everypixel-cli 0.2.0__tar.gz → 0.2.1__tar.gz

{everypixel_cli-0.2.0 → everypixel_cli-0.2.1}/CHANGELOG.md

@@ -1,5 +1,9 @@
 # Changelog
 
+## [0.2.1] - 2026-09-11
+
+- Support up to 9 image, 3 video, and 3 audio references for MiniMax H3/Turbo.
+
 ## [0.2.0] - 2026-09-09
 
 - Add MiniMax H3/Turbo, WAN 3.0, Flux 3, and Seedance 2.5 video models.

{everypixel_cli-0.2.0 → everypixel_cli-0.2.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: everypixel-cli
-Version: 0.2.0
+Version: 0.2.1
 Summary: Official command line interface for Everypixel API
 Project-URL: Homepage, https://labs.everypixel.com
 Project-URL: Repository, https://github.com/everypixel/everypixel-cli
@@ -230,20 +230,22 @@ New video models use these provider defaults and limits:
 
 | Model | Duration | Default resolution / aspect ratio | Inputs |
 | --- | --- | --- | --- |
-| `minimax-h3-turbo`, `minimax-h3` | 3–15 s | 768p / 7:4 | Text, first/last frames, or one `--reference-image` |
+| `minimax-h3-turbo`, `minimax-h3` | 3–15 s | 768p / 7:4 | Text, first/last frames, or up to 9 image, 3 video, and 3 audio references |
 | `wan3.0` | 2–30 s | 1080p / adaptive | Text, frames, or up to 10 reference images and 5 reference videos |
 | `flux3` | 5–20 s generation; 5–15 s continuation | 720p / 16:9 | Text, frames, or `video edit --video` for continuation |
 | `seedance2.5` | 4–30 s | 720p / 16:9 | Text or `video edit` references, including audio alone |
 
-MiniMax reference images cannot be combined with first/last frames. WAN 3.0
-also separates frame and reference inputs. Seedance 2.5 supports up to 1080p;
+Use repeatable `--reference-image`, `--reference-video`, and `--reference-audio`
+options for MiniMax references. References cannot be combined with first/last
+frames. WAN 3.0 also separates frame and reference inputs. Seedance 2.5 supports
+up to 1080p;
 Seedance 2 and Mini retain their 15-second limit. Grok video model names are
 `grok-imagine` and `grok-imagine-1.5` (image required); `grok` and `grok15`
 remain accepted aliases. The CLI's default video model remains `ltx23`.
 
 Topaz upscaling models are `topaz-prob-4`, `topaz-slp-2.5`, and `topaz-ast-2`.
 They support 4K output. The default `seedvr2` supports up to 1440p, with a
-20-second source limit at 1440p.
+20-second source limit at 1080p and a 10-second limit at 1440p, inclusive.
 
 ### Lipsync and Audio
 

{everypixel_cli-0.2.0 → everypixel_cli-0.2.1}/README.md

@@ -196,20 +196,22 @@ New video models use these provider defaults and limits:
 
 | Model | Duration | Default resolution / aspect ratio | Inputs |
 | --- | --- | --- | --- |
-| `minimax-h3-turbo`, `minimax-h3` | 3–15 s | 768p / 7:4 | Text, first/last frames, or one `--reference-image` |
+| `minimax-h3-turbo`, `minimax-h3` | 3–15 s | 768p / 7:4 | Text, first/last frames, or up to 9 image, 3 video, and 3 audio references |
 | `wan3.0` | 2–30 s | 1080p / adaptive | Text, frames, or up to 10 reference images and 5 reference videos |
 | `flux3` | 5–20 s generation; 5–15 s continuation | 720p / 16:9 | Text, frames, or `video edit --video` for continuation |
 | `seedance2.5` | 4–30 s | 720p / 16:9 | Text or `video edit` references, including audio alone |
 
-MiniMax reference images cannot be combined with first/last frames. WAN 3.0
-also separates frame and reference inputs. Seedance 2.5 supports up to 1080p;
+Use repeatable `--reference-image`, `--reference-video`, and `--reference-audio`
+options for MiniMax references. References cannot be combined with first/last
+frames. WAN 3.0 also separates frame and reference inputs. Seedance 2.5 supports
+up to 1080p;
 Seedance 2 and Mini retain their 15-second limit. Grok video model names are
 `grok-imagine` and `grok-imagine-1.5` (image required); `grok` and `grok15`
 remain accepted aliases. The CLI's default video model remains `ltx23`.
 
 Topaz upscaling models are `topaz-prob-4`, `topaz-slp-2.5`, and `topaz-ast-2`.
 They support 4K output. The default `seedvr2` supports up to 1440p, with a
-20-second source limit at 1440p.
+20-second source limit at 1080p and a 10-second limit at 1440p, inclusive.
 
