docs: sync built-in nodes documentation (2026-09-16 to 09-20)

built-in-nodes/MoGeInference.mdx

@@ -1,12 +1,12 @@
 ---
 title: "MoGeInference - ComfyUI Built-in Node Documentation"
-description: "Run MoGe on a single image to estimate depth and geometry."
+description: "Run MoGe on images to estimate depth and geometry."
 sidebarTitle: "MoGeInference"
 icon: "circle"
 mode: wide
 ---
 
-Run MoGe on a single image to estimate depth and geometry. This node processes an input image through the MoGe model to generate a 3D point cloud, depth map, camera intrinsics, a mask, and surface normals.
+Run MoGe on images to estimate depth and geometry. This node processes an input image through the MoGe model to generate a 3D point cloud, depth map, camera intrinsics, a mask, and surface normals.
 
 ## Inputs
 
@@ -19,6 +19,7 @@ Run MoGe on a single image to estimate depth and geometry. This node processes a
 | `batch_size` | Images per inference call. Lower if you OOM on a long video / image set. (default: 4) | INT | Yes | 1 to 64 |
 | `force_projection` | (Advanced) Forces projection of the predicted points. (default: True) | BOOLEAN | Yes | True/False |
 | `apply_mask` | (Advanced) Set masked-out (sky / invalid) pixels to inf in points and depth so meshing culls them. Disable to keep the raw predicted geometry everywhere; the mask is still returned separately. (default: True) | BOOLEAN | Yes | True/False |
+| `refine_steps` | (Advanced) MoGe-3 only: sparse volumetric refinement passes over the predicted depth. More passes sharpen fine detail and edges at a roughly linear cost. 0 disables refinement. Ignored by MoGe-1 / MoGe-2. (default: 3) | INT | Yes | 0 to 8 |
 
 Note: When the input `image` contains more frames than `batch_size`, the node processes them in multiple inference calls and combines the results into a single output geometry.
 
@@ -31,4 +32,4 @@ Note: When the input `image` contains more frames than `batch_size`, the node pr
 > This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! [Edit on GitHub](https://github.com/Comfy-Org/embedded-docs/blob/main/comfyui_embedded_docs/docs/MoGeInference/en.md)
 
 ---
-**Source fingerprint (SHA-256):** `59f6b8b1ab65147a47f5dc7ebee7b965a5ab37c6a0843a2c80d50c767ad98db4`
+**Source fingerprint (SHA-256):** `10f3399d9b6bc4ff8a940c940f538a8ec8f38a15e7d65f162499c5ab264fad65`

built-in-nodes/MoGePanoramaInference.mdx

@@ -18,6 +18,7 @@ This node performs depth estimation on equirectangular panorama images. It split
 | `split_resolution` | Resolution of each perspective split (default: 512). | INT | Yes | 256 to 1024 |
 | `merge_resolution` | Long-side resolution of the merged equirect distance map (default: 1920). | INT | Yes | 256 to 8192 |
 | `batch_size` | Views per inference batch (12 splits total) (default: 4). | INT | Yes | 1 to 12 |
+| `refine_steps` | MoGe-3 only: sparse volumetric refinement passes over the predicted depth. More passes sharpen fine detail and edges at a roughly linear cost. 0 disables refinement. Ignored by MoGe-1 / MoGe-2 (default: 3). | INT | Yes | 0 to 8 |
 
 **Notes:**
 
@@ -33,4 +34,4 @@ This node performs depth estimation on equirectangular panorama images. It split
 > This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! [Edit on GitHub](https://github.com/Comfy-Org/embedded-docs/blob/main/comfyui_embedded_docs/docs/MoGePanoramaInference/en.md)
 
 ---
-**Source fingerprint (SHA-256):** `d35b6d42a5bb17c184bc56fe3867d3a183017084dc81649c0663a9fba2362770`
+**Source fingerprint (SHA-256):** `7f21452d035b2fe9d30b0cd15ddad10916439492b1d682a8b28f1a79e375df58`

built-in-nodes/OpenAIGPTImage1.mdx

@@ -35,7 +35,6 @@ Generates images synchronously through OpenAI's GPT Image endpoint. It can creat
 - The longest edge of a custom resolution must be 3840 or less.
 - Custom resolution aspect ratio must not exceed 3:1.
 - Custom resolution total pixels must be between 655,360 and 8,294,400.
-- Transparent background is not supported for the `gpt-image-2` model.
 - The `gpt-image-1` and `gpt-image-1.5` models only support the sizes `auto`, `1024x1024`, `1024x1536`, and `1536x1024`. Other sizes are only supported by the `gpt-image-2` model.
 
