Directorio de recursos
Encuentra lo que necesita tu flujo.
Guides: techniques, setup, and troubleshootingBusca Pipes de Sweet Tea, flujos externos, modelos, paquetes de nodos y guías en un catálogo con evidencias.
Empieza con un flujo ↗Catálogo actual
Recursos destacados
Google Veo 3 Video
Generate native video with Google Veo 3 and transport the original bitstream through SweetTeaPreviewVideo without a duplicate ComfyUI save.
Ideogram 4 image generator
Generate images from scene prompts using the fixed Ideogram 4 conditional and unconditional model pair. Enter your completed scene description in Positive prompt; structured JSON is recommended for compositions containing text and positioned objects. No uploaded media is required. For a first run, keep the supplied monochrome knight-and-horse magazine-cover prompt, square aspect ratio, 1 megapixel output, Default sampling preset and Image count 1, then generate. The example mentions image_0f5a20.png, but there is no reference-image input and that file is not supplied to the model. Replace that mention with a written typography description when adapting the example. Technical configuration text inside the prompt is requested lettering in the image; it does not change the actual sampler, steps or seed. Output size controls both the latent canvas and scheduler dimensions. Dimensions round upward to multiples of 16, with a minimum of 256 pixels per side. Changing the canvas aspect ratio does not rewrite the prompt's JSON aspect_ratio or bounding boxes; adjust those manually to suit your composition. Optional boxes use [y1,x1,y2,x2] coordinates normalized to 0–1000. Sampling preset selects coordinated scheduler settings: Default uses 20 steps, mu 0 and std 1.75; Quality uses 48 steps, mu 0 and std 1.5; Turbo uses 12 steps, mu 0.5 and std 1.75. These names identify sampling configurations rather than guaranteed output quality. Advanced controls expose guidance, a guidance override interval, seed, sampler and image count. Main CFG defaults to 7; Override CFG defaults to 3 between progress 0.7 and 1.0. Keep Override start below Override end. Set the override equal to main CFG to remove their guidance difference. Seed -1 selects fresh noise each run; a fixed seed allows comparisons with the same noise initialization. Euler is the source sampler default. Larger image counts and resolutions require more memory and computation. The workflow uses ideogram4_int8_convrot.safetensors and ideogram4_unconditional_int8_convrot.safetensors in diffusion_models, qwen3vl_8b_fp8_scaled.safetensors in text_encoders, and flux2-vae.safetensors in vae. These fixed models must be available in the execution environment. The supplied text-encoder installation source is https://huggingface.co/Comfy-Org/Qwen3-VL/resolve/main/text_encoders/qwen3vl_8b_fp8_scaled.safetensors. There is no independent negative prompt: negative conditioning is zeroed from positive conditioning and passed to the unconditional model. The source's separate external-LLM instruction template is outside this image output. This Pipe does not call an LLM or automatically turn an idea into JSON; paste your finished prompt directly into Positive prompt.
