Model
Trends
.ai
Model
Trends
.ai
ang pinakamahuhusay na open source AI model

DualCLIPLoader node sa ComfyUI

DualCLIPLoader loads two text encoders at once, for example clip_l plus t5xxl for Flux or clip_g plus t5xxl for SD3, and outputs one CLIP object for the prompt nodes.
Node pack: ComfyUI core
Kategorya: Mga loader
Hindi pa naisasalin ang pahinang ito, kaya ipinapakita ito sa Ingles.

Ano ang ginagawa ng DualCLIPLoader

Models like Flux and SD3 read the prompt with two different encoders and merge the result. DualCLIPLoader loads both files from models/text_encoders (models/clip on older installs) and presents them as a single CLIP output, so the rest of the graph works exactly like an SD workflow: CLIP goes into CLIP Text Encode.
The type option tells ComfyUI which model family the pair belongs to, because the two encoders are combined differently for flux, sd3, hunyuan_video and hidream. The order of clip_name1 and clip_name2 does not matter.

Mga input

Pangalan
Uri
Ano ito
clip_name1
COMBO
First text encoder file, for example clip_l.safetensors.
clip_name2
COMBO
Second text encoder file, for example t5xxl_fp16.safetensors or t5xxl_fp8_e4m3fn.safetensors.
type
COMBO
sdxl, sd3, flux, hunyuan_video or hidream: the model family that will use the encoders.
device
COMBO
Optional on recent builds: default or cpu, to keep the encoders out of VRAM.

Mga output

Pangalan
Uri
Ano ito
CLIP
CLIP
The combined text encoder for CLIP Text Encode.
AD
Laktawan ang setup: mga ComfyUI workflow na naka-preinstall sa cloud
Pinapatakbo ng BitVector ang mga handang ComfyUI workflow sa sarili nitong mga GPU. Walang install, walang nawawalang node, walang pulang kahon. Buksan ito sa iyong telepono o laptop at gumawa sa loob ng isang minuto.

Paano gamitin ang DualCLIPLoader

  1. Download the two encoder files into models/text_encoders.
  2. Add DualCLIPLoader and select both files.
  3. Set type to the model family (flux for Flux Dev and Schnell).
  4. Connect CLIP to the CLIP Text Encode node that holds your prompt.

Mga setting at tip

  • t5xxl_fp8_e4m3fn saves about 5 GB of memory with almost no change in output; use fp16 only on 24 GB cards.
  • Flux ignores the negative prompt at cfg 1, so one CLIP Text Encode is enough; connect the same conditioning to both sampler inputs or use a ConditioningZeroOut for the negative.
  • For SD3.5 the three-encoder TripleCLIPLoader exists, but clip_g plus t5xxl in DualCLIPLoader also works.
  • Setting device to cpu frees VRAM at the cost of slower prompt encoding; the encoder only runs once per prompt change.

Pag-aayos ng DualCLIPLoader

The dropdown does not show my t5xxl file

Bakit ito nangyayari
The file is in the wrong folder. Text encoders are read from models/text_encoders and models/clip, not from checkpoints or diffusion_models.

Paano ayusin
Move the file to models/text_encoders and press R to refresh.

ERROR: clip input is invalid: None, or "shape mismatch" during text encoding

Bakit ito nangyayari
type does not match the files, for example type sd3 with Flux encoders, or one of the two files is a full checkpoint instead of an encoder.

Paano ayusin
Set type to the family you are running and make sure both files are real text encoders (clip_l, clip_g, t5xxl, llama for Hunyuan).

Prompt encoding takes a very long time or spills to system RAM

Bakit ito nangyayari
T5 in fp16 is about 9.5 GB. With a 12 GB card it is swapped in and out for every prompt.

Paano ayusin
Use the fp8 T5 file, or set device to cpu so the encoder runs once in system memory.

AD
Patakbuhin ang workflow na ito mula sa iyong telepono
Bawat workflow sa pahinang ito ay naka-preinstall sa BitVector kasama ang mga model at custom node. Pumili ng isa, mag-type ng prompt, tapos. Zero setup, walang ida-download.

Mga tanong tungkol sa DualCLIPLoader

Do I need DualCLIPLoader for SDXL?

Not normally. SDXL checkpoints carry both encoders inside the file and Load Checkpoint provides them. The sdxl type exists for split SDXL encoder files.

Which type do I pick for Flux Kontext or Flux Fill?

flux. They share the same clip_l and t5xxl encoders as Flux Dev.

Mga kaugnay na node

Load Diffusion Model
ComfyUI core
Load Diffusion Model (UNETLoader) loads a bare denoising network such as Flux, SD3.5, Wan or HunyuanVideo from models/diffusion_models, with a weight_dtype option for fp8 to save VRAM.
CLIP Text Encode (Prompt)
ComfyUI core
CLIP Text Encode turns a text prompt into CONDITIONING using the model text encoder. One node holds the positive prompt, a second one the negative prompt.
Flux Guidance
ComfyUI core
Flux Guidance (FluxGuidance) writes the guidance value into the conditioning that Flux Dev was distilled to expect, replacing cfg; 3.5 is the default and 2 to 5 the useful range.
Load VAE
ComfyUI core
Load VAE (VAELoader) loads a standalone VAE file from models/vae so a checkpoint with a missing or weak VAE decodes clean colours, and so split models like Flux and Wan get their decoder.
AD
Pagod na sa pag-aayos ng mga node? Hayaan ang cloud
Pinapanatili ng BitVector ang daan-daang ComfyUI workflow na naka-install, na-update at nasubok sa mabibilis na cloud GPU. Walang Python, walang CUDA error, walang limitasyon sa VRAM. Gumagana mula sa anumang browser.

Higit pang ComfyUI node

ComfyUI Manager
ComfyUI-Manager
ComfyUI Manager is the extension that installs, updates and fixes custom node packs and models from inside the interface, resolves missing nodes in imported workflows and snapshots your setup.
KSampler
ComfyUI core
KSampler runs the denoising loop: it takes the model, prompts and a latent and produces the finished latent image, controlled by seed, steps, cfg, sampler, scheduler and denoise.
Load Checkpoint
ComfyUI core
Load Checkpoint (CheckpointLoaderSimple) opens a .safetensors or .ckpt model file and hands out the three parts every workflow needs: the diffusion MODEL, the CLIP text encoder and the VAE.
Save Image
ComfyUI core
Save Image writes the IMAGE tensor to ComfyUI/output as a PNG, with the whole workflow embedded in the file metadata so the picture can be dragged back into ComfyUI to restore the graph.
VAE Decode
ComfyUI core
VAE Decode converts the sampled LATENT into a pixel IMAGE with the VAE; VAE Decode (Tiled) does the same in tiles for very large images.
Empty Latent Image
ComfyUI core
Empty Latent Image creates the blank latent canvas (width, height, batch_size) that text-to-image sampling starts from; dimensions must be multiples of 8 and match the model family.