> ## Documentation Index
> Fetch the complete documentation index at: https://dripart-chore-sync-comfy-api-v2-spec.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Use Sync 3 with Comfy Router

> Call synclabs/sync-3 through Comfy Router: endpoint, request shape and the response Router returns.

API Reference for `synclabs/sync-3`, served by Comfy Router from Synclabs.

## Quick start

Create a key in [your Comfy workspace](https://platform.comfy.org/profile/api-keys?onboarding=router) and export it as `COMFY_API_KEY`. The Python, TypeScript and Swift snippets use the Comfy SDKs (`pip install comfy-sdk`, `npm install @comfyorg/sdk`, and the [`ComfySwiftSDK`](https://github.com/Comfy-Org/comfy-swift-sdk) Swift package); the cURL snippet is the same call over raw HTTP.

**Model ID:** `synclabs/sync-3`

**Endpoint:** `POST https://api.comfy.org/v2/models/synclabs/sync-3`

<Tabs>
  <Tab title="Wait for the result">
    <CodeGroup>
      ```python Python theme={null}
      from comfy_sdk import Comfy

      # Reads COMFY_API_KEY from the environment.
      # The SDK automatically creates an idempotency key and reuses it for automatic retries.
      with Comfy() as client:
          result = client.models.run(
              "synclabs/sync-3",
              {
                  "input": [
                      {
                          "type": "video",
                          "url": "https://example.invalid/synclabs/sync-3/speaker.mp4",
                      },
                      {
                          "type": "audio",
                          "url": "https://example.invalid/synclabs/sync-3/voiceover.wav",
                      },
                  ],
                  "options": {
                      "sync_mode": "bounce",
                  },
              },
          )

      print(result)
      ```

      ```typescript TypeScript theme={null}
      import { comfy } from "@comfyorg/sdk";

      // Reads COMFY_API_KEY from the environment.
      // The SDK automatically creates an idempotency key and reuses it for automatic retries.
      const { data } = await comfy.models.run("synclabs/sync-3", {
        input: [
          {
            type: "video",
            url: "https://example.invalid/synclabs/sync-3/speaker.mp4",
          },
          {
            type: "audio",
            url: "https://example.invalid/synclabs/sync-3/voiceover.wav",
          },
        ],
        options: {
          sync_mode: "bounce",
        },
      });

      console.log(data);
      ```

      ```swift Swift theme={null}
      import Foundation
      import ComfySwiftSDK

      // Reads COMFY_API_KEY from the environment.
      // The SDK mints an idempotency key per call and reuses it for automatic retries.
      let client = ComfyCloudClient(apiKey: ProcessInfo.processInfo.environment["COMFY_API_KEY"]!)
      let result = try await client.models.run(
          "synclabs/sync-3",
          input: [
              "input": [
                  [
                      "type": "video",
                      "url": "https://example.invalid/synclabs/sync-3/speaker.mp4",
                  ],
                  [
                      "type": "audio",
                      "url": "https://example.invalid/synclabs/sync-3/voiceover.wav",
                  ],
              ],
              "options": [
                  "sync_mode": "bounce",
              ],
          ]
      )

      print(result.output)
      ```

      ```bash cURL theme={null}
      curl https://api.comfy.org/v2/models/synclabs/sync-3 \
        -H "X-API-Key: $COMFY_API_KEY" \
        -H "Idempotency-Key: $(uuidgen)" \
        -H "Content-Type: application/json" \
        -d "{\"input\": [{\"type\":\"video\",\"url\":\"https://example.invalid/synclabs/sync-3/speaker.mp4\"},{\"type\":\"audio\",\"url\":\"https://example.invalid/synclabs/sync-3/voiceover.wav\"}], \"options\": {\"sync_mode\":\"bounce\"}}"
      ```
    </CodeGroup>
  </Tab>

  <Tab title="Queue and collect later">
    The same body, sent to `POST https://api.comfy.org/v2/models/synclabs/sync-3/requests`. Router answers `201` with a `request_id` as soon as the run is admitted, and the result is collected once it is ready, from this process or another one. [Queued delivery](/development/comfy-router/queue) walks through status, cancellation and collection.

