34 KiB
Production Pricing Strategy - Basic Tier Launch (OSS-Focused)
Executive Summary
This document provides a comprehensive pricing strategy for ALwrity's production launch with Basic Tier only. All features and tools will be accessible to Basic tier users, requiring careful cost calculation and limit setting to ensure sustainability while providing value.
Critical Goals:
- OSS-First Strategy: Prioritize Open-Source AI models (WaveSpeed OSS models) for cost efficiency
- Hard Cost Cap: $40-50 per user per month maximum (protects against losses)
- Maximum User Value: Provide generous limits while staying within cost constraints
- Fair Pricing: Balance between sustainability and user value (not excessive profit margins)
Strategy: Use WaveSpeed's OSS models (Qwen, FLUX, Ideogram, WAN 2.5) which offer better pricing than proprietary alternatives, allowing us to provide more value to users while maintaining profitability.
Current State Analysis
Current Basic Tier (Code Implementation)
Price: $29/month ($290/year)
Limits:
- AI Text Generation: 10 unified calls/month (across all LLM providers)
- Tokens: 20,000 per provider (Gemini, OpenAI, Anthropic, Mistral)
- Search APIs: 200 Tavily, 200 Serper, 100 Metaphor, 100 Firecrawl, 500 Exa
- Image Generation: 5 Stability AI images/month
- Image Editing: 30 AI image edits/month
- Video Generation: 20 videos/month
- Audio Generation: 50 TTS generations/month
- Monthly Cost Cap: $50.00
Problem: 10 unified AI text generation calls is too restrictive for production launch where users need to experience all features.
ALwrity Tools & Content Generation Analysis
Content Generation Tools
1. Text Generation Tools (Primary LLM Usage)
| Tool | API Calls per Generation | Typical Usage | Cost per Generation |
|---|---|---|---|
| Blog Writer | 3-5 calls | 1 blog = research (1) + outline (1) + content (1-3) | $0.01 - $0.05 |
| Story Writer | 2-3 calls | 1 story = outline (1) + script (1-2) | $0.01 - $0.03 |
| Podcast Maker | 3-4 calls | 1 podcast = research (1) + script (1) + outline (1-2) | $0.01 - $0.04 |
| Facebook Writer | 1-2 calls | 1 post = generation (1) + optional optimization (1) | $0.005 - $0.01 |
| LinkedIn Writer | 1-2 calls | 1 post = generation (1) + optional optimization (1) | $0.005 - $0.01 |
| SEO Tools | 1-3 calls | Varies by tool complexity | $0.005 - $0.02 |
| Content Planning | 2-4 calls | Strategy generation + analysis | $0.01 - $0.03 |
Average: ~2-3 LLM calls per content generation workflow
2. Image Generation Tools
| Tool | API Calls | Cost per Generation |
|---|---|---|
| Image Generator | 1 Stability call | $0.04 per image |
| Image Editor | 1 Image Edit call | $0.04 per edit operation |
Current Limit: 5 images/month (too low for production)
3. Video Generation Tools
| Tool | API Calls | Cost per Video | Notes |
|---|---|---|---|
| Video Studio | 1 video call | $0.10 - $0.42 | Depends on model/duration |
| YouTube Creator | 1 video call per scene | $0.10 - $0.42 per scene | 5 scenes = $0.50 - $2.10 |
| Story Writer Video | 1 video call per scene | $0.10 - $0.42 per scene | Variable scenes |
| Podcast Maker Video | 1 video call per scene | $0.10 - $0.42 per scene | Optional video generation |
Current Limit: 20 videos/month (reasonable)
4. Audio Generation Tools
| Tool | API Calls | Cost per Generation | Notes |
|---|---|---|---|
| Audio Generator | 1 audio call | $0.05 per 1,000 chars | ~$0.10 - $0.50 per audio |
| Podcast Maker TTS | 1 audio call per scene | $0.05 per 1,000 chars | Multiple scenes |
| Story Writer Narration | 1 audio call per scene | $0.05 per 1,000 chars | Multiple scenes |
Current Limit: 50 audio generations/month (reasonable)
API Cost Breakdown
LLM Provider Costs (Per 1M Tokens)
| Provider | Model | Input Cost | Output Cost | Typical Use |
|---|---|---|---|---|
| Gemini | 2.5 Flash | $0.30 | $2.50 | Default (cost-effective) |
| Gemini | 2.5 Pro | $1.25 | $10.00 | Premium quality |
| OpenAI | GPT-4o Mini | $0.15 | $0.60 | Cost-effective |
| OpenAI | GPT-4o | $2.50 | $10.00 | Premium quality |
| Anthropic | Claude 3.5 Sonnet | $3.00 | $15.00 | Premium quality |
| HuggingFace | GPT-OSS-120B | $1.00 | $3.00 | Alternative option |
Average Cost per LLM Call (assuming 1K input + 2K output tokens):
- Gemini Flash: ~$0.0056 per call
- GPT-4o Mini: ~$0.0015 per call
- Claude 3.5: ~$0.033 per call
Recommendation: Use Gemini Flash as default for cost efficiency.
