Update to VideoGen - AI Video Generator
Snapshot Nov 18, 2026 · 06:11 UTC · version 2.6.0
Collection source: downloaded plugin package.
Instructions updated for videogen-getting-started
Skill instructions — Now says: | Tool | Inputs ...
Observed in instructions or declared skills. Runtime behavior has not been tested.
Up to 3 plugins free · no card required.
All fields and original values.
Skill instructions
Nov 5, 2026 · 06:08 UTC
| Tool | Inputs | | --- | --- | --- | | Narrated explainer, ad, or recap from **text**, especially about one minute or longer | `script_to_video` | `script`, optional `style`, `aspectRatio` `{ width, height }` | | They have a voiceover a...
Nov 18, 2026 · 06:11 UTC
| Tool | Inputs ...
Supporting files
Nov 5, 2026 · 06:08 UTC
[{"relative_path":"agents/openai.yaml","size_in_bytes":437},{"relative_path":"references/always.md","size_in_bytes":2551}]
Nov 18, 2026 · 06:11 UTC
[{"relative_path":"agents/openai.yaml","size_in_bytes":437},{"relative_path":"references/always.md","size_in_bytes":2664}]
Compare saved observations
Download comparison JSONSKILL.md text changes
− removed · + added
--- before +++ after @@ -24,20 +24,21 @@ 1. Call `get_me`. If it fails, stop and ask the user to link VideoGen in ChatGPT. 2. Tell them generation takes several minutes for a full video, then pick a workflow. If they have not named one, give short pros/cons and wait: -| Prefer when | Tool | Inputs | -| --- | --- | --- | -| Narrated explainer, ad, or recap from **text**, especially about one minute or longer | `script_to_video` | `script`, optional `style`, `aspectRatio` `{ width, height }` | -| They have a voiceover audio file | `voiceover_to_video` | `fileId` from `open_uploader` | -| They have a PDF or PowerPoint | `slideshow_to_video` | `fileId` from `open_uploader` | -| Short shot list, frame-by-frame control | `storyboard_to_video` | `scenes` with a visual `prompt` each; **at most 3 scenes** unless they explicitly ask for more | -| One short clip (up to about 30s) as a project | `prompt_to_video_clip` | `prompt` | -| Restyle an existing video (new look, same timing; up to 120s per run) | `video_to_video` | `prompt`, plus `videoFileId` from `open_uploader` or a public `videoUrl` | -| One still, clip, voice, or transform only | a `generate_*` media tool | per that tool | +| Prefer when | Tool | Inputs | +| ------------------------------------------------------------------------------------------- | ------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------- | +| Narrated explainer, ad, or recap from **text**, especially about one minute or longer | `script_to_video` | `script`, optional `style`, `aspectRatio` `{ width, height }` | +| They have a voiceover audio file | `voiceover_to_video` | `fileId` from `open_uploader` | +| They have a PDF or PowerPoint | `slideshow_to_video` | `fileId` from `open_uploader` | +| Short shot list, frame-by-frame control | `storyboard_to_video` | `scenes` with a visual `prompt` each; **at most 3 scenes** unless they explicitly ask for more | +| One short clip (up to about 30s) as a project | `prompt_to_video_clip` | `prompt` | +| Restyle an existing video (new look, same timing; up to 120s per run) | `video_to_video` | `prompt`, plus `videoFileId` from `open_uploader` or a public `videoUrl` | +| Translate an existing video's speech into other languages (one editable video per language) | `translate_video` | `videoFileId` from `open_uploader`, `outputLanguages`; export each language with `export_project` and its `editableVideoId` from the run's `outputs` | +| One still, clip, voice, or transform only | a `generate_*` media tool | per that tool | Do not silently default to `storyboard_to_video`. Do not start a long storyboard for a one-minute explainer. 3. Start the workflow. If the result is still `pending` or `running`, poll `get_workflow_run` until terminal. Keep `projectId`. -4. Optionally `remix_project` with at least two of `CAPTIONS`, `TRANSITIONS`, `ANIMATE_IMAGES`, `ZOOM`. +4. Optionally `start_video_edits` with at least two of `ENABLE_CAPTIONS`, `ADD_TRANSITIONS`, `CONVERT_IMAGES_TO_VIDEOS`, `ADD_ZOOM`. 5. Call `export_project`, poll `get_project_export` until `succeeded`, and share `downloadUrl`. ## Output @@ -49,3 +50,5 @@ - Chain several `generate_*` tools to fake a full video. - Name internal model vendors. - Start work before `get_me` succeeds. + +For `start_video_edits`, pass `projectId`, the intended `editableVideoId` from the completed workflow's `outputs`, and the requested operations as `videoEdits` objects with a `type` field. Poll the returned ids before exporting that video.
