Finden Sie, was Sie für Ihren Workflow benötigen.

Qwen-Image1024x10248GB VRAM

Qwen-Image edge-guided restyling

The source image is resized to the chosen megapixel count while preserving its aspect ratio. It supplies both the starting image latent and the Canny edges used for structural guidance. Denoise controls how strongly the image is regenerated; control strength sets adherence to its edges. At the source default of full denoise, edge guidance provides the main structural constraint. Edge thresholds tune which details become guidance. The optional Lightning LoRA starts disabled, matching the source workflow. The source notes suggest 4 steps and CFG 1.0 when enabling it, versus the preserved defaults of 20 steps and CFG 2.5 without it. These settings remain separate controls.

Bria FIBO1024x10248GB VRAM

Bria FIBO image edit

Edit an uploaded image with text instructions using Bria FIBO and receive one edited image in the Pipe's preview output. For a first run, upload a Source image, replace the architectural night-scene example in Edit instruction with your desired changes, and adapt Negative prompt to describe unwanted features. Leave Structured prompt empty and retain the initial Guidance and Steps values, then run the Pipe. The edit applies to the entire image. This workflow has no mask input or output-size controls, and it does not preserve uploaded transparency. The selected output returns an image preview rather than using the original SaveImage stage. The node's secondary text output is not exposed. Guidance ranges from 3 to 5; higher values make the image follow the instruction more closely. Steps sets sampling iterations from 20 to 50. Seed -1 requests fresh randomization on each run; use a fixed seed to control generation randomness. For programmatic editing, enter a Bria-compatible structured edit prompt as JSON in Structured prompt, using it instead of the ordinary edit instruction. Leave this field empty for normal text editing; it requires the provider's structured edit format rather than arbitrary JSON. Moderation settings controls whether the individual provider checks are supplied. With true selected, Prompt moderation, Source image moderation and Result moderation control their respective checks. The supplied defaults select true with all three checks disabled. Selecting false omits the individual settings. Generation uses the Bria partner API, sends the source image and editing inputs to the provider, and requires the runtime's Comfy API authentication. Provider charges apply.

Nano Banana 21024x10248GB VRAM

Nano Banana 2 image generator

Generate images from a text prompt with Nano Banana 2 or Nano Banana 2 Lite and return the final images to Sweet Tea. A configured, authenticated Gemini API connection is required. Requests send your prompt and provider instructions to Gemini and use provider billing. This workflow has no reference-image input. For a first run, keep Nano Banana 2 Lite, aspect ratio 1:1 and resolution 1K. Use the supplied racing-poster prompt or replace it with a description of your subject, composition, style and desired text, then run the Pipe. Put unwanted elements into the same prompt as constraints; there is no separate negative prompt. Choose the model before resolution. Lite uses 1K; the existing guidance specifies 1K, 2K and 4K for Nano Banana 2. Return resolution to 1K before switching back to Lite. Choose an explicit aspect ratio for the desired framing; auto has no reference image to match in this workflow. Thinking level selects MINIMAL or HIGH. Temperature controls sampling randomness from 0 to 2, and top P controls the nucleus sampling threshold from 0 to 1. Keep their supplied defaults for a first run. Seed -1 requests fresh randomization; a fixed seed provides only best-effort consistency, and changing model or sampling settings can change the result. System instructions guide the provider's interpretation of your prompt; the supplied instructions prioritize image generation. IMAGE+TEXT can request accompanying provider text, but this Pipe returns only the final image output. Text and intermediate thinking images are not included. The output stage transports images to Sweet Tea without permanently saving PNG files.