Search API Costs
| Provider | Cost per Search | Typical Usage |
|---|---|---|
| Tavily | $0.001 | Research operations |
| Serper | $0.001 | Research operations |
| Metaphor | $0.003 | Research operations |
| Exa | $0.005 | Neural search (premium) |
| Firecrawl | $0.002 | Web page extraction |
Average: ~$0.002 per search operation
Media Generation Costs (OSS-Focused via WaveSpeed)
Image Generation (OSS Models via WaveSpeed)
| Model | Cost | Type | Notes |
|---|---|---|---|
| Qwen Image | $0.03 per image | OSS | Fast generation, cost-effective |
| Ideogram V3 Turbo | $0.05 per image | OSS | Photorealistic, text rendering |
| Default (Qwen) | $0.03 per image | OSS | Recommended for Basic tier |
Image Editing (OSS Models via WaveSpeed)
| Model | Cost | Type | Use Case |
|---|---|---|---|
| Qwen Image Edit | $0.02 per edit | OSS | Budget editing, bilingual |
| Qwen Image Edit Plus | $0.02 per edit | OSS | Multi-image editing |
| FLUX Kontext Pro | $0.04 per edit | OSS | Typography, professional |
| Default (Qwen Edit) | $0.02 per edit | OSS | Recommended for Basic tier |
Video Generation (OSS Models via WaveSpeed)
| Model | Cost | Type | Duration | Notes |
|---|---|---|---|---|
| WAN 2.5 | $0.05/sec | OSS | 5-15 sec | Text-to-Video, Image-to-Video |
| Seedance 1.5 Pro | $0.08/sec | OSS | 10-30 sec | Longer duration |
| Kling v2.5 Turbo (5s) | $0.21 per video | OSS | 5 sec | Image-to-Video |
| Kling v2.5 Turbo (10s) | $0.42 per video | OSS | 10 sec | Extended duration |
| Default (WAN 2.5) | $0.25 per video | OSS | ~5 sec | Recommended for Basic tier |
Audio Generation (OSS Models via WaveSpeed)
| Model | Cost | Type | Notes |
|---|---|---|---|
| Minimax Speech 02 HD | $0.05 per 1K chars | OSS | High-quality TTS |
| Default | $0.05 per 1K chars | OSS | ~$0.10-0.50 per audio |
Face Swap & Specialized (OSS Models via WaveSpeed)
| Operation | Cost | Type | Notes |
|---|---|---|---|
| Face Swap | $0.01-$0.03 | OSS | Basic to premium quality |
| Image Upscaling | $0.01-$0.06 | OSS | 2K/4K/8K options |
| 3D Generation | $0.02-$0.30 | OSS | Budget to premium |
OSS Advantage: WaveSpeed provides access to OSS models (Qwen, FLUX, Ideogram, WAN 2.5) at significantly lower costs than proprietary alternatives, enabling better value for users.
Production-Ready Basic Tier Proposal
Revised Limits for Production Launch
Price: $29/month ($290/year) - KEEP CURRENT PRICING
Rationale: Competitive pricing point, allows for sustainable margins with proper limits.