Full technical diff · 2 changed fields
changed /included_files
[
{
"relative_path": "agents/openai.yaml",
"size_in_bytes": 437
},
{
"relative_path": "references/always.md",
"size_in_bytes": 2551
}
][
{
"relative_path": "agents/openai.yaml",
"size_in_bytes": 437
},
{
"relative_path": "references/always.md",
"size_in_bytes": 2664
}
]changed /skill_md_contents
"---\nname: videogen-getting-started\ndescription: >-\n Connect a VideoGen account, pick the right workflow, and make a first video.\n Use when the user is new, asks what VideoGen can do, or has not named a\n workflow. Do not use for a specific product ad, training lesson, news recap,\n film scene, or export-only request.\n---\n\n# Getting started with VideoGen\n\nTeach ChatGPT how to go from a linked VideoGen account to a finished MP4.\n\nAlso follow `references/always.md`.\n\nIf a tool fails because of credits or account access, tell the user to open https://app.videogen.io and manage their VideoGen account, then retry.\n\n## When to use\n\nThe user is new, asks what VideoGen can do, or wants a video without naming a workflow.\n\n## Steps\n\n1. Call `get_me`. If it fails, stop and ask the user to link VideoGen in ChatGPT.\n2. Tell them generation takes several minutes for a full video, then pick a workflow. If they have not named one, give short pros/cons and wait:\n\n| Prefer when | Tool | Inputs |\n| --- | --- | --- |\n| Narrated explainer, ad, or recap from **text**, especially about one minute or longer | `script_to_video` | `script`, optional `style`, `aspectRatio` `{ width, height }` |\n| They have a voiceover audio file | `voiceover_to_video` | `fileId` from `open_uploader` |\n| They have a PDF or PowerPoint | `slideshow_to_video` | `fileId` from `open_uploader` |\n| Short shot list, frame-by-frame control | `storyboard_to_video` | `scenes` with a visual `prompt` each; **at most 3 scenes** unless they explicitly ask for more |\n| One short clip (up to about 30s) as a project | `prompt_to_video_clip` | `prompt` |\n| Restyle an existing video (new look, same timing; up to 120s per run) | `video_to_video` | `prompt`, plus `videoFileId` from `open_uploader` or a public `videoUrl` |\n| One still, clip, voice, or transform only | a `generate_*` media tool | per that tool |\n\nDo not silently default to `storyboard_to_video`. Do not start a long storyboard for a one-minute explainer.\n\n3. Start the workflow. If the result is still `pending` or `running`, poll `get_workflow_run` until terminal. Keep `projectId`.\n4. Optionally `remix_project` with at least two of `CAPTIONS`, `TRANSITIONS`, `ANIMATE_IMAGES`, `ZOOM`.\n5. Call `export_project`, poll `get_project_export` until `succeeded`, and share `downloadUrl`.\n\n## Output\n\nA real `projectId`, then an MP4 `downloadUrl`. Never invent ids.\n\n## Do not\n\n- Chain several `generate_*` tools to fake a full video.\n- Name internal model vendors.\n- Start work before `get_me` succeeds.\n""---\nname: videogen-getting-started\ndescription: >-\n Connect a VideoGen account, pick the right workflow, and make a first video.\n Use when the user is new, asks what VideoGen can do, or has not named a\n workflow. Do not use for a specific product ad, training lesson, news recap,\n film scene, or export-only request.\n---\n\n# Getting started with VideoGen\n\nTeach ChatGPT how to go from a linked VideoGen account to a finished MP4.\n\nAlso follow `references/always.md`.\n\nIf a tool fails because of credits or account access, tell the user to open https://app.videogen.io and manage their VideoGen account, then retry.\n\n## When to use\n\nThe user is new, asks what VideoGen can do, or wants a video without naming a workflow.\n\n## Steps\n\n1. Call `get_me`. If it fails, stop and ask the user to link VideoGen in ChatGPT.\n2. Tell them generation takes several minutes for a full video, then pick a workflow. If they have not named one, give short pros/cons and wait:\n\n| Prefer when | Tool | Inputs |\n| ------------------------------------------------------------------------------------------- | ------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------- |\n| Narrated explainer, ad, or recap from **text**, especially about one minute or longer | `script_to_video` | `script`, optional `style`, `aspectRatio` `{ width, height }` |\n| They have a voiceover audio file | `voiceover_to_video` | `fileId` from `open_uploader` |\n| They have a PDF or PowerPoint | `slideshow_to_video` | `fileId` from `open_uploader` |\n| Short shot list, frame-by-frame control | `storyboard_to_video` | `scenes` with a visual `prompt` each; **at most 3 scenes** unless they explicitly ask for more |\n| One short clip (up to about 30s) as a project | `prompt_to_video_clip` | `prompt` |\n| Restyle an existing video (new look, same timing; up to 120s per run) | `video_to_video` | `prompt`, plus `videoFileId` from `open_uploader` or a public `videoUrl` |\n| Translate an existing video's speech into other languages (one editable video per language) | `translate_video` | `videoFileId` from `open_uploader`, `outputLanguages`; export each language with `export_project` and its `editableVideoId` from the run's `outputs` |\n| One still, clip, voice, or transform only | a `generate_*` media tool | per that tool |\n\nDo not silently default to `storyboard_to_video`. Do not start a long storyboard for a one-minute explainer.