directly impacts revenue, or large-scale meme projects where character consistency is critical for brand recognition.
Pricing & Plans
Subscription options and add-on services
Plan Comparison
All plans generate 1024x1024 images by default, with dedicated GPU plans offering fully customizable resolution and aspect ratios.
| Plan | Price | Compute | Generation Time | Best For |
|---|---|---|---|---|
| Telegram Shared | $199/mo | Shared (1/4 GPU) | 11 seconds | Testing, smaller communities |
| Telegram Dedicated | $449/mo | Dedicated | 5 seconds | Established projects |
| X Agent | $899/mo | Dedicated | 7-37 seconds | Twitter presence |
| Combo (TG + X) | $1,199/mo | Dedicated | Both platforms | Maximum reach |
Shared Compute ($199/mo)
- Unlimited image generation
- Round-robin queue with up to 3 other projects
- 11 second generation times
- 1024x1024 resolution
- Unlimited images/month
- Default watermark (removable)
Dedicated Compute ($449/mo)
- 100% VRAM allocation
- 5-second guaranteed generation (11s with multi-character)
- Unlimited images/month
- Up to 8 characters (free)
- Custom resolutions available
- No watermark
- Priority support
X Agent ($899/mo)
Users must both mention @yourbot AND include the trigger word. Order doesn't matter, but both are required.
Combo Package ($1,199/mo)
- Everything from Telegram Dedicated
- Everything from X Agent
- VIP support with founder's direct line
- Optional website embedding included
- Proof-of-work bot blocker protection
Add-on Services
Multi-character functionality, watermark removal, and professional dataset creation are available as optional add-ons.
| Add-on | Price | Type | Available On |
|---|---|---|---|
| Additional Character | Free | On request | Dedicated plans only |
| Watermark Removal | $300 | One-time | Shared compute |
| Dataset Creation | $100 | One-time | All plans |
Multi-Character Deployment
Dedicated plan users can add up to seven additional characters beyond the base model for free, supporting a maximum of eight total characters accessible through different trigger words. Loading multiple custom models requires additional compute, so generation time increases from 5 to 11 seconds — but all extra characters are entirely free and can be removed at any time to restore 5-second generations. To request extra characters, submit a request through the Payments tab in your dashboard. We'll reach out to your Telegram to collect the training images, and your new character will be live by the next day.
Multi-character support was implemented due to popular demand but is not mandatory—the majority of successful projects operate with a single character.
Watermark Policy
- Applied only to shared server Telegram bots
- X agent images never include watermarks
- Dedicated compute = watermark-free by default
- One-time $300 fee to remove on shared compute
If watermark removal matters, consider dedicated compute at $449/mo—you get watermark-free images plus 5-second generations and multi-character support.
Payment Information
All Memescale pricing is denominated in USD and paid via cryptocurrency. When you see "$199" or "$449," these represent USD-equivalent amounts calculated at payment time using current exchange rates.
Account Management
Plan modifications and billing policies
Plan Changes
All plan changes take effect at your next billing cycle, ensuring you receive the full service period you've paid for.
All upgrades, downgrades, and cancellations take effect at the beginning of your next billing cycle rather than immediately.
How Plan Changes Work
| Change Type | When It Applies | Billing Impact |
|---|---|---|
| Upgrade | Next billing cycle | New rate from next payment |
| Downgrade | Next billing cycle | Keep current tier until cycle ends |
| Cancellation | Next billing cycle | Full access through paid period |
Example Scenario
If you upgrade from shared to dedicated on day 15 of your 31-day billing cycle, you continue using shared compute until day 31, at which point the upgrade activates and billing adjusts to $449/month.
Profile Modifications
| Platform | Processing Time | Method |
|---|---|---|
| Telegram | Instant | Dashboard API integration |
| X (Twitter) | 2-4 hours (max 12) | Manual (prevents shadow bans) |
Twitter's strict API guidelines can trigger shadow bans if profile changes are made programmatically. Manual processing protects your verified account status.
Billing & Grace Periods
Billing occurs monthly with manual cryptocurrency payment required. A 24-hour grace period applies before bot shutdown.
Payment Process
- Dashboard displays next payment due date
- Send cryptocurrency to provided address
- Payment processed via NOWPayments
- Order ID generated for tracking
Memescale never holds your private keys, never accesses your wallet, and cannot automatically debit payments. You maintain complete custody of your funds.