Proposed Limits
1. AI Text Generation (Unified Limit)
- Current: 10 calls/month ❌ TOO LOW
- Proposed: 50 calls/month ✅
- Rationale:
- Allows ~16-25 content generations/month (assuming 2-3 calls each)
- Enables users to experience Blog Writer, Story Writer, Podcast Maker, Social Writers
- Sustainable cost: ~$0.28/month (50 calls × $0.0056 average)
2. Token Limits (Per Provider)
- Current: 20,000 tokens/provider
- Proposed: 100,000 tokens/provider ✅
- Rationale:
- Allows ~33-50 LLM calls per provider (assuming 2K tokens/call)
- Provides buffer for longer content generation
- Aligns with unified call limit (50 calls × 2K tokens = 100K tokens)
3. Search APIs
- Tavily: 200 calls/month ✅ (Keep)
- Serper: 200 calls/month ✅ (Keep)
- Metaphor: 100 calls/month ✅ (Keep)
- Firecrawl: 100 calls/month ✅ (Keep)
- Exa: 500 calls/month ✅ (Keep)
- Rationale: Sufficient for research-heavy tools (Blog Writer, Podcast Maker, SEO tools)
4. Image Generation (OSS Models via WaveSpeed)
- Current: 5 images/month ❌ TOO LOW
- Proposed: 50 images/month ✅ (INCREASED - OSS models are cheaper)
- Rationale:
- OSS models (Qwen Image $0.03) are cheaper than Stability ($0.04)
- Allows users to generate images for Story Writer, Blog Writer, Social Media
- Cost: ~$1.50/month (50 × $0.03 using Qwen Image OSS model)
- Enables visual content creation workflows
- Default to Qwen Image OSS model for cost efficiency
5. Image Editing (OSS Models via WaveSpeed)
- Current: 30 edits/month
- Proposed: 50 edits/month ✅ (INCREASED - OSS models are cheaper)
- Rationale:
- OSS models (Qwen Edit $0.02) are cheaper than Stability ($0.04)
- Cost: ~$1.00/month (50 × $0.02 using Qwen Edit OSS model)
- Sufficient for image optimization workflows
- Default to Qwen Edit OSS model for cost efficiency
6. Video Generation (OSS Models via WaveSpeed)
- Current: 20 videos/month
- Proposed: 30 videos/month ✅ (INCREASED - OSS models available)
- Rationale:
- OSS models (WAN 2.5 $0.25 per 5s video) provide good value
- Allows ~6-10 full video projects/month (assuming 3-5 scenes each)
- Cost: ~$7.50/month (30 × $0.25 using WAN 2.5 OSS model)
- Enables Video Studio, YouTube Creator, Story Writer video features
- Default to WAN 2.5 OSS model for cost efficiency
7. Audio Generation (OSS Models via WaveSpeed)
- Current: 50 generations/month
- Proposed: 100 generations/month ✅ (INCREASED - OSS models are affordable)
- Rationale:
- OSS models (Minimax Speech 02 HD) provide high quality at $0.05/1K chars
- Sufficient for Podcast Maker, Story Writer narration
- Cost: ~$10.00-$25.00/month (depending on length, assuming 2K-5K chars per audio)
- Enables audio content workflows
- Default to Minimax Speech 02 HD OSS model
8. Monthly Cost Cap
- Current: $50.00
- Proposed: $45.00 ✅ (ADJUSTED - aligns with $40-50 target)
- Rationale:
- Protects against unexpected high usage
- Allows flexibility within limits
- Provides safety margin
- Aligns with $40-50 hard limit requirement
Cost Analysis: Proposed Basic Tier (OSS-Focused)
Monthly Cost Breakdown (Per User) - Using OSS Models
| Category | Usage | Cost per Unit (OSS) | Monthly Cost |
|---|---|---|---|
| LLM Calls | 50 calls | $0.0056 avg (Gemini Flash) | $0.28 |
| Search APIs | 200 searches | $0.002 avg | $0.40 |
| Image Generation | 50 images | $0.03 (Qwen Image OSS) | $1.50 |
| Image Editing | 50 edits | $0.02 (Qwen Edit OSS) | $1.00 |
| Video Generation | 30 videos | $0.25 (WAN 2.5 OSS, ~5s) | $7.50 |
| Audio Generation | 100 audios | $0.10-$0.50 avg | $10.00-$25.00 |
| Total Variable Cost | $20.68-$35.68 |
Margin Analysis (OSS-Focused)
Subscription Revenue: $29.00/month Variable Costs (OSS Models): $20.68-$35.68/month (depending on usage) Gross Margin: -$6.68 to +$8.32/month
✅ IMPROVEMENT: OSS models reduce costs significantly:
- Image generation: $0.03 vs $0.04 (25% savings)
- Image editing: $0.02 vs $0.04 (50% savings)
- Video generation: $0.25 vs $0.42 (40% savings)
Mitigation Strategy:
- Cost cap enforcement: Monthly cost cap of $45 prevents extreme losses