 ### Lipsync and Audio
 

everypixel_cli-0.2.1/src/everypixel_cli/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.2.1"

{everypixel_cli-0.2.0 → everypixel_cli-0.2.1}/src/everypixel_cli/application/services.py

@@ -651,6 +651,7 @@ def build_video_generate_payload(**values: Any) -> dict[str, Any]:
     last_image = values.pop("last_image", None)
     reference_images = values.pop("reference_images", [])
     reference_videos = values.pop("reference_videos", [])
+    reference_audios = values.pop("reference_audios", [])
     values = normalize_grok_model(values)
     if values.get("duration") is None:
         model = values.get("model")
@@ -677,6 +678,10 @@ def build_video_generate_payload(**values: Any) -> dict[str, Any]:
         payload["reference_video_urls"] = [
             media_value(value) for value in reference_videos
         ]
+    if reference_audios:
+        payload["reference_audio_urls"] = [
+            media_value(value) for value in reference_audios
+        ]
     validated = _VIDEO_GENERATE_ADAPTER.validate_python(
         {key: value for key, value in payload.items() if value is not None}
     )
@@ -971,6 +976,7 @@ class ApplicationServices:
         aspect_ratio: str | None,
         reference_images: list[str] | None = None,
         reference_videos: list[str] | None = None,
+        reference_audios: list[str] | None = None,
         image: str | None = None,
         last_image: str | None = None,
         seed: int | None = None,
@@ -986,6 +992,7 @@ class ApplicationServices:
             aspect_ratio=aspect_ratio,
             reference_images=reference_images or [],
             reference_videos=reference_videos or [],
+            reference_audios=reference_audios or [],
             image=image,
             last_image=last_image,
             seed=seed,
@@ -1427,7 +1434,7 @@ def normalize_video_edit_content(payload: dict[str, Any]) -> None:
         value = payload.get(key)
         if isinstance(value, str):
             payload[key] = media_value(value)
-    for key in ("reference_image_urls", "reference_video_urls"):
+    for key in ("reference_image_urls", "reference_video_urls", "reference_audio_urls"):
         values = payload.get(key)
         if isinstance(values, list):
             payload[key] = [

{everypixel_cli-0.2.0 → everypixel_cli-0.2.1}/src/everypixel_cli/cli.py

@@ -807,7 +807,14 @@ def video_generate(
         list[str],
         typer.Option(
             "--reference-video",
-            help="Wan reference video URL or local path. Can be repeated.",
+            help="MiniMax or Wan reference video URL or local path. Can be repeated.",
+        ),
+    ] = [],
+    reference_audio: Annotated[
+        list[str],
+        typer.Option(
+            "--reference-audio",
+            help="MiniMax reference audio URL or local path. Can be repeated.",
         ),
     ] = [],
     seed: Optional[int] = None,
@@ -839,6 +846,7 @@ def video_generate(
             aspect_ratio=aspect_ratio,
             reference_images=reference_image,
             reference_videos=reference_video,
+            reference_audios=reference_audio,
             seed=seed,
             generate_audio=generate_audio,
             callback_url=callback_url,
@@ -1025,7 +1033,8 @@ def video_upscale(
         typer.Option(
             "--resolution",
             help=(
-                "720p, 1080p, 1440p, or 4k (Topaz only). SeedVR2: up to 20s at 1440p."
+                "720p, 1080p, 1440p, or 4k (Topaz only). "
+                "SeedVR2: up to 20s at 1080p, 10s at 1440p."
             ),
         ),
     ] = "1080p",
@@ -1036,7 +1045,8 @@ def video_upscale(
 ) -> None:
     """Start video upscale from a URL/file or task_id.
 
-    SeedVR2 source videos longer than 20 seconds cannot be upscaled to 1440p.
+    SeedVR2 accepts source videos up to 20 seconds at 1080p or 10 seconds at
+    1440p, inclusive.
     """
 
     runtime = apply_common_options(

{everypixel_cli-0.2.0 → everypixel_cli-0.2.1}/src/everypixel_cli/mcp_server.py

@@ -225,6 +225,7 @@ class VideoGenerateInput(MCPToolInput):
     aspect_ratio: VideoGenerateAspectRatio | None = None
     reference_images: list[str] | None = None
     reference_videos: list[str] | None = None
+    reference_audios: list[str] | None = None
     image: str | None = None
     last_image: str | None = None
     seed: int | None = None
@@ -576,6 +577,7 @@ def create_mcp_server(
             aspect_ratio=arguments.aspect_ratio,
             reference_images=arguments.reference_images,
             reference_videos=arguments.reference_videos,
+            reference_audios=arguments.reference_audios,
             image=arguments.image,
             last_image=arguments.last_image,
             seed=arguments.seed,

{everypixel_cli-0.2.0 → everypixel_cli-0.2.1}/src/everypixel_cli/resources/openapi.json

