A survey of AI image generation services and inference providers, ranked by permissiveness for adult content, violence, and political figures. Focused on API access for scripted/automated workflows.
The Problem with Mainstream Providers
OpenRouter, DeepInfra, fal.ai, and Replicate serve open-weight models (FLUX, SDXL variants) but enforce content policies on their hosted endpoints. Standard FLUX checkpoints refuse explicit adult content, and nearly all cloud services have trained out the ability to generate recognizable political figures by name.
Service Comparison
| Service | Style | Adult content | Political figures | API | Cost |
|---|---|---|---|---|---|
| NovelAI | Anime/illustrated | Full | Partial | Community wrappers only | $15-25/mo |
| Runware | Any (400K+ CivitAI models) | Verify ToS | Partial | Yes (REST) | ~$0.002/image |
| CivitAI | Any (SDXL/FLUX) | Full (toggle on) | Partial | Yes (Orchestration API) | Pay-per-gen |
| Mage.space | Photorealistic | Full (paid tier) | Better than most | Yes | $8-16/mo |
| tensor.art | Anime/mixed | Full | Limited | No | Free tier |
| SeaArt.ai | Mixed | Full | Limited | No | Free tier |
| Venice.ai | Any | Full | Better than most | Yes | Pay-per-token |
Key Findings
NovelAI
Built specifically for illustrated adult content. The strongest option for anime and comic-book styles. Official image generation API deprecated March 2024 – access is now via community Python wrappers that authenticate as your account:
- HanaokaYuzu/NovelAI-API (async Python)
- LlmKira/novelai-python
Not available through OpenRouter, DeepInfra, or any other aggregator.
Runware
The only inference aggregator with direct access to CivitAI’s 400K+ community checkpoint catalog. Many CivitAI checkpoints are uncensored fine-tunes of SDXL and FLUX. REST API, ~$0.002/image, pay-per-generation with no monthly commitment. Verify adult content policy before committing – their web product and API policies may differ.
CivitAI’s model catalog is not accessible through OpenRouter, DeepInfra, fal.ai, or Replicate. Runware is the only aggregator with that integration.
CivitAI
Direct web UI access to community checkpoints with adult content toggle. Official Orchestration API available at developer.civitai.com for scripted generation. No official CLI – use the REST API directly or civitai-dl (pip) for model downloads.
Mage.space
Best option for photorealistic adult content and political satire. Handles political figures better than most services (describe rather than name for best results). REST API available on paid tiers.
Venice.ai
Privacy-focused, uncensored LLM and image inference. Does not log prompts. Available via OpenRouter (venice/uncensored:free) – useful for script and story generation where no-logging matters. Not available on DeepInfra (Venice uses DeepInfra as backend infrastructure but does not expose their privacy wrapper through DeepInfra’s public API).
Political Figures
No cloud service reliably generates photorealistic political figures by name – most have that trained out. Two workarounds:
Option A: Describe rather than name. “A 70-year-old American politician, grey hair, blue suit, large build” produces usable results on Mage.space and Runware/CivitAI checkpoints.
Option B: Fine-tune a LoRA on the target person using HuggingFace AutoTrain (free tier). Upload 15-25 images of the target, train a DreamBooth or LoRA checkpoint (~20-60 minutes), then use the trigger word on any service that accepts custom LoRAs (Runware, CivitAI, Mage.space).
LLM Text Generation (Scripts, Dialogue, Story)
For adult or political content in text generation:
| Provider | Access | Notes |
|---|---|---|
| Venice.ai | Direct API or via OpenRouter | No logging, explicitly uncensored |
| DeepInfra | Direct API | Llama/Mistral models less filtered than hosted GPT/Claude |
| OpenRouter | Direct API | Venice models available; other providers vary |
Venice is the strongest option for text that must not be logged or filtered.
Pricing Comparison (Image Generation, 1024x1024)
| Provider | Model | Cost/image |
|---|---|---|
| DeepInfra | FLUX.1-schnell | ~$0.002 |
| Runware | Community SDXL/FLUX | $0.001-0.010 |
| OpenRouter | FLUX.2 Klein | $0.014 |
| OpenRouter | FLUX.2 Pro | $0.030 |
| DeepInfra | FLUX.2 Pro | $0.015 |
| Mage.space | Photorealistic | ~$0.01 (subscription) |
| NovelAI | Illustrated | ~$0.01 (subscription) |
DeepInfra is cheapest for standard FLUX models but does not carry uncensored community checkpoints. Runware’s community model access makes it the best value for adult content specifically.
Reference
- NovelAI: https://novelai.net
- Runware: https://runware.ai
- Runware API docs: https://developer.runware.ai
- CivitAI: https://civitai.com
- CivitAI API docs: https://developer.civitai.com
- Mage.space: https://www.mage.space
- Venice.ai: https://venice.ai
- Venice on OpenRouter: https://openrouter.ai/venice
- HuggingFace AutoTrain (LoRA): https://huggingface.co/autotrain
Want to stay in touch?
- Join my Signal announce-only group to be notified when I have a new essay up and other important announcements.
- For discussion, join the libertygardeners Signal group.
- Subscribe to my mailing list.
- Email: [email protected]
- Signal: archerships.43
- Website: archerships.com
- Other social media: Substack | Twitter | Facebook | Nostr | Odysee
If you’d like to support my work:
- Share my posts.
- Become a subscriber to my newsletter.
- Attend my live events (dinner parties, conferences, pop-up cities, etc).
- Introduce me to like-minded people.
- Make a one-time donation to support my work: Crypto | Fiat
- Hire me for privacy / crypto / censorship consulting.
If there is a topic you’d like me to cover, please let me know!
Questions, comments, and suggestions are welcome.