- OSS model defaults: Default to cheaper OSS models (Qwen, WAN 2.5)
- Realistic usage: Most users won't hit all limits simultaneously
- Average usage assumption: ~60-70% of limits = $12-25 cost = $4-17 margin
- Hard limit protection: $45 cap ensures we never exceed $50/user/month
Revised Basic Tier Limits (Production-Ready, OSS-Focused)
{
"name": "Basic",
"tier": SubscriptionTier.BASIC,
"price_monthly": 29.0,
"price_yearly": 290.0,
# AI Text Generation (Unified Limit)
"ai_text_generation_calls_limit": 50, # INCREASED from 10
# Token Limits (Per Provider)
"gemini_tokens_limit": 100000, # INCREASED from 20,000
"openai_tokens_limit": 100000, # INCREASED from 20,000
"anthropic_tokens_limit": 100000, # INCREASED from 20,000
"mistral_tokens_limit": 100000, # INCREASED from 20,000
# Search APIs
"tavily_calls_limit": 200, # Keep
"serper_calls_limit": 200, # Keep
"metaphor_calls_limit": 100, # Keep
"firecrawl_calls_limit": 100, # Keep
"exa_calls_limit": 500, # Keep
# Media Generation (OSS Models via WaveSpeed)
"stability_calls_limit": 50, # INCREASED from 5 (using Qwen Image OSS $0.03)
"image_edit_calls_limit": 50, # INCREASED from 30 (using Qwen Edit OSS $0.02)
"video_calls_limit": 30, # INCREASED from 20 (using WAN 2.5 OSS $0.25)
"audio_calls_limit": 100, # INCREASED from 50 (using Minimax Speech OSS)
# Cost Protection
"monthly_cost_limit": 45.0, # ADJUSTED from 50.0 (aligns with $40-50 target)
# OSS Model Defaults
"default_image_model": "qwen-image", # OSS model via WaveSpeed
"default_image_edit_model": "qwen-edit", # OSS model via WaveSpeed
"default_video_model": "wan-2.5", # OSS model via WaveSpeed
"default_audio_model": "minimax-speech-02-hd", # OSS model via WaveSpeed
# Features
"features": [
"full_content_generation",
"advanced_research",
"basic_analytics",
"all_tools_access", # All ALwrity tools accessible
"billing_dashboard",
"usage_tracking",
"oss_models_priority" # NEW: OSS models prioritized for cost efficiency
],
"description": "Perfect for individuals and small teams. Access all ALwrity features with generous limits powered by OSS AI models."
}
Tool Usage Scenarios & Limits
Scenario 1: Blog Writer User
- Workflow: 1 blog post = 3-5 LLM calls + 3-5 search calls + 1-2 images
- Monthly Capacity: ~10-16 blog posts (with 50 LLM calls)
- Cost: ~$0.50-$1.00 per blog post
- Status: ✅ FEASIBLE
Scenario 2: Story Writer User
- Workflow: 1 story = 2-3 LLM calls + 5-10 images + 5-10 audio + 5-10 videos
- Monthly Capacity: ~16-25 stories (LLM limit) OR ~3-6 stories (image/video limits)
- Cost: ~$2.00-$5.00 per story
- Status: ✅ FEASIBLE (limited by media, not LLM)
Scenario 3: Podcast Maker User
- Workflow: 1 podcast = 3-4 LLM calls + 3-5 search calls + 5-10 audio + optional 5-10 videos
- Monthly Capacity: ~12-16 podcasts (LLM limit) OR ~5-10 podcasts (audio limit)
- Cost: ~$1.00-$3.00 per podcast (without video)
- Status: ✅ FEASIBLE
Scenario 4: Social Media Content Creator
- Workflow: 1 post = 1-2 LLM calls + 1 image (optional)
- Monthly Capacity: ~25-50 posts (LLM limit) OR ~30 posts (image limit)
- Cost: ~$0.10-$0.15 per post
- Status: ✅ FEASIBLE
Scenario 5: Video Creator (YouTube Creator)
- Workflow: 1 video = 2-3 LLM calls + 5 scenes × (1 image + 1 audio + 1 video)
- Monthly Capacity: ~4-5 full videos (video limit) OR ~16-25 videos (LLM limit)
- Cost: ~$3.00-$5.00 per video
- Status: ✅ FEASIBLE (limited by video limit, not LLM)
Risk Mitigation Strategies
1. Cost Cap Enforcement
- Monthly cost cap: $50.00 (hard limit)
- Behavior: When cap reached, all API calls blocked until next billing period
- Protection: Prevents losses from extreme usage
2. Pre-flight Validation
- Implementation: Already in place
- Function: Validates limits BEFORE making API calls
- Benefit: Prevents wasted API calls on operations that would fail
3. Usage Monitoring & Alerts
- 80% Warning: Alert users at 80% of limits
- 100% Block: Block operations at 100% of limits
- Dashboard: Real-time usage tracking
4. Optimized Default Models