    <CodeGroup>
      ```python Python theme={null}
      from comfy_sdk import Comfy

      # Reads COMFY_API_KEY from the environment.
      # Each submit() call mints its own Idempotency-Key and reuses it for automatic retries.
      with Comfy() as client:
          handle = client.models.submit(
              "synclabs/sync-3",
              {
                  "input": [
                      {
                          "type": "video",
                          "url": "https://example.invalid/synclabs/sync-3/speaker.mp4",
                      },
                      {
                          "type": "audio",
                          "url": "https://example.invalid/synclabs/sync-3/voiceover.wav",
                      },
                  ],
                  "options": {
                      "sync_mode": "bounce",
                  },
              },
          )
          print("request_id:", handle.request_id)  # with the model ID, all another process needs

          # Poll until the request completes, waiting the Retry-After the server names.
          for update in handle.iter_events():
              print(update.status, update.queue_position)

          # The provider's own payload, the same value models.run() returns.
          # A request that failed or was cancelled raises the typed Router error here.
          result = handle.get()

      print(result)
      ```

      ```typescript TypeScript theme={null}
      import { comfy } from "@comfyorg/sdk";

      // Reads COMFY_API_KEY from the environment.
      // Each submit() call mints its own Idempotency-Key and reuses it for automatic retries.
      const handle = await comfy.models.submit("synclabs/sync-3", {
        input: [
          {
            type: "video",
            url: "https://example.invalid/synclabs/sync-3/speaker.mp4",
          },
          {
            type: "audio",
            url: "https://example.invalid/synclabs/sync-3/voiceover.wav",
          },
        ],
        options: {
          sync_mode: "bounce",
        },
      });
      console.log("requestId:", handle.requestId); // with the model ID, all another process needs

      // Poll until the request completes, waiting the Retry-After the server names.
      for await (const update of handle.events()) {
        console.log(update.status, update.queuePosition);
      }

      // The same result models.run() returns. A request that failed or was cancelled rejects here.
      const result = await handle.get();

      console.log(result.data);
      ```

      ```swift Swift theme={null}
      import Foundation
      import ComfySwiftSDK

      // Reads COMFY_API_KEY from the environment.
      // Each submit() call mints its own Idempotency-Key and reuses it for automatic retries.
      let client = ComfyCloudClient(apiKey: ProcessInfo.processInfo.environment["COMFY_API_KEY"]!)
      let handle = try await client.models.submit(
          "synclabs/sync-3",
          input: [
              "input": [
                  [
                      "type": "video",
                      "url": "https://example.invalid/synclabs/sync-3/speaker.mp4",
                  ],
                  [
                      "type": "audio",
                      "url": "https://example.invalid/synclabs/sync-3/voiceover.wav",
                  ],
              ],
              "options": [
                  "sync_mode": "bounce",
              ],
          ]
      )
      print("requestId:", handle.requestId)  // with the model ID, all another process needs

      // Poll until the request completes, waiting the Retry-After the server names.
      for try await update in handle.events() {
          print(update.state.rawValue, update.queuePosition.map(String.init) ?? "unknown")
      }

      // The provider's own payload, the same value models.run() returns.
      // A request that failed or was cancelled throws the typed Router error here.
      let result = try await handle.result()

      print(result.output)
      ```

      ```bash cURL theme={null}
      # 1. Submit. Router answers 201 with request_id, status_url, response_url and cancel_url.
      curl https://api.comfy.org/v2/models/synclabs/sync-3/requests \
        -H "X-API-Key: $COMFY_API_KEY" \
        -H "Idempotency-Key: $(uuidgen)" \
        -H "Content-Type: application/json" \
        -d "{\"input\": [{\"type\":\"video\",\"url\":\"https://example.invalid/synclabs/sync-3/speaker.mp4\"},{\"type\":\"audio\",\"url\":\"https://example.invalid/synclabs/sync-3/voiceover.wav\"}], \"options\": {\"sync_mode\":\"bounce\"}}"

      # 2. Poll until status is COMPLETED, waiting the Retry-After seconds each response names.
      REQUEST_ID="<request_id from the 201 body>"
      curl -i https://api.comfy.org/v2/models/synclabs/sync-3/requests/$REQUEST_ID/status \
        -H "X-API-Key: $COMFY_API_KEY"

      # 3. Collect. 200 with the model's native output, 202 with the status body while it is still running.
      curl https://api.comfy.org/v2/models/synclabs/sync-3/requests/$REQUEST_ID \
        -H "X-API-Key: $COMFY_API_KEY"
      ```
    </CodeGroup>
  </Tab>
</Tabs>

## Schema

### Input

<ParamField body="dubParams" type="object">
  Dubbing parameters attached to a Sync Labs generate request
</ParamField>

