All plugins

gmi-video

GMI Cloud text-to-video and image-to-video as a video_gen provider (video_gen.provider: gmi-video). Reads each model's live schema, so duration, resolution, aspect ratio and audio options use the values that model accepts. Anything it lacks is dropped and reported. Disclosure — sends prompts, the optional image URLs and your GMI_API_KEY (read per request through Hermes secret scope) to console.gmicloud.ai; every generation spends GMI credits with no Hermes approval prompt (GMI shows its own price per model, and most video models bill per second). The result is downloaded from the URL GMI returns into the Hermes video cache; the model catalog is read live and cached in memory for one hour. No telemetry, shell commands, background processes, self-updates or extra Python dependencies.

zkasuranToolsv0.1.0Updated 3 hours ago

About

GMI Cloud text-to-video and image-to-video as a video_gen provider (video_gen.provider: gmi-video). Reads each model's live schema, so duration, resolution, aspect ratio and audio options use the values that model accepts. Anything it lacks is dropped and reported. Disclosure — sends prompts, the optional image URLs and your GMI_API_KEY (read per request through Hermes secret scope) to console.gmicloud.ai; every generation spends GMI credits with no Hermes approval prompt (GMI shows its own price per model, and most video models bill per second). The result is downloaded from the URL GMI returns into the Hermes video cache; the model catalog is read live and cached in memory for one hour. No telemetry, shell commands, background processes, self-updates or extra Python dependencies.

gmi-video

Creates video from text or a public image URL with Hailuo, Kling and other live GMI models.
Requires Hermes 0.21.6 or later and Python 3.11 or later.
Install with hermes plugins install gmi-video once listed in the catalog.

Setup

Set GMI_API_KEY through Hermes setup for the active profile.
Get a key at https://console.gmicloud.ai.

video_gen:
  provider: gmi-video
  model: Minimax-Hailuo-02

Example: Generate a short video of a blue boat. Use gmi-video.
An explicit model argument selects another GMI model.

Cost and data disclosure

Every generation call spends the user's GMI credits with no Hermes approval prompt.
Prices come live from GMI and are shown as supplied. No total cost is computed.
Prompts and image URLs go to console.gmicloud.ai. Results are fetched from returned URLs.
No telemetry or background processes. No user media files are read outside the Hermes cache.

Limits

Use a public image URL for image-to-video. Local files are rejected.
Reference image lists are ignored and reported.
Text and image jobs use different default models.
Duration is clamped to the selected model schema. Video may cost money per second.
Unsupported options are dropped and reported where the response supports metadata.
See the repository README for offline tests and manual live checks.

README from the reviewed commit 6760b29; it updates when the author re-pins.

More by zkasuran

3 plugins by zkasuran in the catalog.

zkasurangmi-imageGMI Cloud image generation and URL-based image editing as an image_gen provider (image_gen.provider: gmi-image). Maps Hermes aspect ratios onto each model's own schema, so seedream, gemini-image, gpt-image and luma models share one backend. Disclosure — sends prompts, the optional image URLs and your GMI_API_KEY (read per request through Hermes secret scope) to console.gmicloud.ai; every generation spends GMI credits with no Hermes approval prompt (GMI shows its own price per model). The result is downloaded from the URL GMI returns into the Hermes image cache; the model catalog is read live and cached in memory for one hour. No telemetry, shell commands, background processes, self-updates or extra Python dependencies.Communitytoolszkasurangmi-speechGMI Cloud text-to-speech as a tts provider (tts.provider: gmi-speech). Picks the voice, speed and output format parameters from the chosen model's schema and writes a file whose extension matches the real format; music models are excluded. Disclosure — sends the text to speak and your GMI_API_KEY (read per request through Hermes secret scope) to console.gmicloud.ai; every synthesis spends GMI credits with no Hermes approval prompt (GMI shows its own price per model). The audio is downloaded over HTTPS from the URL GMI returns and written to the output path Hermes supplies; the model catalog is read live and cached in memory for one hour. No telemetry, shell commands, background processes, self-updates or extra Python dependencies.Communityvoice