- Strategy: Use cost-effective models by default (Gemini Flash, GPT-4o Mini)
- Benefit: Reduces costs while maintaining quality
- User Control: Allow model selection for power users
5. Efficient API Usage
- Batching: Batch multiple operations where possible
- Caching: Cache research results and common queries
- Optimization: Continue optimizing tool workflows to reduce API calls
Pricing Page Updates Required
Current Issues
- Pricing page shows outdated limits
- Missing unified
ai_text_generation_calls_limitexplanation - Token limits don't match code (shows 1M/500K, code has 20K)
- Missing video/audio/image editing limits
- Missing cost transparency information
Required Updates
Basic Tier Display
💰 Basic Plan - $29/month ($290/year)
✨ All ALwrity Features Included:
✅ Blog Writer, Story Writer, Podcast Maker
✅ Image Generator & Editor
✅ Video Studio & YouTube Creator
✅ Audio Generator
✅ All Social Media Writers
✅ All SEO Tools & Dashboards
✅ Content Planning & Strategy Tools
📊 Usage Limits:
• 50 AI Text Generations/month (unified across all LLM providers)
• 100,000 tokens per provider (Gemini, OpenAI, Anthropic, Mistral)
• 200 Research Searches/month (Tavily, Serper)
• 500 Neural Searches/month (Exa)
• 30 AI Images/month
• 30 Image Edits/month
• 20 AI Videos/month
• 50 AI Audio Generations/month
• $50 Monthly Cost Cap (protects you from overages)
💡 Perfect for: Individuals, content creators, small teams
Implementation Checklist
Phase 1: Update Code Limits
- Update
pricing_service.pyBasic tier limits:ai_text_generation_calls_limit: 10 → 50gemini_tokens_limit: 20,000 → 100,000openai_tokens_limit: 20,000 → 100,000anthropic_tokens_limit: 20,000 → 100,000mistral_tokens_limit: 20,000 → 100,000stability_calls_limit: 5 → 30
- Run database migration script
- Test limit enforcement
Phase 2: Update Pricing Page
- Update
docs-site/docs/features/subscription/pricing.md - Update frontend pricing page component
- Add cost transparency section
- Add tool usage examples
- Add FAQ section
Phase 3: Update Documentation
- Update subscription rule file (
.cursor/rules/subscription.mdc) - Update API documentation
- Create user-facing pricing guide
Phase 4: Testing
- Test all tools with new limits
- Verify cost calculations
- Test limit enforcement
- Test cost cap enforcement
- Verify pre-flight validation
Cost Calculation Examples
Example 1: Blog Writer - 1 Blog Post (OSS Models)
Research: 3 Exa searches = $0.015
Outline: 1 LLM call (Gemini Flash) = $0.0056
Content: 2 LLM calls (Gemini Flash) = $0.0112
Image: 1 Qwen Image OSS = $0.03 (vs $0.04 Stability)
Total: ~$0.06 per blog post (saved $0.01 with OSS)
Example 2: Story Writer - 1 Story (5 scenes, OSS Models)
Outline: 1 LLM call = $0.0056
Script: 1 LLM call = $0.0056
Images: 5 × $0.03 (Qwen Image OSS) = $0.15 (vs $0.20)
Audio: 5 × $0.10 = $0.50
Videos: 5 × $0.25 (WAN 2.5 OSS) = $1.25 (vs $0.50-$2.10)
Total: ~$1.96 per story (higher video cost, but better quality)
Example 3: Podcast Maker - 1 Episode (10 min, 5 scenes, OSS Models)
Research: 3 Exa searches = $0.015
Script: 1 LLM call = $0.0056
Outline: 1 LLM call = $0.0056
Audio: 5 × $0.20 (Minimax Speech OSS) = $1.00
Video (optional): 5 × $0.25 (WAN 2.5 OSS) = $1.25
Total: ~$1.03 per podcast (without video)
Total: ~$2.28 per podcast (with video, OSS models)
Example 4: Social Media - 10 Posts (OSS Models)
Generation: 10 × 1 LLM call = 10 calls × $0.0056 = $0.056
Images: 10 × $0.03 (Qwen Image OSS) = $0.30 (vs $0.40)
Total: ~$0.36 for 10 posts (saved $0.10 with OSS)
Competitive Analysis
Similar AI Content Platforms
| Platform | Price | Limits | Notes |
|---|---|---|---|
| Jasper | $49/month | 50K words | Text-focused |
| Copy.ai | $49/month | Unlimited words | Text-focused |
| Writesonic | $19/month | 100K words | Text-focused |
| ALwrity Basic | $29/month | 50 LLM calls + media | Full platform |
ALwrity Advantage:
- Lower price point ($29 vs $49)
- Includes video, image, audio generation (competitors don't)
- Comprehensive tool suite (not just text)