\n\n3. Start the workflow. If the result is still `pending` or `running`, poll `get_workflow_run` until terminal. Keep `projectId`.\n4. Optionally `start_video_edits` with at least two of `ENABLE_CAPTIONS`, `ADD_TRANSITIONS`, `CONVERT_IMAGES_TO_VIDEOS`, `ADD_ZOOM`.\n5. Call `export_project`, poll `get_project_export` until `succeeded`, and share `downloadUrl`.\n\n## Output\n\nA real `projectId`, then an MP4 `downloadUrl`. Never invent ids.\n\n## Do not\n\n- Chain several `generate_*` tools to fake a full video.\n- Name internal model vendors.\n- Start work before `get_me` succeeds.\n\nFor `start_video_edits`, pass `projectId`, the intended `editableVideoId` from the completed workflow's `outputs`, and the requested operations as `videoEdits` objects with a `type` field. Poll the returned ids before exporting that video.\n"Full snapshot data
{
"description": "Connect a VideoGen account, pick the right workflow, and make a first video. Use when the user is new, asks what VideoGen can do, or has not named a workflow. Do not use for a specific product ad, training lesson, news recap, film scene, or export-only request.",
"included_files": [
{
"relative_path": "agents/openai.yaml",
"size_in_bytes": 437
},
{
"relative_path": "references/always.md",
"size_in_bytes": 2664
}
],
"name": "videogen-getting-started",
"skill_md_contents": "---\nname: videogen-getting-started\ndescription: >-\n Connect a VideoGen account, pick the right workflow, and make a first video.\n Use when the user is new, asks what VideoGen can do, or has not named a\n workflow. Do not use for a specific product ad, training lesson, news recap,\n film scene, or export-only request.\n---\n\n# Getting started with VideoGen\n\nTeach ChatGPT how to go from a linked VideoGen account to a finished MP4.\n\nAlso follow `references/always.md`.\n\nIf a tool fails because of credits or account access, tell the user to open https://app.videogen.io and manage their VideoGen account, then retry.\n\n## When to use\n\nThe user is new, asks what VideoGen can do, or wants a video without naming a workflow.\n\n## Steps\n\n1. Call `get_me`. If it fails, stop and ask the user to link VideoGen in ChatGPT.\n2. Tell them generation takes several minutes for a full video, then pick a workflow. If they have not named one, give short pros/cons and wait:\n\n| Prefer when | Tool | Inputs |\n| ------------------------------------------------------------------------------------------- | ------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------- |\n| Narrated explainer, ad, or recap from **text**, especially about one minute or longer | `script_to_video` | `script`, optional `style`, `aspectRatio` `{ width, height }` |\n| They have a voiceover audio file | `voiceover_to_video` | `fileId` from `open_uploader` |\n| They have a PDF or PowerPoint | `slideshow_to_video` | `fileId` from `open_uploader` |\n| Short shot list, frame-by-frame control | `storyboard_to_video` | `scenes` with a visual `prompt` each; **at most 3 scenes** unless they explicitly ask for more |\n| One short clip (up to about 30s) as a project | `prompt_to_video_clip` | `prompt` |\n| Restyle an existing video (new look, same timing; up to 120s per run) | `video_to_video` | `prompt`, plus `videoFileId` from `open_uploader` or a public `videoUrl` |\n| Translate an existing video's speech into other languages (one editable video per language) | `translate_video` | `videoFileId` from `open_uploader`, `outputLanguages`; export each language with `export_project` and its `editableVideoId` from the run's `outputs` |\n| One still, clip, voice, or transform only | a `generate_*` media tool | per that tool |\n\nDo not silently default to `storyboard_to_video`. Do not start a long storyboard for a one-minute explainer.\n\n3. Start the workflow. If the result is still `pending` or `running`, poll `get_workflow_run` until terminal. Keep `projectId`.\n4. Optionally `start_video_edits` with at least two of `ENABLE_CAPTIONS`, `ADD_TRANSITIONS`, `CONVERT_IMAGES_TO_VIDEOS`, `ADD_ZOOM`.\n5. Call `export_project`, poll `get_project_export` until `succeeded`, and share `downloadUrl`.\n\n## Output\n\nA real `projectId`, then an MP4 `downloadUrl`. Never invent ids.\n\n## Do not\n\n- Chain several `generate_*` tools to fake a full video.\n- Name internal model vendors.\n- Start work before `get_me` succeeds.\n\nFor `start_video_edits`, pass `projectId`, the intended `editableVideoId` from the completed workflow's `outputs`, and the requested operations as `videoEdits` objects with a `type` field. Poll the returned ids before exporting that video.\n"
}SHA-256 of public snapshot: cb4e06dc1fa454a8d6af694c8bac9df4da373083ab1ed4c986d6a8dc6b45585b