TheBotique

sogni-creative-agent-skill

CLI and agent skill for personal LoRAs, image, video, and music generation using Sogni AI's decentralized GPU network. Supports Pixal3D single- and multi-view image-to-GLB, BiRefNet background removal, Qwen3-TTS speech/voice cloning/design, MiniMax Music 3, SAM 3 object selection, promptless RTX

as observed 2026-11-21T05:34:10.121Z
Identifier
sogni-creative-agent-skill
Source
ClawHub
Version observed
3.72.1
Source repository
not published
Repository observation
No source repository listed
First observed here
2026-08-28T22:54:49.073Z
Observations recorded
26
Installs (reported upstream)
2
Weekly downloads (upstream)
1,008
Declared license
MIT-0

Observation history

2026-11-21T05:34:10.121Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.71.1: Seedance videos made from a first or last frame (`--ref`, `--ref-end`) now render at the resolution you ask for. Use Sogni SDK 5.86.1 and intelligence-client 4.32.1." "3.72.1: With the content filter off (`--no-filter`), MiniMax H3 video from loose references now uses MiniMax H3 Prism (`minimax-h3-r2v-prism`, or `minimax-h3-r2v-prism-2stage` for
latestVersion "3.71.1" "3.72.1"
2026-11-20T04:34:16.795Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.71.0: Render Seedance 2.5 and Seedance 2.5 Uncensored videos in 4K with `-m seedance2-5 --target-resolution 2160` (or `-m seedance2-5-uncensored`), in text, image, reference, au "3.71.1: Seedance videos made from a first or last frame (`--ref`, `--ref-end`) now render at the resolution you ask for. Use Sogni SDK 5.86.1 and intelligence-client 4.32.1."
latestVersion "3.71.0" "3.71.1"
2026-11-19T03:34:19.174Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.70.0: Render MiniMax H3 Prism at 2K, 1080p or two-stage 768p with `-m minimax-h3-r2v-prism-2stage --no-filter`: the Prism reference-to-video request, delivered at twice the canv "3.71.0: Render Seedance 2.5 and Seedance 2.5 Uncensored videos in 4K with `-m seedance2-5 --target-resolution 2160` (or `-m seedance2-5-uncensored`), in text, image, reference, au
latestVersion "3.70.0" "3.71.0"
2026-11-18T02:34:26.003Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.69.0: Generate and edit images with Seedream 5.5 Uncensored by asking for it with `-m seedream-5-5-uncensored` or \"Seedream 5.5\"." "3.70.0: Render MiniMax H3 Prism at 2K, 1080p or two-stage 768p with `-m minimax-h3-r2v-prism-2stage --no-filter`: the Prism reference-to-video request, delivered at twice the canv
latestVersion "3.69.0" "3.70.0"
2026-11-17T01:34:38.249Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.68.1: Use Sogni SDK 5.84.1, intelligence-client 4.30.1 and the synchronized creative-agent 2.34.1 runtime." "3.69.0: Generate and edit images with Seedream 5.5 Uncensored by asking for it with `-m seedream-5-5-uncensored` or \"Seedream 5.5\"."
latestVersion "3.68.1" "3.69.0"
2026-11-16T00:36:19.090Z

Fields that differed: changelog latestVersion summary

FieldBeforeAfter
changelog "3.66.1: Document the SAM 3.1 video selection options for SDK callers: a drawn region, 2 to 16 objects tracked together, the same selection on several frames, and how tracking carr "3.68.1: Use Sogni SDK 5.84.1, intelligence-client 4.30.1 and the synchronized creative-agent 2.34.1 runtime."
latestVersion "3.66.1" "3.68.1"
summary "CLI and agent skill for personal LoRAs, image, video, and music generation using Sogni AI's decentralized GPU network. Supports Pixal3D single- and multi-view image-to-GLB, BiRefN "CLI and agent skill for personal LoRAs, image, video, and music generation using Sogni AI's decentralized GPU network. Supports Pixal3D single- and multi-view image-to-GLB, BiRefN
2026-11-10T19:47:06.599Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.65.1: Improve optional inline image previews while keeping full-resolution result links available. Refresh agent instructions and Claude plugin presentation, and clarify how net "3.66.1: Document the SAM 3.1 video selection options for SDK callers: a drawn region, 2 to 16 objects tracked together, the same selection on several frames, and how tracking carr
latestVersion "3.65.1" "3.66.1"
2026-11-06T15:44:01.797Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Seedance 2.5 draft mode: iterate on 480p drafts with --seedance-draft and render the kept one at 1080p with --seedance-finalize. New prompt-guide command prints each model's promp "3.65.1: Improve optional inline image previews while keeping full-resolution result links available. Refresh agent instructions and Claude plugin presentation, and clarify how net
latestVersion "3.63.0" "3.65.1"
2026-11-04T13:43:41.377Z