@@ -2073,14 +2073,30 @@
             "title": "Prompt",
             "type": "string"
           },
+          "reference_audio_urls": {
+            "items": {
+              "type": "string"
+            },
+            "maxItems": 3,
+            "title": "Reference Audio Urls",
+            "type": "array"
+          },
           "reference_image_urls": {
             "items": {
               "type": "string"
             },
-            "maxItems": 1,
+            "maxItems": 9,
             "title": "Reference Image Urls",
             "type": "array"
           },
+          "reference_video_urls": {
+            "items": {
+              "type": "string"
+            },
+            "maxItems": 3,
+            "title": "Reference Video Urls",
+            "type": "array"
+          },
           "resolution": {
             "const": "768p",
             "default": "768p",
@@ -4852,6 +4868,62 @@
         ]
       }
     },
+    "/v1/tasks/{task_id}": {
+      "delete": {
+        "description": "Cancel an owned task while cancellation is still available. Repeated cancellation is idempotent.",
+        "operationId": "delete_task_v1_tasks__task_id__delete",
+        "parameters": [
+          {
+            "in": "path",
+            "name": "task_id",
+            "required": true,
+            "schema": {
+              "format": "uuid",
+              "title": "Task Id",
+              "type": "string"
+            }
+          }
+        ],
+        "responses": {
+          "200": {
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ImgenStatusResponse"
+                }
+              }
+            },
+            "description": "Successful Response"
+          },
+          "401": {
+            "description": "Authentication required"
+          },
+          "404": {
+            "description": "Task not found for this client"
+          },
+          "409": {
+            "description": "Task cannot be cancelled in its current state"
+          },
+          "422": {
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/HTTPValidationError"
+                }
+              }
+            },
+            "description": "Validation Error"
+          },
+          "503": {
+            "description": "Cancellation temporarily unavailable; retry the request"
+          }
+        },
+        "summary": "Cancel a queued task",
+        "tags": [
+          "imgen"
+        ]
+      }
+    },
     "/v1/transcribe": {
       "post": {
         "operationId": "transcribe_v1_transcribe_post",

{everypixel_cli-0.2.0 → everypixel_cli-0.2.1}/src/everypixel_cli/schemas.py

@@ -6,6 +6,7 @@ media sources, model limits, enum values, and numeric ranges.
 
 from __future__ import annotations
 
+import re
 from typing import Annotated, Any, Literal
 
 from pydantic import (
@@ -424,7 +425,9 @@ class MiniMaxH3VideoGenRequest(_StrictVideoPayload):
     seed: int = Field(default=-1, ge=-1, le=18_446_744_073_709_551_615)
     image_url: str | None = None
     image_last_url: str | None = None
-    reference_image_urls: list[str] = Field(default_factory=list, max_length=1)
+    reference_image_urls: list[str] = Field(default_factory=list, max_length=9)
+    reference_video_urls: list[str] = Field(default_factory=list, max_length=3)
+    reference_audio_urls: list[str] = Field(default_factory=list, max_length=3)
     callback_url: str | None = None
 
     @model_validator(mode="after")
@@ -433,8 +436,29 @@ class MiniMaxH3VideoGenRequest(_StrictVideoPayload):
             raise ValueError("prompt must not be blank")
         validate_frame_inputs(self.image_url, self.image_last_url)
         validate_reference_inputs(self.reference_image_urls, [])
-        if self.reference_image_urls and (self.image_url or self.image_last_url):
-            raise ValueError("Frame images and reference_image_urls cannot be combined")
+        for media_type, values in (
+            ("video", self.reference_video_urls),
+            ("audio", self.reference_audio_urls),
+        ):
+            for value in values:
+                if value.startswith("data:"):
+                    if not re.fullmatch(
+                        rf"data:{media_type}/[a-zA-Z0-9.+\-]+;base64,[A-Za-z0-9+/]+=*",
+                        value,
+                    ):
+                        raise ValueError(
+                            f"reference_{media_type}_urls requires {media_type} "
+                            "base64 data URIs"
+                        )
+                else:
+                    HttpUrl(value)
+        references = (
+            self.reference_image_urls
+            or self.reference_video_urls
+            or self.reference_audio_urls
+        )
+        if references and (self.image_url or self.image_last_url):
+            raise ValueError("Frame images and reference media cannot be combined")
         return self
 
 
@@ -975,7 +999,10 @@ class VideoUpscalePayload(BaseModel):
     callback_url: str | None = None
     resolution: VideoUpscaleResolution = Field(
         default="1080p",
-        description="Output resolution; SeedVR2 accepts up to 20 seconds at 1440p.",
+        description=(
+            "Output resolution; SeedVR2 source limits: "
+            "20s at 1080p, 10s at 1440p, inclusive."
+        ),
     )
 
     @model_validator(mode="after")

everypixel_cli-0.2.0/src/everypixel_cli/__init__.py

@@ -1 +0,0 @@
-__version__ = "0.2.0"