Wan 2.1 VACE 14B1024x10248GB VRAM

Wan 2.1 first-to-last frame video

Generate a short, silent video transition between a first image and a last image, guided by a motion prompt. Both images are required. Wan 2.1 VACE conditions the endpoints; it does not guarantee pixel-exact reproduction. Uploaded alpha masks are not used. For a first run, upload endpoint images with matching dimensions and compatible framing, replace the example motion prompt, keep the 512 × 512 output, choose length setting 4 and leave Lightning disabled. Keep batch size 1 and the supplied sampling and encoding defaults. The negative prompt lists unwanted appearance and motion; its supplied Chinese exclusions can be retained or replaced. Choose output width and height as multiples of 16, each at least 16. Other values round down. The source recommends resolutions up to 720p and clips shorter than five seconds. Length is a positive integer setting converted to 16 × setting + 1 generated frames. Setting 4 gives 65 frames, approximately 4.06 seconds at 16 FPS. The preserved default of 5 gives 81 frames, approximately 5.06 seconds. Playback FPS changes speed and duration without generating additional frames. Lightning enables the fixed Lightx2v LoRA and switches to its separate steps and CFG controls: defaults of 6 steps and CFG 1, compared with standard defaults of 20 steps and CFG 6. Lightning LoRA strength applies only in Lightning mode. Frame guidance strength scales endpoint conditioning; CFG controls text guidance. Both modes share seed, sampler, scheduler, denoising amount and sampling shift. Seed -1 randomizes each run; use a fixed seed to compare changes with the same initial noise. Advanced playback and encoding settings include bit depth, file type, codec and CRF. The default uses 8-bit video creation and H.264 encoding with automatic file type. AV1 and other containers are selectable; choose compatible container and codec settings. Ten-bit video requires compatible encoding and playback. The graph has no audio input and this interface returns one video output.

SDXL Refiner 1.01024x10248GB VRAM

Replace video background

Replace an uploaded video's background with a selected solid color using Bria. For a first run, upload your source video, keep Background color set to Black, and run the Pipe to download the processed video. Choose another listed background color to change the fill behind the foreground subject. The output uses a solid background; no transparent-output mode is provided. Rerun seed controls whether Bria processes the video again. Its default of -1 selects a random value on each run. Bria results remain non-deterministic even when you reuse a seed. Under advanced Video output settings, you can choose the file type and codec or leave both on auto. Quality (CRF) defaults to -1 for automatic quality and accepts explicit values from 0 to 63. These settings configure output encoding rather than background removal.

SDXL with separate refiner prompts

Generate an image or batch from text with SDXL Base and Refiner, using separate prompts for composition and finishing detail. Both fixed checkpoints are required: sd_xl_base_1.0.safetensors and sd_xl_refiner_1.0.safetensors, available to ComfyUI in its checkpoints directory. For a first run, keep 1024 × 1024 pixels, Image count 1 and the default sampling settings. Enter the subject, setting and composition in Positive prompt, then describe compatible finishing details in Refiner prompt. You can also run the supplied blue galaxy bottle example. The refiner continues the base latent; its prompt influences final denoising rather than generating a separate composition. Negative prompt and Refiner negative prompt apply to their respective stages, so edit both for exclusions you want throughout generation. Width and height are pixels in multiples of eight. Image count generates a batch at those dimensions. Larger dimensions and batches require more memory. In advanced settings, Seed controls the base stage's initial noise: -1 randomizes each run, while a fixed value reuses initial noise. Refiner seed controls randomness consumed by stochastic refiner samplers despite fresh initial noise being disabled. Fix both seeds when comparing settings with those samplers. Advanced controls provide separate prompt guidance, samplers and schedulers for each stage. Defaults use Euler and the normal scheduler for both stages. Start with matching sampler and scheduler choices. Each stage's Schedule steps value describes the full noise schedule, not just the steps that stage executes. Defaults use 25 schedule steps: the base starts at 0 and hands off at 20 with leftover noise; the refiner resumes at 20 and finishes without adding fresh initial noise. When adjusting the split, keep both schedule lengths equal and Base handoff step aligned with Refiner handoff step under matching schedules. The default Refiner end step of 10000 means finish the available schedule, not run 10000 steps. This tool takes text inputs and always runs both generation stages. It has no uploaded-image mode, model selector or refiner bypass.