 ## Outputs
@@ -47,4 +46,4 @@ Generates images synchronously through OpenAI's GPT Image endpoint. It can creat
 > This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! [Edit on GitHub](https://github.com/Comfy-Org/embedded-docs/blob/main/comfyui_embedded_docs/docs/OpenAIGPTImage1/en.md)
 
 ---
-**Source fingerprint (SHA-256):** `bf588bffced6e66536b4cb54655ef6ebb9cf988d9739e3c379a8ebda1486e20a`
+**Source fingerprint (SHA-256):** `f0f0db7fd2cdf8efd2155522b289aa8f3f939fa79a6e9060b0ffbb2dd20efa1e`

built-in-nodes/OpenAIGPTImageNodeV2.mdx

@@ -44,7 +44,7 @@ These inputs appear when `model` is set to `gpt-image-2`.
 | `model.size` | Image size. Select "Custom" to use the custom width and height (default: `"auto"`). | COMBO | Yes | `"auto"`<br />`"1024x1024"`<br />`"1024x1536"`<br />`"1536x1024"`<br />`"2048x2048"`<br />`"2048x1152"`<br />`"1152x2048"`<br />`"3840x2160"`<br />`"2160x3840"`<br />`"Custom"` |
 | `model.custom_width` | Used only when `model.size` is "Custom". Must be a multiple of 16 (default: `1024`). | INT | No | 480 to 3840 (step 16) |
 | `model.custom_height` | Used only when `model.size` is "Custom". Must be a multiple of 16 (default: `1024`). | INT | No | 480 to 3840 (step 16) |
-| `model.background` | Return image with or without background (default: `"auto"`). | COMBO | Yes | `"auto"`<br />`"opaque"` |
+| `model.background` | Return image with or without background (default: `"auto"`). | COMBO | Yes | `"auto"`<br />`"opaque"`<br />`"transparent"` |
 | `model.quality` | Image quality, affects cost and generation time (default: `"low"`). | COMBO | Yes | `"low"`<br />`"medium"`<br />`"high"` |
 | `model.images` | Optional reference image(s) for image editing. Up to 16 images. See Reference Inputs for details. | IMAGE | No | 0 to 16 |
 | `model.mask` | Optional mask for inpainting (white areas will be replaced). Requires exactly one reference image. | MASK | No | N/A |
@@ -65,7 +65,7 @@ These inputs appear when `model` is set to `gpt-image-1.5` or `gpt-image-1`. Bot
 
 | Parameter | Description | Data Type | Required | Range |
 | --- | --- | --- | --- | --- |
-| `model.images` | Growable slot: connect 1..N items (e.g. `image_1`...`image_16`); up to 16 reference images for all models. | IMAGE | No | 1 to 16 |
+| `model.images` | Growable slot: connect 1..N items (e.g. `image_1`...`image_16`); up to 16 reference images for all models. | IMAGE | No | 0 to 16 |
 | `model.mask` | Optional mask for inpainting (white areas will be replaced). Requires exactly one reference image. | MASK | No | N/A |
 
 **Parameter Constraints and Limitations:**
@@ -76,7 +76,6 @@ These inputs appear when `model` is set to `gpt-image-1.5` or `gpt-image-1`. Bot
 - When `model.images` is provided, the node runs in image editing mode; without `model.images`, it generates images from the prompt alone.
 - Reference images and the mask are downscaled before being sent to the API.
 - `"xhigh"` and `"max"` quality levels are only available for `gpt-image-2.5-flare` and `gpt-image-2.5-sunburst`.
-- The `"transparent"` background option is available for `gpt-image-2.5-flare`, `gpt-image-2.5-sunburst`, `gpt-image-1.5`, and `gpt-image-1`, but not for `gpt-image-2`.
 - `seed` is currently not implemented in the backend.
 