<ParamField body="dubParams.numSpeakers" type="integer">
  Number of speakers in the source video; 0 enables auto-detection
</ParamField>

<ParamField body="dubParams.providerName" type="string" required>
  Provider to use for dubbing (e.g. elevenlabs)
</ParamField>

<ParamField body="dubParams.sourceLang" type="string">
  Source language code; defaults to auto
</ParamField>

<ParamField body="dubParams.targetLang" type="string" required>
  Target language code for dubbing
</ParamField>

<ParamField body="input" type="object[]" required>
  Input items; exactly one visual input (video or image) and one audio or text input
</ParamField>

<ParamField body="input[].assetId" type="string">
  ID of an asset from the Sync Labs media library
</ParamField>

<ParamField body="input[].provider" type="object">
  Text-to-speech provider configuration for a Sync Labs text input
</ParamField>

<ParamField body="input[].provider.name" type="string" required>
  TTS provider name (e.g. elevenlabs)
</ParamField>

<ParamField body="input[].provider.script" type="string" required>
  Script to be used for generation
</ParamField>

<ParamField body="input[].provider.similarityBoost" type="number">
  How closely the AI should adhere to the original voice

  Format: `double`
</ParamField>

<ParamField body="input[].provider.stability" type="number">
  Voice stability; lower values introduce broader emotional range

  Format: `double`
</ParamField>

<ParamField body="input[].provider.voiceId" type="string" required>
  Sync voice id (cloned voice from the Studio) or ElevenLabs voice ID
</ParamField>

<ParamField body="input[].refId" type="string">
  Reference identifier used to link this input to segment definitions
</ParamField>

<ParamField body="input[].segments_frames" type="integer[][]">
  Deprecated - use the top-level segments array instead
</ParamField>

<ParamField body="input[].segments_secs" type="number[][]">
  Deprecated - use the top-level segments array instead
</ParamField>

<ParamField body="input[].type" type="string" required>
  Input type (video, image, audio, or text)
</ParamField>

<ParamField body="input[].url" type="string">
  URL of the media to be used for generation
</ParamField>

<ParamField body="model" type="string">
  Name of the model to use for generation; only sync-3 is supported. On the Comfy Router route `POST /v2/models/synclabs/{model}` this field is supplied from the path and may be omitted.
</ParamField>

<ParamField body="options" type="object">
  Additional options available for a Sync Labs generation
</ParamField>

<ParamField body="options.active_speaker_detection" type="object">
  Active speaker detection configuration
</ParamField>

<ParamField body="options.active_speaker_detection.auto_detect" type="boolean">
  Whether to automatically detect and apply generation to the active speaker
</ParamField>

<ParamField body="options.active_speaker_detection.bounding_boxes" type="integer[][]">
  Per-frame array of bounding boxes \[x1, y1, x2, y2] for the detected face
</ParamField>

<ParamField body="options.active_speaker_detection.bounding_boxes_url" type="string">
  URL to a JSON file containing bounding boxes
</ParamField>

<ParamField body="options.active_speaker_detection.coordinates" type="integer[]">
  Pixel coordinates \[x, y] in the source video frame identified by frame\_number
</ParamField>

<ParamField body="options.active_speaker_detection.frame_number" type="integer">
  Frame index that corresponds to the provided coordinates for manual speaker selection
</ParamField>

<ParamField body="options.active_speaker_detection.v3" type="boolean">
  Whether to use ASD v3
</ParamField>

<ParamField body="options.model_mode" type="string">
  Edit region for the model (lips, face, head); only works with react-1
</ParamField>

<ParamField body="options.occlusion_detection_enabled" type="boolean">
  Whether to detect occlusion during generation
</ParamField>

<ParamField body="options.prompt" type="string">
  Emotion prompt; only works with react-1
</ParamField>

<ParamField body="options.sync_mode" type="string">
  How to handle duration mismatches between video and audio (bounce, loop, cut\_off, silence, remap)
</ParamField>

<ParamField body="options.temperature" type="number">
  How expressive lipsync can be, 0 to 1

  Format: `double`
</ParamField>

<ParamField body="outputFileName" type="string">
  Base filename for the generated output without extension
</ParamField>

<ParamField body="projectId" type="string">
  Optionally attach this generation to a Sync Labs project
</ParamField>

<ParamField body="segments" type="object[]">
  Segment definitions applying different audio inputs to different video segments
</ParamField>

<ParamField body="segments[].audioInput" type="object" required>
  Audio input configuration for a specific segment
</ParamField>