Anima edge-guided illustration
Generate prompted Anima illustrations guided by an uploaded image’s outlines, and receive an edge-map preview alongside the generated images. For a first run, upload a reference with the composition you want and replace the example positive prompt with your desired subject, colors and style. Review the negative prompt too: its example includes sepia, which conflicts with the positive example’s sepia palette. Keep output size at 1 megapixel, control strength at 1, batch size at 1 and Turbo mode off. Seed -1 chooses fresh random noise each run; use a fixed seed when comparing settings. The reference is processed into dark edges on a light background. Its proportions determine the output proportions, subject to latent sizing. It supplies outline guidance rather than an encoded starting image, so uploaded colors, texture and transparency are not preserved. The edge-map preview is shown before resizing; the guide used for generation is resized to the requested megapixel target. Control strength adjusts outline influence; 0 removes that influence while retaining reference-derived sizing. Edge detection thresholds range from 0.01 to 0.99: keep the low threshold below the high threshold. Lower thresholds can retain more contours; higher thresholds can reduce clutter. Control start and end are fractions of sampling from 0 to 1; keep start below end to define the interval during which guidance applies. Guide interpolation controls how the edge map is resized. Turbo mode jointly selects the fixed Turbo LoRA and the separate Turbo steps and CFG settings. Standard defaults are 30 steps and CFG 4; Turbo defaults are 8 steps and CFG 1. Only the active mode’s steps and CFG affect sampling, and Turbo LoRA strength affects the selected model only when Turbo is enabled. Use positive step counts. Sampler and scheduler apply to both modes; the source presets are Euler and simple. Keep denoise at 1 initially: lowering it changes sampling from the empty latent and does not turn this into image-to-image preservation. Batch size controls the number of generated variants sharing the prompt and guide. Larger batches and output sizes increase memory requirements. The preview remains a guide preview rather than another generated variant. The fixed model setup uses anima-base-v1.0.safetensors in diffusion_models, anima-lllite-any-test-like-v2.safetensors in model_patches, qwen_3_06b_base.safetensors in text_encoders, qwen_image_vae.safetensors in vae, and anima-turbo-lora-v0.2.safetensors in loras. These files are referenced by the workflow; their availability is not established by this review. The base model, text encoder and VAE come from circlestone-labs/Anima on Hugging Face, the patch from Comfy-Org/Anima-LLLite, and the LoRA from circlestone-labs/Anima-Official-LoRAs. Generated-image transport also requires the declared tea-time-labs/sweet-tea-nodes package.
FLUX Canny image generation
The Pipe extracts a Canny edge map from your source image and uses FLUX.1 Canny dev to generate the prompted scene with that structural guidance. Output dimensions follow the source image through conditioning; there are no independent size controls. Edge thresholds tune which outlines guide generation. Guidance controls prompt adherence. Negative conditioning is zeroed internally, so this workflow has no editable negative prompt.
Wan 2.2 camera video — Standard
Generate a silent camera-conditioned video from a scene prompt and an optional starting image with Wan 2.2 14B Fun Camera. For a first run, upload a starting frame with Source image enabled, replace the rover example under Scene and action, and keep the 720 × 720 dimensions, 81 frames and 16 FPS. Choose Static or a pan, zoom or rotation under Camera movement. Leave output container and codec on auto. The default clip is about 5.1 seconds long. To generate from text without a starting image, explicitly bypass Source image; an empty upload alone is not the mode switch. Describe subject action in the prompt and choose camera motion separately. Camera speed applies to moving presets. Width and height are pixel dimensions in increments of 16; frame counts follow 1, 5, 9 and so on. Frame rate controls playback speed and duration, while frame count controls how many frames are generated. Advanced camera intrinsic parameters use the source's 0–1 scales, not millimeters. Advanced settings include unwanted-content text, seed and independent prompt guidance for initial generation and refinement. Seed -1 selects new random variation; use a fixed nonnegative seed when comparing settings. The supplied negative prompt is Chinese and excludes artifacts, poor anatomy, clutter and static imagery; adapt it to your scene. The standard profile uses a high-noise stage followed by low-noise refinement, without the sibling four-step profile's acceleration LoRAs. Both samplers use a total schedule of 20 steps: the initial stage stops at 10 and refinement starts at 10. These are independent controls, not automatically synchronized. When changing the schedule, keep both total-step settings matched, keep the handoff boundaries matched and within the total schedule, and coordinate the noise schedulers. Refinement continues the initial latent without adding new noise. Separate model sampling shifts are also available. Keep latent batch size at 1 for one clip; this graph has a single video assembly and output path and does not expose separate clip downloads for larger batches. Output settings control container, codec and encoding CRF; -1 leaves CRF automatic. Video assembly supports 8-bit and 10-bit, with possible player or encoder compatibility limits for 10-bit. No audio is connected. Uploaded transparency is not preserved through the starting-image path because only the loader's image output is used. Camera choices condition generated motion rather than guaranteeing exact physical camera tracking.