Grace Period Details
If payment isn't received by your due date, you get 24 hours of full bot functionality while resolving payment issues. After 24 hours without payment, your bot turns off automatically.
Refund Policy
Refunds are provided exclusively during the evaluation period (before you approve your bot). Once subscription begins, all payments are final with no refunds for partial months.
Model Preservation
When you stop paying, your trained model is preserved indefinitely. Reactivation occurs the same day at current pricing.
Even if you stop paying for weeks, months, or years, your trained LoRA weights and configuration remain stored in our system indefinitely.
What Gets Preserved
- Trained LoRA weights
- Fine-tuned model data
- Original training images
- Bot username and display name
- Bio and description
- Profile configuration
Reactivation Process
| Aspect | Details |
|---|---|
| Timeline | Same-day deployment |
| Pricing | Current rates (not grandfathered) |
| Training Required | None (existing model used) |
| Upload Required | None (images preserved) |
Use Cases
- Seasonal projects needing temporary service
- Community initiatives with fluctuating funding
- Projects testing Memescale before full migration
- Pausing during low-activity periods
To permanently delete your model and training data, contact support with an explicit deletion request. Files are removed within 48 hours.
Content & Moderation
Filtering policies and content responsibility
Filtering Policy
Hand-crafted system prompts achieve 99%+ effectiveness at preventing nudity and NSFW content.
The system prompts don't just filter content—they actively enhance visual quality and character consistency across all outputs.
What's Filtered
| Content Type | Filtered? | Notes |
|---|---|---|
| Nudity/NSFW | Yes | 99%+ effectiveness |
| Gore/Violence | No | At your discretion |
| Weapons | No | At your discretion |
| Mature Themes | Varies | Context-dependent |
If you require additional content restrictions, implement moderation at the community level through Telegram group settings, X account controls, or dashboard user management features.
Content Responsibility
You are solely responsible for moderating the social channels where your bot operates.
There are no terms of service governing bot usage, prompt submissions, or content themes. The only rule is that you have fun.
Why No Restrictions?
- Open-source models with local infrastructure
- No reselling or aggregating of generated images
- Eliminates compliance complexities
- Respects community autonomy
Dashboard Moderation Controls
| Control | Function | Scope |
|---|---|---|
| Whitelist | Allow only specified users | Per-platform |
| Blacklist | Block specific users | Per-platform |
| Rate Limiting | Cap generations per hour/day | Per-user |
| Silent Block | Ignore messages entirely | Per-user |
| Timeout Response | Inform user they're blocked | Per-user |
These controls operate independently for Telegram and X, allowing different moderation policies across platforms.
Technical Specifications
Resolution, parameters, and platform integration
Image Resolution & Parameters
All plans generate 1024x1024 images by default with custom resolutions available on dedicated plans.
| Specification | Default | Customizable? |
|---|---|---|
| Resolution | 1024x1024 | Dedicated plans only |
| Aspect Ratio | 1:1 (square) | Dedicated plans only |
| Format | PNG | No |
| File Size | 400KB - 2MB | Varies by complexity |
| Guidance Scale | Optimized | By request (discouraged) |
| Inference Steps | Optimized | By request (discouraged) |
Custom Resolution Requests
- Available on all dedicated plans at no additional cost
- Widescreen, portrait, and higher resolutions supported
- Submit requests during onboarding or anytime through support
- Higher resolutions slightly increase generation time
Telegram's built-in compression automatically downscales images for efficient transmission, making higher resolutions provide minimal visible improvement on the platform.
Parameter Modifications
Generation parameters are locked by default. While adjustments are possible by DMing @MemescaleAI, this is actively discouraged as it can degrade output quality and character consistency.
Performance Limits
No concurrent user limits exist—unlimited community members can generate simultaneously.
Throughput Example
On dedicated compute, 100 prompts submitted in 10 seconds would complete in approximately 8 minutes (100 × 5 seconds). On shared compute, they process through the round-robin queue based on other project activity.
