ComfyChair

sh.hnet.comfychair

Native Android client for ComfyUI - generate AI images and videos on mobile

ComfyChair brings the power of ComfyUI to your Android device with a native, Material Design 3 interface. Connect to your ComfyUI server and generate AI images and videos right from your phone. What you can do: - Generate images from text prompts with live progress updates - Transform existing images or use inpainting with a built-in mask editor - Create AI videos from text or animate still images - Manage your generation queue, cancel jobs, or prioritize requests - Browse and organize all your creations in a unified gallery - Work offline with cached data when your server is unavailable Flexible workflow support: - Works with both traditional Checkpoint and modern UNET workflows (Flux, Z-Image, and more) - Stack up to 5 LoRAs with individual strength controls - Create new workflows from scratch or import existing ones from ComfyUI - Full node graph visualization and editing with a searchable node browser - Organize nodes into groups and add notes with Markdown support Built for mobile: - Smooth, responsive interface built with Jetpack Compose - Manage multiple servers with per-server settings - HTTP Basic Auth and API key authentication for secured servers - Pinch-to-zoom media viewer with swipe navigation - Batch operations for saving, sharing, or deleting multiple items - Backup and restore your configuration across devices - Available in English, German, French, Polish, Spanish, and Chinese Requirements: - A running ComfyUI server (local or remote) - Android 13 or higher ComfyChair is free, open source (GPL-3.0), and not affiliated with the official ComfyUI project.

First release: Nov 3, 2026, 1 total release.

Most recent release: Nov 3, 2026.

Repo

Appears in 0 app stacks.

0 sats / 0 zaps received in the past year.

Sats Received

Underlying data available via MCP: app_zaps, app_releases.

Zap Count

Underlying data available via MCP: app_zaps, app_releases.

Releases

Nov 3, 2026 0.8.13

## Privacy and security - **Generated media is kept in memory only by default.** Previously, the default settings saved previews and videos on the device when the app went to the background, and video playback wrote temporary files even with the media cache disabled – an oversight. The new "Media storage" setting (Memory only / Memory, saved on exit / On the device) replaces the two former cache switches; switching to Memory only removes all stored media, including leftovers from earlier versions. In Memory only mode the app is also hidden from the recents screen preview. - **Certificates are verified.** Certificates trusted by Android are accepted as usual. Self-signed or private-CA certificates are accepted only after you confirm their SHA-256 fingerprint, and from then on only that certificate is accepted for the server; a changed certificate is reported. Existing users with such servers will be asked once. - **No silent fallback from HTTPS to HTTP.** A server that used HTTPS is never switched to unencrypted HTTP without asking. - **Android cloud backup is disabled.** Prompts, server addresses and credentials are no longer uploaded to Google Drive. Use the backup export in the app's settings to move your configuration. - Prompts, file names and workflow names no longer appear in the debug log; copied metadata is marked as sensitive. ## New features

More…

- Preview of all results of a job, with a pager and thumbnail strip on every generation screen; videos play inline. - "Use as source" sends any generated image to Image to Image or Image to Video. - Built-in templates: Krea-2 Turbo, Krea-2 Edit, Krea-2 Identity Edit (LoRA), Krea-2 Turbo inpainting with LanPaint, SeedVR2 upscaling, and base Qwen Image Edit (#41). - Each workflow, built-in or custom, can be switched on or off; the generation screens offer only the enabled ones. Replaces "Show built-in workflows". - LoRA strength from -100 to 100 with a ±10 stepper and a fine slider (#35, #38). - Prompts can be mapped to primitive String nodes (#36, #42). - Workflows without a positive prompt (e.g. upscalers) are supported (#37). - Readable validation errors, and a warning when ComfyUI skips outputs. ## Bug fixes - Crash loop on servers with very large `/object_info` responses (#44). - Stale connection after editing the connected server (#49). - LoRA chain not injected for non-core model loaders such as GGUF (#29, #48). - Workflow Editor edits overriding uploaded images (#40). - Seed, batch size, upscale method and other common settings ignored outside Text to Image, which failed some workflows with HTTP 400. - Upscale method menu offering a single option outside Text to Image. - First generation after start rejected with HTTP 400 before model lists were loaded. - Seeds above 2147483647 set in the Workflow Editor being truncated. - Metadata showing the positive prompt as negative for models without a negative prompt (Flux, Krea-2). - Smoother saving on exit and faster workflow list.