 ## Outputs
@@ -88,4 +87,4 @@ These inputs appear when `model` is set to `gpt-image-1.5` or `gpt-image-1`. Bot
 > This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! [Edit on GitHub](https://github.com/Comfy-Org/embedded-docs/blob/main/comfyui_embedded_docs/docs/OpenAIGPTImageNodeV2/en.md)
 
 ---
-**Source fingerprint (SHA-256):** `4f77b79f9f432a1f2e0fd814012aebe7cc42a8aa983ee9a61f3b32984bf65148`
+**Source fingerprint (SHA-256):** `804ea35d0e2aa0b2993a293cb10cb41e2f9c6a3732304306253f7f7b1eb59b8a`

built-in-nodes/PrunaImageToVideoNode.mdx

@@ -0,0 +1,53 @@
+---
+title: "PrunaImageToVideoNode - ComfyUI Built-in Node Documentation"
+description: "Animates an image into a video with Pruna's P-Video-2 model."
+sidebarTitle: "PrunaImageToVideoNode"
+icon: "circle"
+mode: wide
+---
+
+Animates an image into a video with Pruna's P-Video-2 model. The first frame is required and fixes the aspect ratio of the output; an optional last frame gives the video an end point to interpolate towards. The prompt describes how the scene moves, and the node either generates its own soundtrack or takes an audio clip that drives the motion.
+
+## Inputs
+
+### Common Inputs
+
+| Parameter | Description | Data Type | Required | Range |
+| --- | --- | --- | --- | --- |
+| `model` | Pruna video model to use. Selecting a model reveals its own inputs below. | DYNAMIC_COMBO | Yes | `"p-video-2"` |
+
+### P-Video-2 Inputs
+
+| Parameter | Description | Data Type | Required | Range |
+| --- | --- | --- | --- | --- |
+| `model.first_frame` | Image the video starts from. The output keeps the aspect ratio of this image. | IMAGE | Yes | - |
+| `model.last_frame` | Image the video ends on. Its aspect ratio must be close to the first frame's. | IMAGE | No | - |
+| `model.prompt` | Describes how the scene moves and sounds. Must contain at least one non-whitespace character, up to 5000 characters (default: empty). | STRING | Yes | Up to 5000 characters |
+| `model.duration` | Length of the video in seconds. `"auto"` lets the model choose the length from the prompt. Ignored when `model.audio` is connected: the video then follows the audio length, rounded up to a whole second, up to 20 seconds (default: `"5"`). | COMBO | Yes | `"auto"`<br />`"1"`<br />`"2"`<br />`"3"`<br />`"4"`<br />`"5"`<br />`"6"`<br />`"7"`<br />`"8"`<br />`"9"`<br />`"10"`<br />`"11"`<br />`"12"`<br />`"13"`<br />`"14"`<br />`"15"`<br />`"16"`<br />`"17"`<br />`"18"`<br />`"19"`<br />`"20"` |
+| `model.resolution` | Output resolution. 720p renders about 0.9 megapixels (1280x704 at 16:9), 1080p about 2 megapixels (1920x1088 at 16:9) (default: `"720p"`). | COMBO | Yes | `"720p"`<br />`"1080p"` |
+| `model.fps` | Frames per second. 48 fps is not available with draft at 1080p (default: `"24"`). | COMBO | Yes | `"24"`<br />`"48"` |
+| `model.draft` | Faster, less detailed render, billed at a lower rate than a standard render (default: False). | BOOLEAN | Yes | True/False |
+| `model.generate_audio` | Generate a soundtrack for the video. Ignored when `model.audio` is connected, which becomes the soundtrack instead (default: True). | BOOLEAN | Yes | True/False |
+| `model.enhance_prompt` | Rewrite the prompt with more detail before generation; short prompts need it. Turn it off to reproduce a result exactly with the same seed (default: True). Advanced setting. | BOOLEAN | Yes | True/False |
+| `model.audio` | Audio that drives the motion and becomes the soundtrack. At least 1 second long; audio longer than 20 seconds is truncated. Sets the video length instead of `model.duration`. | AUDIO | No | - |
+| `model.seed` | Seed for the generation. The same seed reproduces a result exactly only when `model.enhance_prompt` is off (default: 42). | INT | Yes | 0 to 2147483647 |
+
+**Notes:**
+
+- `model.first_frame` is required and fixes the output aspect ratio, so this node has no aspect ratio input.
+- `model.last_frame` is optional, but its aspect ratio has to be close to the first frame's or the node raises an error.
+- `model.prompt` is required and limited to 5000 characters.
+- Draft at 1080p cannot be combined with 48 fps: the node raises an error, so either turn draft off or use 24 fps.
+- Connected audio must be at least 1 second long; anything past 20 seconds is ignored.
+- With `model.audio` connected the audio sets the video length, so `model.duration` and `model.generate_audio` have no effect.
+
+## Outputs
+
+| Output Name | Description | Data Type |
+| --- | --- | --- |
+| `video` | The generated video with its soundtrack. | VIDEO |
+
+> This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! [Edit on GitHub](https://github.com/Comfy-Org/embedded-docs/blob/main/comfyui_embedded_docs/docs/PrunaImageToVideoNode/en.md)
+
+---
+**Source fingerprint (SHA-256):** `da8952e478eee593543fa7ae1aa329ad5bd5078024f905cdba185d18e76db130`

