diff --git a/README.md b/README.md index 2c3bf66..0a0c5e2 100644 --- a/README.md +++ b/README.md @@ -1,42 +1,34 @@ # ComfyUI venice.ai Flux/SDXL/LLM API Nodes -An unofficial custom node implementation for ComfyUI that integrates with venice.ai's Flux and SDXL image generation models (and upscale + text gen). This project is adapted from [ComfyUI-FLUX-TOGETHER-API](https://github.com/BZcreativ/ComfyUI-FLUX-TOGETHER-API) to work with the venice.ai API. +An unofficial custom node implementation for ComfyUI that integrates with venice.ai's Generative AI services such as Image, Text and TTS Generation models (as well as Upscale and Enhance). This project is adapted from [ComfyUI-FLUX-TOGETHER-API](https://github.com/BZcreativ/ComfyUI-FLUX-TOGETHER-API) to work with the venice.ai API. +**Note**: TTS is in Beta per Venice (as of 15th June 2025). The node for it "Generate Speech (Venice)" is set to BETA/Experimental too. To be able to find it with node search you have to enable "Show experimental nodes in search" setting in ComfyUI. -Disclaimer: I just made this on a whim because someone wanted something similar to Together.AI custom nodes but have them use venice instead and also, I don't have access to any API keys for any of the mentioned services. i guess it should work, sometimes? +Disclaimer: I originally made this on a whim because someone wanted something similar to Together.AI custom nodes but have them use venice instead. I'm also not affiliated with Venice.AI. Idk is this good enough for a disclaimer or something -Speech API/TTS - Done (its in beta, Comfyui Settings > search for experimental and turn on the setting for having it show experimental nodes in search) -New models may have different voices which are not handled right now so any new model might not be surported until code updated to dynamically update voices upon model change but this may pose an issue for workflow sharing so i probably have to just bunch all the voices together, unless theres a better option, maybe a string input +## Nodes: (Text gen node is missing but will be updated soon) -Generate Image node: now has safe mode toggle and some more tooltips, and revised seed limits +![3 of the main nodes showcased](./gh_assets/some_nodes.png) +Image and Text models (as of 15th June 2025 | for Speech gen only tts-kokoro is available) +![Image gen models](./gh_assets/image_gen_models_15june.png) ![Text gen models](./gh_assets/text_gen_models15june.png) -~~# todo: fix upscale node (see updated api doc)~~ DONE, its now called "I2I Enhance + Upscale (Venice), its a new node so the old one wont work anymore +~~# todo: inpainting~~ Deprecated by venice, new thing is coming for it at some point -~~# todo: inpainting~~ Deprecated by venice, new thing coming ig +todo: actual log, maybe separate logging file for less clutter from comfyui stuff, maybe, maybe... eeeee -# todo: LLM characters list +todo: LLM characters list -# todo: use variants api (currently in beta) | idk what happened to this +todo: use variants api (currently in beta) | idk what happened to this -# Todo: chat history/memory/context for LLM +Todo: chat history/memory/context for LLM -*Implemented stuff from venice api:* - -- Image generation endpoint -- Chat completion/Text generation endpoint -- Upscale image endpoint (untested) - -*Missing-ish Stuff from venice api:* - -- A way to update hardcoded things like: - - available models using list models enpoint - - available style presets using styles endpoint +todo - Some Validation on queue for API limits like prompt max length or width/height - these are different for different models so this isnt planned to be implemented unless the api exposes those limits somehow + - i'd have an error thrown before it gets sent to api about some value being too high or too low (though api should send error back for now either way) + - maybe the api already exposes those limits, the steps limit is afaik at least, need to look into it -If there's no "untested" on any of the points it means it should work but there's not been extensive testing on it - -## Below ReadMe text is slightly altered from original Flux Together API readme, it was not reworked or anything so it might not be correct or up to date +## Below ReadMe text is slightly altered from original Flux Together API readme, it was not reworked or anything so its likely not correct or up to date ### Features @@ -50,6 +42,8 @@ If there's no "untested" on any of the points it means it should work but there' ### Installation - these instructions are a mess +0. Before proceeding, check if you can find these nodes through ComfyUI-Manager interface rather than following the instructions. + 1. Clone this repository into your ComfyUI custom_nodes directory: ```bash diff --git a/gh_assets/image_gen_models_15june.png b/gh_assets/image_gen_models_15june.png new file mode 100644 index 0000000..7ab8a1c Binary files /dev/null and b/gh_assets/image_gen_models_15june.png differ diff --git a/gh_assets/some_nodes.png b/gh_assets/some_nodes.png new file mode 100644 index 0000000..9d1b44a Binary files /dev/null and b/gh_assets/some_nodes.png differ diff --git a/gh_assets/text_gen_models15june.png b/gh_assets/text_gen_models15june.png new file mode 100644 index 0000000..cacbcfd Binary files /dev/null and b/gh_assets/text_gen_models15june.png differ