Image to imageWan 2.21024x10248GB VRAM

Wan 2.2 Image to Video

Animate a starting image into a silent video using Wan 2.2 and a fixed four-step LightX2V LoRA pair. For a first run, upload a source image, leave its stage enabled, and enter a prompt describing the subject's action, scene, and camera movement. Keep the default dimensions, frame count, and sampling settings. The default result is 81 frames at 640 × 640, played at 16 FPS, for about 5.1 seconds. The supplied negative prompt lists unwanted artifacts and can be retained or edited. To generate without a starting image, bypass the Source image stage and describe the complete scene in the prompt. Leaving the upload empty while that stage is enabled does not select this mode. The same image-to-video models remain in use. Width and height specify generated pixels in multiples of 16. Frame count uses increments of four starting at 1. Playback FPS changes playback speed and duration without creating additional frames; duration is frame count divided by FPS. Keep latent batch size at 1 for a single clip: decoded frames feed one video output, with no separate per-item file outputs. The graph includes no audio input, upscaling stage, end-frame control, or transparency-preservation path. Advanced controls expose separate high-noise generation and low-noise refinement settings. Seed -1 randomizes the initial noise; a fixed nonnegative seed reuses that variation. The default Euler/simple schedules both have length 4, with the first stage running from 0 to 2 and refinement continuing from 2 to 4 without fresh noise. Keep schedule lengths and handoff boundaries coordinated. These defaults match the fixed four-step acceleration LoRAs; changing steps alone does not establish a supported higher-quality preset. Guidance and schedule shift remain independent for each stage. The model pair and acceleration LoRAs are fixed; this interface does not offer custom LoRA selection or strength controls. Video encoding offers container, codec, compression quality, and bit-depth choices. For the first run, retain auto container and codec, CRF -1, and 8-bit video. The source notes possible compatibility limitations for 10-bit playback.

ModellCivitai

Kristen Nobody

Follow me to make sure you see new poses and Nobodys when I post them. Things move fast on this site, it's easy to miss.

TextualInversionSD 1.5

ACE-Step1024x10248GB VRAM

ACE-Step 1.5 music to MP3

Generate a music track from a musical description and lyrics with ACE-Step 1.5 Turbo, then export it as MP3. No media upload is required. For a first run, keep the supplied neo-soul prompt and lyrics, both duration fields at 120 seconds, Batch size at 1, and the supplied sampling settings, then run the tool and retrieve the MP3 output. To create your own song, describe the genre, instruments, mood, vocal character, and groove in Music prompt, then replace Lyrics with the words to sing. Optional markers such as [Verse] and [Chorus] communicate song structure. Choose the lyrics language, tempo in beats per minute, time signature, and musical key. Language uses the supplied language-code choices; en means English and unknown leaves the language unspecified. Time signature offers 2, 3, 4, and 6. Audio duration controls generated length in seconds. Conditioning duration is an independent field supplied to music conditioning: change both to the same value when changing track length. They are not synchronized. Batch size controls the number of audio latents; use 1 for a single-track run. The supplied evidence does not establish how larger batches are presented in the exported result. Generate audio codes is enabled by default and adds a language-model stage before audio sampling. You can disable it while retaining text-conditioned audio generation. Audio-code seed, guidance, temperature, top-p, top-k, and min-p apply when this mode is enabled. They are separate from the audio sampler's seed and guidance. Both seeds default to -1 for randomization on each run; set both explicitly when comparing settings with audio-code generation enabled. Advanced audio sampling starts with eight steps, Euler, the simple scheduler, guidance 1, and full denoising. Sampling shift is a separate model configuration control with source default 3. No further shift tuning guidance is supplied. This workflow starts from an empty audio latent and has no uploaded audio reference or editable negative prompt; lowering denoising strength does not add audio editing. MP3 quality offers V0 variable-rate encoding or fixed 128k/320k bitrates. The checkpoint is fixed to ace_step_1.5_turbo_aio.safetensors; this tool does not offer model selection.

ModellCivitai

VAE: Final Prune VAE

This is an earlier version of a stable VAE. Compared to other VAEs, it has a higher level of stability. I am not the creator, but I could not find this VAE on the website, so I am sharing it here.

Glass Style

Textual Inversion Style Embedding trained on glass and crystal . It's an odd one, a little experimental! Sometimes it will give you lovely translucent clothing details, or even clothing made of glass. Sometimes it...