Spam Protection
| Protection | Method | User Impact |
|---|---|---|
| Proof-of-Work | WebAssembly computation | Imperceptible |
| Difficulty | 8 leading zeroes | Blocks bots only |
| Rate Limiting | Per-user configurable | Optional |
User Access Controls
- Whitelists for allowed users
- Blacklists for blocked users
- Rate limits per hour or per day
- Independent controls for Telegram and X
Platform Integration
X agents include full X API integration. Telegram bots make zero external API calls.
| Platform | Integration Method | External APIs |
|---|---|---|
| X (Twitter) | Managed X API | 30-second polling |
| Telegram | Webhook-based | Zero external calls |
X Agent Integration
- Managed X API access included
- 30-second polling for mentions
- No customer API key management required
- Automatic authentication handling
Telegram Integration
Telegram bots operate through webhooks but make zero external API calls during generation. All prompt parsing, inference, and delivery happens on isolated local infrastructure.
Memescale does not offer customer API access for triggering generations, managing settings, or retrieving images through code. Contact @MemescaleAI on Twitter to discuss custom solutions.
Support
Support tiers and contact methods
Support Tiers
Support response times and access levels vary by subscription tier.
| Tier | Response Time | Channel | Included With |
|---|---|---|---|
| Standard | Within 1 business day | Dashboard messaging | All plans |
| Priority | Within hours | Dashboard messaging | Dedicated plans ($449+) |
| VIP | Immediate 24/7 | Founder's direct cell | Combo package ($1,199) |
Standard Support
- Included with all subscription tiers
- Handles billing, training status, profile modifications
- Technical assistance and troubleshooting
- Dashboard messaging system with ticket tracking
Priority Support
Dedicated plans ($449/mo and above) include priority support with response times measured in hours rather than days—ideal for projects requiring rapid resolution during critical deployment periods.
VIP Support
Combo package subscribers receive the founder's personal cell phone for immediate 24/7 assistance during major announcements, partnership launches, or high-traffic events.
Contact Methods
Support inquiries are handled through the dashboard messaging system, with VIP access for combo subscribers.
| Method | Access | Best For |
|---|---|---|
| Dashboard Messaging | All users | Standard inquiries |
| Founder's Cell Phone | Combo subscribers only | Urgent issues |
Dashboard Messaging
- Access from any dashboard page
- Tracked tickets persist until resolution
- Covers all inquiry types
- Standard: 1 business day response
- Priority: Within hours
VIP Direct Access
Combo package subscribers receive the founder's personal cell phone during onboarding, providing direct 24/7 access for urgent situations.
Technical emergencies during high-traffic events, urgent profile modifications before major announcements, or immediate assistance with complex customization requests.
Security & Privacy
Account security and data retention policies
Account Security
Session security is cryptographically enforced. Memescale never holds private keys or accesses wallets.
Account Protection
| Security Feature | Description |
|---|---|
| Wallet Security | Never hold keys, never access wallets |
| X API Auth | Handled internally, no user credentials |
Infrastructure Security
- Cryptographically protected SSH tunnels
- Complete port closure on all GPU servers
- Air-gapped webhook architecture
- Zero attack surface from open endpoints
This architecture maintains uptime even during widespread outages affecting major cloud providers—because we don't depend on them.
For complete security architecture documentation including session binding, HMAC reconstruction, and threat detection, see Security Documentation.
Data Retention
Training images and model weights are preserved indefinitely. Explicit deletion requests processed within 48 hours.
What's Preserved
| Data Type | Retained? | Duration |
|---|---|---|
| LoRA Weights | Yes | Indefinitely |
| Training Images | Yes | Indefinitely |
| Bot Configuration | Yes | Indefinitely |
| Generated Images | No | Not stored |
Even 50 cancelled bots over five years consume only ~10GB of total storage, making indefinite preservation operationally sustainable.
Preservation Benefits
- Pause service for weeks, months, or years
- Reactivate without retraining
- Same-day deployment on return
- No additional upload required
Data Deletion
Advanced Features
Multi-character deployment and customization
Multi-Character Deployment
Dedicated plans support up to eight total characters accessible through different trigger words.
Multi-character support was implemented due to popular demand but remains optional. Most successful projects operate with a single character.
| Specification | Value |
|---|---|
| Base Characters | 1 included |
| Additional Characters | Up to 7 (free) |
| Maximum Total | 8 characters |
| Generation Time | 11 seconds (5s with single character) |
| Trigger Words | Unique per character |
| Profile Sharing | Same username/bio for all |
How It Works
- Each character requires its own unique trigger word
- Community member uses appropriate trigger to select character
- Dynamic LoRA hotswapping loads correct weights
- Same bot profile serves all characters
Technical Isolation
Backend infrastructure ensures complete isolation through Group IDs, Chat IDs, and unique internal request tags. Even if two bots on different shared servers use identical trigger commands, there is zero risk of cross-contamination.
