All plugins

leonardo

Leonardo.AI video generation backend for video_generate: Motion 2.0 text-to-video and image-to-video through the Leonardo REST API, using only the Python standard library. Disclosure — the prompt and any reference image URL are sent to cloud.leonardo.ai and the finished clip is fetched from cdn.leonardo.ai, authenticated with your LEONARDO_API_KEY, which the plugin reads through Hermes's secret scope and never from a file or the ambient environment; generation is billed to your Leonardo API token balance (about 200 tokens per 5-second clip, purchased separately from a Leonardo web subscription); the job is polled in-process with a bounded wall-clock deadline, and the token cost is reported back but no budget is enforced, so an unattended run can spend your balance; no telemetry, no shell commands, no background processes, no stored credentials and no self-updates. Zero Python dependencies.

RichardChua1234Toolsv1.0.1Updated 5 hours ago

About

Leonardo.AI video generation backend for video_generate: Motion 2.0 text-to-video and image-to-video through the Leonardo REST API, using only the Python standard library. Disclosure — the prompt and any reference image URL are sent to cloud.leonardo.ai and the finished clip is fetched from cdn.leonardo.ai, authenticated with your LEONARDO_API_KEY, which the plugin reads through Hermes's secret scope and never from a file or the ambient environment; generation is billed to your Leonardo API token balance (about 200 tokens per 5-second clip, purchased separately from a Leonardo web subscription); the job is polled in-process with a bounded wall-clock deadline, and the token cost is reported back but no budget is enforced, so an unattended run can spend your balance; no telemetry, no shell commands, no background processes, no stored credentials and no self-updates. Zero Python dependencies.

Leonardo AI — video generation provider

Adds Leonardo.AI as a video generation backend for Hermes.

video_generate(prompt="a lantern glowing at dusk over still water")            # text-to-video
video_generate(prompt="gentle ripples", image_url="https://…/frame.jpg")       # image-to-video

Why this exists

Hermes ships video backends for FAL, xAI and DeepInfra — and that is the entire
list, because video_gen has no hardcoded providers: every backend must arrive
as a plugin. Anything Hermes cannot reach is simply unavailable to the agent.

Leonardo is a large hosted video provider that Hermes could not reach at all. It
is also the cheapest practical door to a hosted video model, because one
Leonardo account fronts several families — Leonardo's own Motion models plus
Veo, Kling, Seedance and Wan — instead of requiring a separate account, key and
billing relationship with every vendor.

Install

hermes plugins install <this-directory-or-repo>
hermes plugins enable leonardo
hermes config set video_gen.provider leonardo

Then make sure the toolset is on:

hermes tools          # → Video Generation → Leonardo AI

Requires LEONARDO_API_KEY in $HERMES_HOME/.env.

API credits are separate from web-app tokens. A Leonardo account can show
web-subscription tokens while having zero API credits, in which case every
call fails. Generate a key under API Access → Create New Key and buy API
credits there; the /me endpoint reports the API balance as apiPaidTokens.

Verified behaviour

Tested end to end against the live API, not just type-checked:

| Check | Result |
|---|---|
| hermes plugins doctor | OK: runtime discovery, manifest parsing, import, and registration passed |
| Live text-to-video | submitted → COMPLETE → 832×480 h264 MP4, 5.03 s, 2.29 MB |
| Output verified | downloaded and probed with ffprobe — real, playable video |
| Reported cost | 0.299 USD per 480p clip |
| Failure paths | missing key and invalid key both return clean errors, no exception |

Response shapes

The create endpoint has shipped three different id shapes. Missing any one of
them throws away a generation you have already been billed for, so all are
handled:

{"generate":       {"generationId": "...", "cost": {"amount": "0.299", "unit": "DOLLARS"}}}  // v2 — observed live
{"generationId":   "..."}                                                                    // flat
{"sdGenerationJob":{"generationId": "..."}}                                                  // v1

Polling reads generations_by_pk.status and takes
generated_images[0].motionMP4URL as the result.

Capabilities

| | |
|---|---|
| Modalities | text → video, image → video |
| Aspect ratios | 16:9, 9:16, 4:5, 2:3 |
| Resolutions | 480p, 720p |
| Duration | fixed by the model (~5 s) — the Motion API accepts no duration parameter |
| Audio | no |
| Negative prompt | yes (≤1000 chars) |
| Seed | yes |

Dimensions follow Leonardo's documented table for Motion 2.0, e.g. 16:9 480p =
832×480 and 16:9 720p = 1280×720.

Limitations

  • One model family today (motion-2.0). Leonardo also fronts Veo, Kling,
    Seedance and Wan; adding them means adding verified model ids to the
    FAMILIES dict — the rest of the plugin is family-agnostic. They are not
    included on guessed ids.
  • Image-to-video passes the frame as image.type = "URL". This path is
    documented by Leonardo but was not exercised in the live test; if it rejects a
    URL, seed with an uploaded image id instead.
  • Video URLs are on Leonardo's CDN, which rejects requests without a browser
    User-Agent     (403). Hermes' own downloader handles this.
  • Costs Leonardo API credits. Nothing here meters spend for you.

Disclosure (catalog rule 13)

What a user should know before installing:

  • Network. Outbound HTTPS to cloud.leonardo.ai (API calls) and
    cdn.leonardo.ai (fetching the finished clip). No other hosts.
  • Telemetry. None. Nothing is reported to the plugin author or anyone else.
  • Credentials. Reads LEONARDO_API_KEY through
    agent.secret_scope.get_secret(), which honours the active profile. It does
    not read .env, auth.json, a vendor CLI's token file, or any other
    credential store, and it never writes or rotates a token.
  • Reads outside its own data. None. The plugin never touches Hermes core
    modules, other plugins, or user files.
  • Shell commands. None. HTTP is stdlib urllib only.
  • Background processes. None. Generation is submitted and polled in-process,
    bounded at 600 s, then the call returns.
  • Stored credentials. None written by the plugin.
  • Spend. Each generation bills Leonardo API credits. The plugin reports the
    cost the API returns but does not enforce a budget.

Files

plugin.yaml    manifest — kind: backend, requires_env: LEONARDO_API_KEY
__init__.py    LeonardoVideoGenProvider + register(ctx)

MIT.

README from the reviewed commit f8fc01d; it updates when the author re-pins.

Developer Mode

Explore built-in Hermes toolsets and actions.

HermesAgent

Developer?

Create a Hermes extension for Desktop or agent workflows.

Developer Section