- Better value proposition
Recommendations Summary
✅ APPROVED: Production-Ready Basic Tier (OSS-Focused)
Price: $29/month ($290/year) - KEEP
Key Changes (OSS-Focused):
- ✅ Increase AI Text Generation: 10 → 50 calls/month
- ✅ Increase Token Limits: 20K → 100K per provider
- ✅ Increase Image Generation: 5 → 50 images/month (OSS: Qwen Image $0.03)
- ✅ Increase Image Editing: 30 → 50 edits/month (OSS: Qwen Edit $0.02)
- ✅ Increase Video Generation: 20 → 30 videos/month (OSS: WAN 2.5 $0.25)
- ✅ Increase Audio Generation: 50 → 100 generations/month (OSS: Minimax Speech)
- ✅ Adjust Cost Cap: $50 → $45 (aligns with $40-50 target)
- ✅ Default to OSS Models: Qwen, WAN 2.5, Minimax Speech (cost-efficient)
Expected Outcomes:
- Users can experience all ALwrity features with generous limits
- Sustainable cost structure (~$20-35/user/month average with OSS models)
- Competitive pricing ($29 vs competitors $49+)
- Room for margin ($4-17/user/month average)
- Cost cap ($45) protects against losses (hard limit $40-50)
- OSS models provide 25-50% cost savings vs proprietary alternatives
Risk Level: 🟢 LOW (with cost cap enforcement and OSS model defaults)
Implementation Plan
Phase 1: Update Pricing Service & Database (Priority: HIGH)
1.1 Update pricing_service.py Basic Tier Limits
File: backend/services/subscription/pricing_service.py
Changes Required:
# In initialize_default_plans() method
{
"name": "Basic",
"tier": SubscriptionTier.BASIC,
"price_monthly": 29.0,
"price_yearly": 290.0,
# AI Text Generation (Unified Limit)
"ai_text_generation_calls_limit": 50, # Changed from 10
# Token Limits (Per Provider)
"gemini_tokens_limit": 100000, # Changed from 20,000
"openai_tokens_limit": 100000, # Changed from 20,000
"anthropic_tokens_limit": 100000, # Changed from 20,000
"mistral_tokens_limit": 100000, # Changed from 20,000
# Search APIs (Keep existing)
"tavily_calls_limit": 200,
"serper_calls_limit": 200,
"metaphor_calls_limit": 100,
"firecrawl_calls_limit": 100,
"exa_calls_limit": 500,
# Media Generation (OSS Models via WaveSpeed)
"stability_calls_limit": 50, # Changed from 5 (now includes WaveSpeed OSS)
"image_edit_calls_limit": 50, # Changed from 30
"video_calls_limit": 30, # Changed from 20
"audio_calls_limit": 100, # Changed from 50
# Cost Protection
"monthly_cost_limit": 45.0, # Changed from 50.0
}
Action Items:
- Update
initialize_default_plans()method inpricing_service.py - Run database migration to update existing Basic tier subscriptions
- Test limit enforcement with new values
- Verify cost calculations reflect OSS model pricing
1.2 Update WaveSpeed Model Pricing in pricing_service.py
File: backend/services/subscription/pricing_service.py
Changes Required:
# In initialize_default_pricing() method, update/add WaveSpeed OSS model pricing:
# Image Generation (OSS Models via WaveSpeed)
{
"provider": APIProvider.IMAGE,
"model_name": "qwen-image",
"cost_per_request": 0.03, # OSS model via WaveSpeed
"description": "WaveSpeed Qwen Image (OSS) - Fast generation"
},
{
"provider": APIProvider.IMAGE,
"model_name": "ideogram-v3-turbo",
"cost_per_request": 0.05, # OSS model via WaveSpeed
"description": "WaveSpeed Ideogram V3 Turbo (OSS) - Photorealistic"
},
# Image Editing (OSS Models via WaveSpeed)
{
"provider": APIProvider.IMAGE_EDIT,
"model_name": "qwen-edit",
"cost_per_request": 0.02, # OSS model via WaveSpeed
"description": "WaveSpeed Qwen Image Edit (OSS) - Budget editing"
},
{
"provider": APIProvider.IMAGE_EDIT,
"model_name": "qwen-edit-plus",
"cost_per_request": 0.02, # OSS model via WaveSpeed
"description": "WaveSpeed Qwen Image Edit Plus (OSS) - Multi-image"
},
{
"provider": APIProvider.IMAGE_EDIT,
"model_name": "flux-kontext-pro",
"cost_per_request": 0.04, # OSS model via WaveSpeed
"description": "WaveSpeed FLUX Kontext Pro (OSS) - Professional"
},
# Video Generation (OSS Models via WaveSpeed)
{
"provider": APIProvider.VIDEO,
"model_name": "wan-2.5",
"cost_per_request": 0.25, # OSS model via WaveSpeed (~5 seconds)