built-in-nodes/PrunaTextToVideoNode.mdx

@@ -0,0 +1,50 @@
+---
+title: "PrunaTextToVideoNode - ComfyUI Built-in Node Documentation"
+description: "Generates a video from a text prompt with Pruna's P-Video-2 model."
+sidebarTitle: "PrunaTextToVideoNode"
+icon: "circle"
+mode: wide
+---
+
+Generates a video from a text prompt with Pruna's P-Video-2 model. The prompt describes the scene, its motion and its sound, and the node either generates its own soundtrack or takes an audio clip that drives the motion and becomes the soundtrack.
+
+## Inputs
+
+### Common Inputs
+
+| Parameter | Description | Data Type | Required | Range |
+| --- | --- | --- | --- | --- |
+| `model` | Pruna video model to use. Selecting a model reveals its own inputs below. | DYNAMIC_COMBO | Yes | `"p-video-2"` |
+
+### P-Video-2 Inputs
+
+| Parameter | Description | Data Type | Required | Range |
+| --- | --- | --- | --- | --- |
+| `model.prompt` | Describes the video, its motion and its sound. Must contain at least one non-whitespace character, up to 5000 characters (default: empty). | STRING | Yes | Up to 5000 characters |
+| `model.aspect_ratio` | Aspect ratio of the output video (default: `"16:9"`). | COMBO | Yes | `"16:9"`<br />`"9:16"`<br />`"4:3"`<br />`"3:4"`<br />`"3:2"`<br />`"2:3"`<br />`"1:1"` |
+| `model.duration` | Length of the video in seconds. `"auto"` lets the model choose the length from the prompt. Ignored when `model.audio` is connected: the video then follows the audio length, rounded up to a whole second, up to 20 seconds (default: `"5"`). | COMBO | Yes | `"auto"`<br />`"1"`<br />`"2"`<br />`"3"`<br />`"4"`<br />`"5"`<br />`"6"`<br />`"7"`<br />`"8"`<br />`"9"`<br />`"10"`<br />`"11"`<br />`"12"`<br />`"13"`<br />`"14"`<br />`"15"`<br />`"16"`<br />`"17"`<br />`"18"`<br />`"19"`<br />`"20"` |
+| `model.resolution` | Output resolution. 720p renders about 0.9 megapixels (1280x704 at 16:9), 1080p about 2 megapixels (1920x1088 at 16:9) (default: `"720p"`). | COMBO | Yes | `"720p"`<br />`"1080p"` |
+| `model.fps` | Frames per second. 48 fps is not available with draft at 1080p (default: `"24"`). | COMBO | Yes | `"24"`<br />`"48"` |
+| `model.draft` | Faster, less detailed render, billed at a lower rate than a standard render (default: False). | BOOLEAN | Yes | True/False |
+| `model.generate_audio` | Generate a soundtrack for the video. Ignored when `model.audio` is connected, which becomes the soundtrack instead (default: True). | BOOLEAN | Yes | True/False |
+| `model.enhance_prompt` | Rewrite the prompt with more detail before generation; short prompts need it. Turn it off to reproduce a result exactly with the same seed (default: True). Advanced setting. | BOOLEAN | Yes | True/False |
+| `model.audio` | Audio that drives the motion and becomes the soundtrack. At least 1 second long; audio longer than 20 seconds is truncated. Sets the video length instead of `model.duration`. | AUDIO | No | - |
+| `model.seed` | Seed for the generation. The same seed reproduces a result exactly only when `model.enhance_prompt` is off (default: 42). | INT | Yes | 0 to 2147483647 |
+
+**Notes:**
+
+- `model.prompt` is required and limited to 5000 characters.
+- Draft at 1080p cannot be combined with 48 fps: the node raises an error, so either turn draft off or use 24 fps.
+- Connected audio must be at least 1 second long; anything past 20 seconds is ignored.
+- With `model.audio` connected the audio sets the video length, so `model.duration` and `model.generate_audio` have no effect.
+
+## Outputs
+
+| Output Name | Description | Data Type |
+| --- | --- | --- |
+| `video` | The generated video with its soundtrack. | VIDEO |
+
+> This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! [Edit on GitHub](https://github.com/Comfy-Org/embedded-docs/blob/main/comfyui_embedded_docs/docs/PrunaTextToVideoNode/en.md)
+
+---
+**Source fingerprint (SHA-256):** `bb7bacf2591618220c72525c1e32382174abe3a55069720adbc84fc1d8081718`