TextualInversionSD 1.5

Recraft V41024x10248GB VRAM

Kling video editor

Edit an uploaded video with text instructions and, optionally, one reference image. The Pipe returns one edited video with the same duration as the source. For a first run, upload your source video, then rewrite Edit instructions to describe what to change and what to preserve. The supplied example refers to a man, an ornate bird, and a desert background; replace those details with instructions for your footage. Keep the default kling-v3-omni model and 720p resolution to start. Reference image is enabled by default. Upload an image and explain how it should guide the edit, or bypass the Reference image stage for text-only editing. When bypassing it, remove references to an uploaded image from your instructions. This Pipe provides one reference-image input. Choose 720p or 1080p output. Keep original audio retains the source video's audio when enabled. Editing model also offers kling-video-o1. Re-run seed defaults to -1 for a fresh value each run; fixed seed values do not make results reproducible. Include desired changes and unwanted changes in the same instruction field. Requests to preserve characters, actions, or details guide the model and are not guarantees. Editing uses the Kling provider API, which receives the supplied media and instructions and involves a paid provider call.

WaveSpeed image upscale

Upscale an uploaded image to a 2K, 4K, or 8K provider preset with SeedVR2 or Ultimate through WaveSpeed. For a first run, upload your source image, leave the model at SeedVR2 and target resolution at 4K, then run to view the returned upscaled image. The source image is required; no bundled example image is needed. Choose SeedVR2 or Ultimate to change the upscaling model. Both use the same target-resolution control. To compare their results, run each model with the same source image and resolution preset. The resolution choices are provider presets rather than explicit pixel width and height settings. This operation sends your image to WaveSpeed and uses a paid provider call. The workflow uses only the loaded RGB image; uploaded alpha transparency is not preserved.

HappyHorse text to video

Generate a 3–15 second video from an English or Chinese text prompt with HappyHorse. Configured ComfyUI partner API access is required; no uploaded media is needed. For a first run, use the supplied floral fashion prompt or describe your own scene, action, visual style, and camera movement. Keep HappyHorse 1.1, 720P, 16:9, and a 5-second duration, then run the tool to receive the video in Sweet Tea. Choose 720P or 1080P and a duration from 3 to 15 whole seconds. Requests such as “8K” or “60 fps” in the prompt do not override the resolution control; there is no frame-rate control, separate negative prompt, or reference-image input. HappyHorse 1.0 is an alternative model. When selecting it, choose 16:9, 9:16, 1:1, 4:3, or 3:4. HappyHorse 1.1 also supports 21:9, 9:21, 5:4, and 4:5. Check the aspect ratio whenever you switch models. Seed -1 requests a fresh random seed each run; a nonnegative seed controls generation variation. Enable Watermark to request the provider's AI-generated watermark; use this control rather than relying on watermark instructions in the prompt. Advanced Video output settings offer automatic formatting or MP4, MKV, and WebM file types, with automatic, H.264, or AV1 codec choices. Leave both settings on auto for the first run.

Color and edit images with Gemini

Color an illustration or apply prompted image edits with Gemini, or generate an image from text alone. For a first run, leave Source image enabled, upload an illustration, and use “Color this image.” Start with Auto aspect ratio and 1K resolution. Add palette, style and details to preserve or change to your prompt. An upload is required while Source image is enabled. For text-only generation, disable Source image and replace the coloring prompt with a complete description of the image you want. No upload is needed in this mode. Choose an explicit aspect ratio to specify the composition. With a reference image, Auto follows its aspect ratio. Resolution selects a 1K, 2K or 4K target; 2K and 4K use Gemini's native upscaler. Model selects Gemini 3 Pro Image or Nano Banana 2 (Gemini 3.1 Flash Image). Seed -1 randomizes each run; a fixed seed requests best-effort consistency without guaranteeing identical results. System instructions control Gemini's underlying interpretation of the prompt; the default directs it to produce an image. Keep Response format at IMAGE for image-only requests. IMAGE+TEXT also requests text, but this Pipe returns only generated images and does not retain the text response. This workflow sends the prompt and any enabled reference image to the Gemini service and uses a provider-priced request. Access to that service is required. Results are returned through an image preview output. The reference connection carries RGB image data only and does not preserve uploaded transparency.