Fields that differed: changelog latestVersion summary

FieldBeforeAfter
changelog "3.60.1: Prefer FastH3 Omni for MiniMax reference-to-video, including the two-stage option for 1080p and 2K. 3.60.0: Create images and edit from up to ten references with Seedream "Seedance 2.5 draft mode: iterate on 480p drafts with --seedance-draft and render the kept one at 1080p with --seedance-finalize. New prompt-guide command prints each model's promp
latestVersion "3.60.1" "3.63.0"
summary "CLI and agent skill for personal LoRAs, image, video, and music generation using Sogni AI's decentralized GPU network. Supports Pixal3D single- and multi-view image-to-GLB, BiRefN "CLI and agent skill for personal LoRAs, image, video, and music generation using Sogni AI's decentralized GPU network. Supports Pixal3D single- and multi-view image-to-GLB, BiRefN
2026-10-23T01:44:40.541Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.58.2: FLUX 3 Image Klein edits now take up to ten references and GPT Image edits up to 16." "3.60.1: Prefer FastH3 Omni for MiniMax reference-to-video, including the two-stage option for 1080p and 2K. 3.60.0: Create images and edit from up to ten references with Seedream
latestVersion "3.58.2" "3.60.1"
2026-10-19T22:50:38.661Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.58.0: Generate uncensored images from text with MOSS-Image and MOSS-Image 8-Step, including compatible Personal LoRAs. Use the published MOSS contracts in intelligence client 4. "3.58.2: FLUX 3 Image Klein edits now take up to ten references and GPT Image edits up to 16."
latestVersion "3.58.0" "3.58.2"
2026-10-18T21:45:32.509Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.57.0: Add FLUX 3 Image Klein (`-m flux3_image_klein_sd_int8_convrot`; hosted selector `flux-3-klein`), Black Forest Labs' fast open image model: text-to-image, image-to-image, a "3.58.0: Generate uncensored images from text with MOSS-Image and MOSS-Image 8-Step, including compatible Personal LoRAs. Use the published MOSS contracts in intelligence client 4.
latestVersion "3.57.0" "3.58.0"
2026-10-17T20:48:50.018Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.56.2: Update the Sogni SDK (`@sogni-ai/sogni-client`) to 5.63.3 and the intelligence client to 4.12.4 (creative agent 2.12.8). Update sharp to 0.35.5, which carries the librsvg "3.57.0: Add FLUX 3 Image Klein (`-m flux3_image_klein_sd_int8_convrot`; hosted selector `flux-3-klein`), Black Forest Labs' fast open image model: text-to-image, image-to-image, a
latestVersion "3.56.2" "3.57.0"
2026-10-15T18:48:12.129Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.56.1: `--upscale-video` rejected every video with \"The source video dimensions, frame count, or frame rate could not be read\" whenever it could not start ffprobe, even when ff "3.56.2: Update the Sogni SDK (`@sogni-ai/sogni-client`) to 5.63.3 and the intelligence client to 4.12.4 (creative agent 2.12.8). Update sharp to 0.35.5, which carries the librsvg
latestVersion "3.56.1" "3.56.2"
2026-10-10T14:15:52.413Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.56.0: Seedance 2.0 Mini Uncensored: `-m seedance2-mini-uncensored` (and `-m seedance2-mini-uncensored-t2v`, or names such as \"Seedance 2.0 Mini Uncensored\", \"Seedance Mini Un "3.56.1: `--upscale-video` rejected every video with \"The source video dimensions, frame count, or frame rate could not be read\" whenever it could not start ffprobe, even when ff
latestVersion "3.56.0" "3.56.1"
2026-10-09T13:20:37.280Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.55.0: Seedance 2.5 Uncensored: `-m seedance2-5-uncensored` (with `-t2v`, `-ia2v` and `-v2v` forms, or the names \"Seedance 2.5 Uncensored\", \"Seedance Uncensored\", \"Seedance "3.56.0: Seedance 2.0 Mini Uncensored: `-m seedance2-mini-uncensored` (and `-m seedance2-mini-uncensored-t2v`, or names such as \"Seedance 2.0 Mini Uncensored\", \"Seedance Mini Un
latestVersion "3.55.0" "3.56.0"
2026-10-07T11:52:53.866Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.54.5: A reconnect to Sogni that took longer than 15 seconds could end the CLI with an error (exit 1) while it was still waiting for a render: the SDK's keep-alive pinged the con "3.55.0: Seedance 2.5 Uncensored: `-m seedance2-5-uncensored` (with `-t2v`, `-ia2v` and `-v2v` forms, or the names \"Seedance 2.5 Uncensored\", \"Seedance Uncensored\", \"Seedance
latestVersion "3.54.5" "3.55.0"
2026-10-05T10:33:23.015Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.54.4: `--music` without `-m` now runs MiniMax Music 3 (60 s by default, 10 to 300 s), like every other Sogni surface; ACE-Step 1.5 XL Turbo stays available by name. `--status <p "3.54.5: A reconnect to Sogni that took longer than 15 seconds could end the CLI with an error (exit 1) while it was still waiting for a render: the SDK's keep-alive pinged the con
latestVersion "3.54.4" "3.54.5"
2026-10-03T08:46:10.142Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "3.54.3: MiniMax H3 keyframes are now named in the prompt, the way MiniMax's own keyframe workflow does it. Update the intelligence client to 4.6.3 and the bundled runtime to creat "3.54.4: `--music` without `-m` now runs MiniMax Music 3 (60 s by default, 10 to 300 s), like every other Sogni surface; ACE-Step 1.5 XL Turbo stays available by name. `--status <p
latestVersion "3.54.3" "3.54.4"
2026-09-28T03:46:57.521Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Update session handling dependencies, worker image exports, prompt metadata opt-out, and per-output JSON labels." "3.54.3: MiniMax H3 keyframes are now named in the prompt, the way MiniMax's own keyframe workflow does it. Update the intelligence client to 4.6.3 and the bundled runtime to creat
latestVersion "3.52.2" "3.54.3"
2026-09-24T00:36:14.063Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Adds MiniMax H3 Two-Stage Reference to Video (minimax-h3-r2v-2stage, minimax-h3-r2v-balanced-2stage) delivering 768p, 1080p or 2K; updates to Sogni SDK 5.53.1, which resends a gen "Update session handling dependencies, worker image exports, prompt metadata opt-out, and per-output JSON labels."
latestVersion "3.52.0" "3.52.2"
2026-09-19T21:21:32.220Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Personal LoRA library, hosted image utilities, multiview 3D and speech controls; updated SDK and agent runtime contracts." "Adds MiniMax H3 Two-Stage Reference to Video (minimax-h3-r2v-2stage, minimax-h3-r2v-balanced-2stage) delivering 768p, 1080p or 2K; updates to Sogni SDK 5.53.1, which resends a gen
latestVersion "3.51.1" "3.52.0"
2026-09-18T20:20:38.905Z