Loading multiple custom models requires additional compute, which increases generation time from 5 to 11 seconds. All additional characters are completely free to add and can be removed at any time — removing all extra characters restores 5-second generation immediately.
Custom Configuration
Dedicated plans support custom resolutions and aspect ratios at no additional cost.
Resolution Options
| Option | Availability | Cost | Impact |
|---|---|---|---|
| 1024x1024 (default) | All plans | Included | 5s generation |
| Custom resolution | Dedicated only | Free | Slight increase |
| Widescreen formats | Dedicated only | Free | Slight increase |
| Print quality | Dedicated only | Free | 15-18s generation |
Higher resolutions provide minimal visible improvement on Telegram due to the platform's built-in compression. Request higher resolutions for print, off-platform distribution, or X where compression is less aggressive.
Parameter Adjustments
Generation parameters (guidance scale, inference steps, sampling methods) are locked by default. Adjustments can introduce inconsistency and degrade quality.
Website Embedding
Combo package subscribers ($1,199/mo) receive optional complimentary website embedding with proof-of-work bot blocker protection, allowing your model to be accessible through a custom web interface.
Profile Customization
Telegram profiles update instantly. X profiles are handled manually to prevent shadow bans.
| Platform | Processing | Reason |
|---|---|---|
| Telegram | Instant | Direct API integration |
| X (Twitter) | 2-4 hours (max 12) | Manual to prevent shadow bans |
Telegram Profile Changes
- Username, display name, bio update instantly
- Profile picture changes propagate within seconds
- Changes reflect across all groups
Trigger Word Mechanics
| Platform | Trigger Format | Example |
|---|---|---|
| Telegram | Trigger word only | /gary wearing a suit |
| X (Twitter) | @username + trigger word | @GaryBot Gary in a suit |
The forward slash is optional in both setup and usage. Community members can use /gary or gary interchangeably.
Both @username mention AND trigger word are required for X generation. Order doesn't matter, but both must be present.
Performance Benchmarks
Generation times and uptime statistics
Generation Times
Shared compute averages 11 seconds during peak hours. Dedicated guarantees 5 seconds regardless of load.
| Platform/Tier | Peak Hours | Off-Peak | Guaranteed |
|---|---|---|---|
| Telegram Shared | 11 seconds | Near-instant | No |
| Telegram Dedicated | 5 seconds | 5 seconds | Yes |
| X Agent | 7-37 seconds* | 7-37 seconds* | Consistent |
*X agent timing includes 30-second polling interval plus generation time.
Shared Compute Variability
- Peak hours: 11 seconds with active queues
- Off-peak: Often near-instant (fewer competing projects)
- Round-robin ensures fair distribution regardless of timing
X Agent Timing Breakdown
A prompt sent immediately after polling takes ~36 seconds (30s wait + 5s generation). A prompt sent 25 seconds into the cycle takes ~11 seconds (5s wait + 5s generation).
Monthly Capacity
| Tier | Theoretical Max* | Practical Reality |
|---|---|---|
| Shared (per project) | Unlimited | Unlimited |
| Dedicated | Unlimited | Unlimited |
*If prompts were sent continuously every second of every day.
Uptime Statistics
Downtime is extremely limited and rare, occurring only during scheduled maintenance.
Why High Uptime
- DDoS-resistant infrastructure
- Cryptographically protected SSH tunnels
- Air-gapped server design
- Zero external dependencies
- Direct hardware control in remote locations
Generation pipelines continue operating even during widespread AWS outages. All GPU processes run on isolated hardware under direct Memescale control.
Scheduled Maintenance
| Aspect | Details |
|---|---|
| Frequency | Rare, as-needed |
| Duration | Few hours to 1 day maximum |
| Notice | Advance dashboard notifications |
| Timing | Scheduled during low-traffic periods |
Maintenance Triggers
- Security patches
- Infrastructure upgrades
- Image generation pipeline improvements
The isolated architecture and local model execution eliminate the unpredictable downtime that plagues API-reliant alternatives, making scheduled maintenance the sole predictable interruption.
Best Practices
Training optimization and deployment strategies
Training Data Selection
Submit high-resolution images with transparent backgrounds showing your character from multiple angles.
Memescale removes backgrounds and upscales low-quality images at no additional cost as part of achieving stellar model quality.
