TextEncodeFibo2 - ComfyUI Built-in Node Documentation

Encode a Fibo 2.0 prompt into conditioning and a matching empty latent. The prompt is either plain text or a structured JSON caption; with images connected it becomes the edit instruction, or a JSON caption of the result with an edit_instruction key, and the images act as reference latents for both outputs when a VAE is connected. Sample the latent output, whose size is the size the images are cropped to.

Inputs

Parameter	Description	Data Type	Required	Range
`clip`	The Fibo 2.0 text encoder used to tokenize and encode the prompt.	CLIP	Yes	-
`prompt`	The caption to generate from, or the edit instruction when images are connected. Refer to the images as `<image_1>`, `<image_2>` and so on. A JSON caption of the expected result works too, with the change described in its `edit_instruction` key.	STRING	Yes	-
`negative_prompt`	Caption to steer away from, encoded as the negative conditioning. A non-empty value is sent inside Fibo 2.0’s chat template as written, not cleaned like a caption. Empty is the null caption Fibo 2.0 was trained with for classifier-free guidance in edit workflows (default: empty).	STRING	Yes	-
`vae`	VAE used to encode the images into reference latents. Without it the images only condition the text encoding.	VAE	No	-
`images`	Images to edit, up to 16. A single image is scaled and center cropped to the result’s size, because the edit lines up with it pixel for pixel; several images are each scaled to their own trained size.	IMAGE	No	0 to 16 images
`width`	Width of the result, rounded down to a multiple of 16. 0 uses the size Fibo 2.0 was trained at for the first image’s aspect ratio, about 1 megapixel, or 1024 without images (default: 0).	INT	Yes	0 to 16384 (step 16)
`height`	Height of the result, rounded down to a multiple of 16. 0 uses the size Fibo 2.0 was trained at for the first image’s aspect ratio, about 1 megapixel, or 1024 without images (default: 0).	INT	Yes	0 to 16384 (step 16)
`batch_size`	Number of latents to create in the batch (default: 1).	INT	Yes	1 to 4096

The reference latents are attached to both outputs, so the images travel with the conditioning rather than only with the positive branch. When the VAE is not connected, the images are only seen by the text encoder and no reference latents are added.

Outputs

Output Name	Description	Data Type
`positive`	Conditioning of the prompt, with the reference latents when a VAE and images are connected.	CONDITIONING
`negative`	Conditioning of `negative_prompt`, text only, with the same reference latents.	CONDITIONING
`latent`	Empty latent at the requested size, ready to be sampled.	LATENT

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

Source fingerprint (SHA-256): 55519ccce6420289b01b5855b80f522057da78f5a17bf7db63e2418d15f9e356

​Inputs

​Outputs

Inputs

Outputs