Fields that differed: changelog latestVersion summary topics

FieldBeforeAfter
changelog "Current image, video and music guidance; FlashVSR video upscaling with server-owned clip limits; Goose skill launcher and packaged MCP command; repaired standalone desktop bundle. "Personal LoRA library, hosted image utilities, multiview 3D and speech controls; updated SDK and agent runtime contracts."
latestVersion "3.42.0" "3.51.1"
summary "Sogni Creative Agent Skill: agent skill and CLI for image, video, and music generation using Sogni AI's decentralized GPU network. Supports promptless RTX VSR image upscaling thro "CLI and agent skill for personal LoRAs, image, video, and music generation using Sogni AI's decentralized GPU network. Supports Pixal3D single- and multi-view image-to-GLB, BiRefN
topics ["Music Generation"] []
2026-09-17T19:21:46.965Z

Fields that differed: license description

FieldBeforeAfter
license null "MIT-0"
description "Sogni Creative Agent Skill: agent skill and CLI for image, video, and music generation using Sogni AI's decentralized GPU network. Supports promptless RTX VSR image upscaling thro null
2026-09-11T13:17:26.274Z

Fields that differed: changelog description displayName latestVersion summary

FieldBeforeAfter
changelog "MiniMax H3 now reports the duration it actually delivers (124 + n*17 frames at 24fps = 5.17-15.08s) instead of an unreachable 5-15s range. SKILL.md, README.md, and llm.txt now doc "Current image, video and music guidance; FlashVSR video upscaling with server-owned clip limits; Goose skill launcher and packaged MCP command; repaired standalone desktop bundle.
description "Sogni Creative Agent Skill: agent skill and CLI for image, video, and music generation using Sogni AI's decentralized GPU network. Supports one-click image-folder loop reels, pers "Sogni Creative Agent Skill: agent skill and CLI for image, video, and music generation using Sogni AI's decentralized GPU network. Supports promptless RTX VSR image upscaling thro
displayName "Clawhub 3.26.1" "sogni-creative-agent-skill"
latestVersion "3.26.1" "3.42.0"
summary "Sogni Creative Agent Skill: agent skill and CLI for image, video, and music generation using Sogni AI's decentralized GPU network. Supports one-click image-folder loop reels, pers "Sogni Creative Agent Skill: agent skill and CLI for image, video, and music generation using Sogni AI's decentralized GPU network. Supports promptless RTX VSR image upscaling thro

Correction

If you maintain this extension and believe anything above is inaccurate, request a correction. Corrections are published, and disputed entries are marked as disputed while under review.