Ideal Training Images
- 8-20 images above 1MB
- Transparent backgrounds
- Multiple angles: front, side, three-quarter
- Consistent lighting without heavy shadows
- Clear visibility without occlusion
- Varied expressions if applicable
Consistency Requirements
| Feature | Recommendation | Why It Matters |
|---|---|---|
| Outfit | Same signature outfit | Character recognition |
| Anatomy | Consistent features | Reliable generation |
| Style | Unified art style | Output consistency |
| Accessories | Include if defining | Complete character |
Varying tail counts, missing ears, different facial structures between images increase the likelihood of inconsistent generated outputs.
Community Management
Use dashboard controls to moderate users and leverage bot content for broader reach.
Moderation Controls
| Control | Behavior Options | Use Case |
|---|---|---|
| Blacklist (Silent) | Ignore messages entirely | Quiet removal |
| Blacklist (Notify) | Inform user of timeout | Clear communication |
| Whitelist | Only approved users generate | Exclusive access |
| Rate Limit | Cap per hour/day | Abuse prevention |
Content Amplification Strategies
- Repurpose high-quality generations as official social posts
- Host meme creation competitions with prizes
- Showcase community-generated content on project X account
- Feature best generations in announcements
Platform Visibility
X provides global public visibility where anyone can discover your character through mentions. Telegram groups limit visibility to current members only.
Unless you're on shared compute Telegram (which includes a watermark), nobody will know you're using Memescale—present the bot as entirely your own creation.
Scaling Deployment
Start with shared compute for testing, upgrade to dedicated when engagement justifies it.
Recommended Scaling Path
- Start with shared compute ($199/mo) to test viability
- Validate community interest and assess usage patterns
- Upgrade to dedicated ($449/mo) when peak-hour performance matters
- Add X agent or combo package for multi-platform presence
- Enable multi-character if managing multiple mascots
| Community Stage | Recommended Tier | Key Benefit |
|---|---|---|
| Testing/Launch | Shared ($199) | Low risk, unlimited generations |
| Growing | Shared ($199) | Sufficient for most communities |
| Established | Dedicated ($449) | Guaranteed 5s, multi-character |
| Multi-Platform | Combo ($1,199) | Maximum reach, VIP support |
Telegram Scaling
Add the same bot to unlimited groups with no additional costs or separate deployments. Scale across multiple communities without managing multiple instances.
Plan Change Timing
All plan changes take effect at your next billing cycle boundary. You always receive the full service period you've paid for before changes apply.
Glossary
Key terms and definitions
Terms & Definitions
Comprehensive glossary of technical terms, platform-specific concepts, and Memescale terminology.
Platform Terms
| Term | Definition |
|---|---|
| Memescale | AI-powered platform providing custom-trained image generation bots for Telegram and X |
| Dashboard | Web interface for managing bots, profiles, billing, and support |
| Trigger Command | The word or phrase that activates image generation (e.g., /gary) |
| Evaluation Period | Testing phase after training where you can request a refund if unsatisfied |
Compute Terms
| Term | Definition |
|---|---|
| Shared Compute | GPU resources distributed across up to 4 projects via round-robin queue ($199/mo) |
| Dedicated Compute | Exclusive 100% GPU allocation with guaranteed 5-second generation ($449/mo) |
| Round-Robin Queue | Fair rotation system that processes one image per project regardless of queue size |
| VRAM | Video RAM - GPU memory used for model inference and image generation |
AI/ML Terms
| Term | Definition |
|---|---|
| LoRA | Low-Rank Adaptation - efficient fine-tuning method for customizing AI models |
| LoRA Weights | Trained parameters specific to your character, stored as .safetensors files |
| Hotswapping | Dynamic switching between different LoRA weights for multi-character support |
| Fine-Tuning | Training process that adapts the base model to your specific character |
| Inference | The process of generating images from prompts using trained models |
| FLUX.1-dev | Base diffusion model used for image generation |
Infrastructure Terms
| Term | Definition |
|---|---|
| SSH Tunnel | Encrypted connection between servers for secure communication |
| Air-Gapped | Architecture where systems are isolated from external networks |
| DDoS Protection | Defense against distributed denial-of-service attacks |
| Webhook | HTTP callback that delivers Telegram messages to our servers |
| Polling | Periodic checking for new X mentions (30-second intervals) |
Account Terms
| Term | Definition |
|---|---|
| Billing Cycle | 31-day subscription period starting from initial payment |
| Grace Period | 24-hour window after missed payment before bot shutdown |
| Model Preservation | Indefinite storage of trained models when subscription lapses |
| Session Security | Cryptographically bound session authentication protecting all account operations |
Plan-Specific Terms
| Term | Definition |
|---|---|
| X Agent | Autonomous Twitter bot with verified account and X API integration ($899/mo) |
| Combo Package | Combined Telegram + X agent with VIP support ($1,199/mo) |
| Multi-Character | Feature allowing up to 8 characters with different triggers (dedicated only) |
| Priority Support | Faster response times included with dedicated plans |
| VIP Support | Direct founder access included with combo package |
Content Terms
| Term | Definition |
|---|---|
| Watermark | Memescale branding on images (shared compute only, removable for $199) |
| System Prompts | Hand-crafted instructions preventing NSFW content (99%+ effective) |
| Whitelist | User access control allowing only specified users |
| Blacklist | User access control blocking specified users |
| Rate Limiting | Restricting specific users to defined generations per hour/day |
FAQ & Troubleshooting
Common questions and solutions
General Questions
Frequently asked questions about Memescale services and capabilities.