built-in-nodes/TextGenerate.mdx

@@ -23,14 +23,15 @@ The TextGenerate node uses a CLIP model to create text based on a user's prompt.
 | `max_length` | The maximum number of tokens the model will generate. The default value is 512. | INT | Yes | 1 to 32768 |
 | `thinking` | Operate in thinking mode if the model supports it. The default value is False. | BOOLEAN | No | True or False |
 | `use_default_template` | Use the built in system prompt/template if the model has one. The default value is True. This is an advanced parameter. | BOOLEAN | No | True or False |
+| `mtp` | Speculative decoding with the checkpoint's multi-token-prediction head. Has no effect without MTP weights. `"auto"` adapts the draft depth, `"2"` to `"5"` pin it. Sampled output stays correctly distributed but differs from non-MTP output for the same seed (default: `"auto"`). | COMBO | No | `"auto"`<br />`"off"`<br />`"2"`<br />`"3"`<br />`"4"`<br />`"5"` |
 
 ### Sampling Parameters (when `sampling_mode` is "on")
 
 | Parameter | Description | Data Type | Required | Range |
 | --- | --- | --- | --- | --- |
 | `temperature` | Controls the randomness of the output. Lower values make the output more predictable, higher values make it more creative. The default value is 0.7. | FLOAT | Yes | 0.01 to 2.0 |
 | `top_k` | Limits the sampling pool to the top K most likely next tokens. A value of 0 disables this filter. The default value is 64. | INT | Yes | 0 to 1000 |
-| `top_p` | Uses nucleus sampling, limiting choices to tokens whose cumulative probability is less than this value. The default value is 0.95. | FLOAT | Yes | 0.0 to 1.0 |
+| `top_p` | Uses nucleus sampling: it keeps the smallest set of most likely tokens whose cumulative probability reaches this value. The default value is 0.95. | FLOAT | Yes | 0.0 to 1.0 |
 | `min_p` | Sets a minimum probability threshold for tokens to be considered. The default value is 0.05. | FLOAT | Yes | 0.0 to 1.0 |
 | `repetition_penalty` | Penalizes tokens that have already been generated to reduce repetition. A value of 1.0 applies no penalty. The default value is 1.05. | FLOAT | Yes | 0.0 to 5.0 |
 | `seed` | A number used to initialize the random number generator for reproducible results. The default value is 0. | INT | Yes | 0 to 18446744073709551615 |
@@ -47,4 +48,4 @@ The TextGenerate node uses a CLIP model to create text based on a user's prompt.
 > This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! [Edit on GitHub](https://github.com/Comfy-Org/embedded-docs/blob/main/comfyui_embedded_docs/docs/TextGenerate/en.md)
 
 ---
-**Source fingerprint (SHA-256):** `6274a2db7c9a963304daf6df494b2b20879155e918d73429fd2ce7f3b5b9da02`
+**Source fingerprint (SHA-256):** `7d9f6aee19e076aa0afb4d57060b09474206ecf2492c0944762965fabac92c51`