Seeking Uncensored Image-to-Image APIs and ComfyUI Workflows
Efficient_Topic_1653 · reddit · 2026-07-09
A developer is building a Silly Tavern extension that requires updating character portraits as the story progresses. Because models like Grok enforce strict censorship on image-to-image requests not generated natively on their platforms, the author is looking for more lenient APIs (like Tongyi Qianwen/Qwen) or local recursive models.
They have a workstation with 48GB of VRAM and have tested decent results on Hugging Face. However, they are struggling to find a streamlined ComfyUI workflow dedicated solely to image-to-image (as existing ones are too bloated or broken). They are asking the community for simple API providers and clean "input image + natural language instruction -> output image" workflow recommendations.
More from coding & agent
- An MCP server signs every AI agent tool call into a verifiable Merkle chain — Funky_Chicken_22 · 2026-07-22
- Claude Code skill uses 10 Markdown rules to make outputs ADHD-friendly — alex_verem · 2026-07-22
- A Firecracker-based platform says it can host 6,000 AI agents on one 256 GB server — maritime_sh · 2026-07-22
- A better path to agent autonomy is running waves, finding friction, and iterating — JnBrymn · 2026-07-22
- AI agent designers map the visual and tonal cues behind companionship products — Unlikely-Platform-47 · 2026-07-22
- Coding agents are heading toward an AI-writes, AI-reviews, human-approves workflow — aftahi_ai · 2026-07-22