Getting Started
| Question | Answer |
|---|---|
| How long until my bot is ready? | Telegram: 12 hours (1 business day). X: 12-24 hours (2 business days). |
| Do I need technical knowledge? | No. We handle all training, deployment, and maintenance. |
| Can I test before committing? | Yes. Full refund available during evaluation period before approval. |
| What payment methods are accepted? | Cryptocurrency only (stablecoins and major crypto assets). |
Pricing & Billing
| Question | Answer |
|---|---|
| Are there hidden fees? | No. One flat monthly fee covers everything. |
| What happens if I miss a payment? | 24-hour grace period, then bot turns off. Model preserved indefinitely. |
| Can I upgrade mid-cycle? | Changes take effect at next billing cycle, not immediately. |
| Is there a contract? | No. Cancel anytime, effective at next billing cycle. |
Features & Limits
| Question | Answer |
|---|---|
| Is generation really unlimited? | Yes. No credits, quotas, or per-image charges. |
| How many groups can use my bot? | Unlimited Telegram groups with same bot. |
| Can multiple users generate simultaneously? | Yes. No concurrent user limits. |
| What resolution are images? | 1024x1024 default. Custom resolutions on dedicated plans. |
Troubleshooting
Solutions for common issues and problems.
Bot Not Responding
- Check subscription status in dashboard
- Verify you're using the correct trigger command
- Ensure the bot is added to your group
- For X: Confirm both @username and trigger word are in tweet
- Contact support if issue persists
Forward slash is optional. Both /gary and gary work.
Generation Quality Issues
| Issue | Solution |
|---|---|
| Inconsistent character | Request revision with better training data |
| Wrong features | Submit more images showing correct features |
| Poor angles | Add images from problematic angles to training set |
| Style mismatch | Ensure training images use consistent art style |
You have 2 free revision rounds included. Use them to refine model quality.
Payment Issues
- Save your order ID for all transactions
- Check blockchain explorer for transaction status
- Contact support immediately with order ID
- 24-hour grace period if payment delayed
X Agent Specific
| Issue | Cause | Solution |
|---|---|---|
| No response to tweet | Missing @mention or trigger | Include both @username AND trigger word |
| Slow response | 30-second polling cycle | Normal behavior, wait up to 45 seconds |
| Profile not updating | Manual processing required | Allow up to 12 hours for changes |
Common Mistakes
Avoid these frequent pitfalls for the best experience.
Training Data Mistakes
- Submitting too few images (minimum 8 recommended)
- Inconsistent character features across images
- Low resolution or blurry images
- Heavily occluded or cropped images
- Mixed art styles in training set
8 excellent images beat 20 mediocre ones. Focus on consistency and clarity.
Platform Mistakes
| Mistake | Impact | Prevention |
|---|---|---|
| Expecting instant X responses | Frustration with 7-37s timing | Understand 30s polling cycle |
| Forgetting @mention on X | Generation doesn't trigger | Always include @username AND trigger |
| Requesting parameter changes | Quality degradation | Trust optimized defaults |
| Skipping evaluation period | Regret after approval | Test thoroughly before approving |
Account Mistakes
- Not saving order IDs for payment tracking
- Expecting mid-cycle plan changes
- Assuming refunds after approval (evaluation period only)
- Waiting until after grace period to resolve payment issues