"description": "WaveSpeed WAN 2.5 (OSS) - Text-to-Video, Image-to-Video"
},
{
"provider": APIProvider.VIDEO,
"model_name": "seedance-1.5-pro",
"cost_per_request": 0.40, # OSS model via WaveSpeed (~5 seconds)
"description": "WaveSpeed Seedance 1.5 Pro (OSS) - Longer duration"
},
# Audio Generation (OSS Models via WaveSpeed)
{
"provider": APIProvider.AUDIO,
"model_name": "minimax-speech-02-hd",
"cost_per_input_token": 0.00005, # $0.05 per 1K chars
"cost_per_output_token": 0.0,
"cost_per_request": 0.0,
"description": "WaveSpeed Minimax Speech 02 HD (OSS) - High-quality TTS"
},
Action Items:
- Add WaveSpeed OSS model pricing entries
- Update default model selection logic to prefer OSS models
- Test cost calculation with OSS models
- Verify pricing accuracy against WaveSpeed API documentation
1.3 Update Default Model Selection Logic
Files:
backend/services/llm_providers/main_image_generation.pybackend/services/image_studio/create_service.pybackend/services/image_studio/edit_service.pybackend/services/video_studio/video_service.pybackend/services/audio_generation/audio_service.py
Changes Required:
- Default image generation to
qwen-image(OSS) instead of Stability - Default image editing to
qwen-edit(OSS) instead of Stability - Default video generation to
wan-2.5(OSS) instead of HuggingFace - Default audio generation to
minimax-speech-02-hd(OSS)
Action Items:
- Update
get_default_provider()methods to prefer WaveSpeed OSS models - Update model selection UI to show OSS models as default/recommended
- Add cost comparison tooltips showing OSS model savings
- Test all tools with OSS model defaults
Phase 2: Update Frontend & Documentation (Priority: HIGH)
2.1 Update Pricing Page
File: docs-site/docs/features/subscription/pricing.md
Changes Required:
- Update Basic tier limits to reflect new values (50 images, 50 edits, 30 videos, 100 audio)
- Add OSS model information and cost savings messaging
- Update cost examples to use OSS model pricing
- Add FAQ about OSS models and cost efficiency
Action Items:
- Update pricing page markdown
- Update frontend pricing component (if exists)
- Add OSS model badges/indicators
- Add cost comparison table (OSS vs proprietary)
2.2 Update Subscription Context & Components
Files:
frontend/src/contexts/SubscriptionContext.tsxfrontend/src/components/billing/EnhancedBillingDashboard.tsxfrontend/src/components/shared/UsageDashboard.tsx
Changes Required:
- Display OSS model indicators in usage dashboard
- Show cost savings from using OSS models
- Update limit displays to show new Basic tier limits
- Add tooltips explaining OSS model benefits
Action Items:
- Update limit displays in billing dashboard
- Add OSS model indicators in cost breakdown
- Update usage statistics to reflect new limits
- Test UI with new limit values
Phase 3: Testing & Validation (Priority: CRITICAL)
3.1 Limit Enforcement Testing
Test Cases:
- Test 50 AI text generation calls limit
- Test 50 image generation limit (OSS models)
- Test 50 image editing limit (OSS models)
- Test 30 video generation limit (OSS models)
- Test 100 audio generation limit (OSS models)
- Test $45 monthly cost cap enforcement
- Test pre-flight validation with new limits
- Test limit exceeded error messages
3.2 Cost Calculation Testing
Test Cases:
- Verify Qwen Image cost: $0.03 per image
- Verify Qwen Edit cost: $0.02 per edit
- Verify WAN 2.5 video cost: $0.25 per video
- Verify Minimax Speech cost: $0.05 per 1K chars
- Test cost aggregation across all operations
- Test cost cap enforcement at $45
- Verify cost display in billing dashboard
3.3 OSS Model Integration Testing
Test Cases:
- Test Qwen Image generation via WaveSpeed
- Test Qwen Edit editing via WaveSpeed
- Test WAN 2.5 video generation via WaveSpeed
- Test Minimax Speech audio generation via WaveSpeed
- Verify default model selection uses OSS models
- Test model fallback if OSS model unavailable
- Verify cost tracking for OSS models
Phase 4: Database Migration (Priority: HIGH)