Image glow

Create an image with a sampled, optionally tinted glow from an uploaded source image. For a first run, upload your image and run with the saved settings: Intensity 2, Radius 29, Glow tint #0a9299, screen blending and Threshold 100. These settings apply a teal-tinted effect rather than a neutral preview. Change intensity or radius and run again to compare. Set Intensity to 0 to preview the source without the effect. Intensity uses a unitless 0 to 100 scale and scales the glow before blending. Radius uses a 0 to 100 scale rather than direct pixel units: the sampling radius is 0.012 × Radius squared in pixels, so the saved value 29 gives about 10 pixels. Sampling includes slight spatial jitter. Radius 0 also disables the effect. Glow tint multiplies the sampled glow's colors. Enter an RGB hex color; #ffffff leaves those colors unchanged. The saved #0a9299 favors teal. Pure black #000000 skips the tint multiplication rather than suppressing the glow. Advanced controls choose brightness selection and blending. Threshold is an inverted 0 to 100 scale: low values favor bright samples, while high values include darker samples. The brightness selection has a soft transition. Choose the exact lowercase blend name add, screen, soft, overlay or lighten; enter it exactly if shown as a text field. Add sums the source and glow, screen combines their inverse values, and lighten keeps the larger value in each channel. Soft and overlay use tonal blends that can darken areas when the sampled glow is weak, so lowering intensity is not necessarily a fade back to the source in those modes. Intensity 0 disables every mode. Strong settings can clip colors to black or white. The active effect also adds a small deterministic dither. Output dimensions follow the source image. Uploaded transparency is not preserved because the image loader's separate alpha mask is unused. This output applies only glow; the other image adjustments in this family are separate tools.

Kling first and last frame video

Generate one video from a starting image, a motion prompt, and an optional ending image using Kling. Provider access is required; generation sends inputs to Kling and uses a paid provider call. The result is returned as a video preview. For a first run, upload the first and end frames, replace the example prompt to match your images, and keep the five-second, 1080p defaults. The end-frame stage starts enabled and needs an upload. To animate only the starting image, bypass the End frame stage; no end-frame upload is then needed. Describe subject motion, camera movement, desired appearance, and the transition in the motion prompt. The same field accepts desired and unwanted content; there is no separate negative prompt. Duration ranges from 3 to 15 seconds. Resolution offers 720p, 1080p, and 4k. Kling v3 Omni is the default model; Kling Video O1 is also selectable. Generate audio is supported only with Kling v3 Omni, so leave it off with Kling Video O1. Seed -1 randomizes each run; a fixed seed does not make provider results reproducible. The advanced Video output stage provides file-type and codec choices. Keep both on auto for the first run. This interface supports first-frame and first-to-last-frame generation. Storyboards remain disabled because editable segment prompts and durations are absent. Reference-image guidance is not connected in this Pipe.

Recraft V4 text to image

Generate one to six images from a written description using Recraft V4 or V4 Pro. No source image is required. The host environment must provide access to the Recraft API; prompts are sent to the provider, and generation incurs provider usage charges. For a first run, replace the example prompt with your subject, composition, lighting and desired style. Keep Recraft V4, Output size 1024x1024 and Image count 1. Leave Negative prompt blank unless you want to describe elements to discourage. Run the Pipe to receive the generated images. Select the model before setting Output size. These are separate controls: switching to V4 Pro does not automatically update the size. The source defaults are 1024x1024 for V4 and 2048x2048 for V4 Pro. Enter sizes as widthxheight in pixels. The complete model-specific size lists are not supplied, so other dimensions must be supported by the selected model or the provider may reject them. Image count ranges from one to six; model choice and image count affect provider usage cost. Leave Seed at -1 to randomize it each run. Seed controls whether the generation node re-runs; a fixed seed does not reproduce the same images because provider results are nondeterministic. This Pipe returns generated images through Sweet Tea's preview output. It does not expose palette or background-color controls: the source's color examples are disconnected from generation.