4.1 Create Migration Script
File: backend/database/migrations/update_basic_tier_limits_oss.py
Script Requirements:
"""
Migration: Update Basic Tier Limits for OSS-Focused Pricing Strategy
- Increase AI text generation: 10 → 50
- Increase token limits: 20K → 100K per provider
- Increase image generation: 5 → 50
- Increase image editing: 30 → 50
- Increase video generation: 20 → 30
- Increase audio generation: 50 → 100
- Adjust cost cap: $50 → $45
"""
def upgrade():
# Update SubscriptionPlan for Basic tier
# Update existing UserSubscription records
# Clear pricing service cache
pass
def downgrade():
# Revert to previous limits if needed
pass
Action Items:
- Create migration script
- Test migration on staging database
- Backup production database before migration
- Run migration during maintenance window
- Verify all subscriptions updated correctly
Phase 5: Monitoring & Adjustment (Priority: MEDIUM)
5.1 Set Up Monitoring
Metrics to Track:
- Average cost per user per month
- Users hitting $45 cost cap
- Users hitting individual limits
- OSS model usage vs proprietary model usage
- Cost savings from OSS models
Action Items:
- Set up cost monitoring dashboard
- Create alerts for cost cap breaches
- Track OSS model adoption rate
- Monitor user satisfaction with limits
5.2 Adjustment Plan
Triggers for Adjustment:
- If average cost > $35/user: Consider reducing limits
- If >15% users hit cost cap: Consider increasing cost cap to $50
- If <20% users use video/audio: Consider reducing those limits
- If OSS models unavailable: Fallback to proprietary models
Action Items:
- Define adjustment criteria
- Create adjustment workflow
- Plan communication strategy for limit changes
Next Steps (Priority Order)
- CRITICAL: Update
pricing_service.pywith new Basic tier limits - CRITICAL: Add WaveSpeed OSS model pricing to
pricing_service.py - HIGH: Update default model selection to prefer OSS models
- HIGH: Create and run database migration
- HIGH: Update pricing page documentation
- HIGH: Test limit enforcement and cost calculations
- MEDIUM: Update frontend components with new limits
- MEDIUM: Set up monitoring and alerts
- LOW: Add OSS model indicators to UI
Monitoring & Adjustment Plan
Key Metrics to Track
- Average LLM calls per user per month
- Average media generation per user per month
- Average cost per user per month
- Users hitting cost cap
- Users hitting individual limits
Adjustment Triggers
- If average cost > $25/user: Consider reducing limits
- If >20% users hit cost cap: Consider increasing cost cap
- If <10% users use video/audio: Consider reducing those limits
- If churn rate high: Consider increasing limits
Review Schedule
- Week 1-2: Daily monitoring
- Month 1: Weekly review
- Month 2-3: Bi-weekly review
- Month 4+: Monthly review
Conclusion
The proposed Basic tier limits (OSS-Focused) provide:
- ✅ Access to all ALwrity features with generous limits
- ✅ Sustainable cost structure using OSS models (25-50% savings)
- ✅ Competitive pricing ($29 vs competitors $49+)
- ✅ Protection against losses ($45 cost cap, hard limit $40-50)
- ✅ Room for growth (can adjust based on usage)
- ✅ OSS-first strategy (Qwen, FLUX, Ideogram, WAN 2.5, Minimax Speech)
- ✅ Maximum user value while staying within cost constraints
Key Advantages of OSS-Focused Strategy:
- Cost Efficiency: 25-50% cost savings vs proprietary models
- Better Limits: Can offer more generations due to lower costs
- User Value: More value for the same $29/month price
- Sustainability: Lower costs = better margins = sustainable business
- Flexibility: Can adjust limits based on actual usage patterns
Recommendation: APPROVE for production launch with OSS-focused strategy.
Confidence Level: 🟢 HIGH (with proper monitoring, cost cap enforcement, and OSS model defaults)
Risk Mitigation:
- $45 cost cap protects against losses (hard limit $40-50)
- OSS model defaults ensure cost efficiency
- Monitoring allows quick adjustment if needed
- Realistic usage assumptions (60-70% of limits)