<ParamField body="segments[].audioInput.endTime" type="number">
  Optional end time in seconds to crop the referenced audio

  Format: `double`
</ParamField>

<ParamField body="segments[].audioInput.refId" type="string" required>
  Reference ID of the audio/text-to-speech input to use for this segment
</ParamField>

<ParamField body="segments[].audioInput.startTime" type="number">
  Optional start time in seconds to crop the referenced audio

  Format: `double`
</ParamField>

<ParamField body="segments[].endTime" type="number" required>
  Segment end time in seconds

  Format: `double`
</ParamField>

<ParamField body="segments[].optionsOverride" type="object">
  Override generation options for a specific segment
</ParamField>

<ParamField body="segments[].optionsOverride.active_speaker_detection" type="object">
  Active speaker detection configuration
</ParamField>

<ParamField body="segments[].optionsOverride.active_speaker_detection.auto_detect" type="boolean">
  Whether to automatically detect and apply generation to the active speaker
</ParamField>

<ParamField body="segments[].optionsOverride.active_speaker_detection.bounding_boxes" type="integer[][]">
  Per-frame array of bounding boxes \[x1, y1, x2, y2] for the detected face
</ParamField>

<ParamField body="segments[].optionsOverride.active_speaker_detection.bounding_boxes_url" type="string">
  URL to a JSON file containing bounding boxes
</ParamField>

<ParamField body="segments[].optionsOverride.active_speaker_detection.coordinates" type="integer[]">
  Pixel coordinates \[x, y] in the source video frame identified by frame\_number
</ParamField>

<ParamField body="segments[].optionsOverride.active_speaker_detection.frame_number" type="integer">
  Frame index that corresponds to the provided coordinates for manual speaker selection
</ParamField>

<ParamField body="segments[].optionsOverride.active_speaker_detection.v3" type="boolean">
  Whether to use ASD v3
</ParamField>

<ParamField body="segments[].optionsOverride.occlusion_detection_enabled" type="boolean">
  Override occlusion detection for this segment
</ParamField>

<ParamField body="segments[].optionsOverride.sync_mode" type="string">
  Override the sync mode for this segment
</ParamField>

<ParamField body="segments[].optionsOverride.temperature" type="number">
  Override temperature (0-1) for this segment

  Format: `double`
</ParamField>

<ParamField body="segments[].startTime" type="number" required>
  Segment start time in seconds

  Format: `double`
</ParamField>

<ParamField body="webhookUrl" type="string">
  Webhook URL for generation status updates
</ParamField>

Generated from the schema Router serves at `GET /v2/models/synclabs/sync-3/openapi.json`, the same document it validates a call against before the request reaches the provider.

### Output

<ResponseField name="createdAt" type="string">
  The date and time the generation was created
</ResponseField>

<ResponseField name="error" type="string">
  The error message if the generation failed
</ResponseField>

<ResponseField name="errorCode" type="string">
  Stable, machine-readable error code if the generation failed
</ResponseField>

<ResponseField name="id" type="string">
  Unique identifier for the generation
</ResponseField>

<ResponseField name="input" type="object[]">
  The input items used for generation
</ResponseField>

<ResponseField name="input[].assetId" type="string">
  ID of an asset from the Sync Labs media library
</ResponseField>

<ResponseField name="input[].provider" type="object">
  Text-to-speech provider configuration for a Sync Labs text input
</ResponseField>

<ResponseField name="input[].provider.name" type="string" required>
  TTS provider name (e.g. elevenlabs)
</ResponseField>

<ResponseField name="input[].provider.script" type="string" required>
  Script to be used for generation
</ResponseField>

<ResponseField name="input[].provider.similarityBoost" type="number">
  How closely the AI should adhere to the original voice

  Format: `double`
</ResponseField>

<ResponseField name="input[].provider.stability" type="number">
  Voice stability; lower values introduce broader emotional range

  Format: `double`
</ResponseField>

<ResponseField name="input[].provider.voiceId" type="string" required>
  Sync voice id (cloned voice from the Studio) or ElevenLabs voice ID
</ResponseField>

<ResponseField name="input[].refId" type="string">
  Reference identifier used to link this input to segment definitions
</ResponseField>

<ResponseField name="input[].segments_frames" type="integer[][]">
  Deprecated - use the top-level segments array instead
</ResponseField>

<ResponseField name="input[].segments_secs" type="number[][]">
  Deprecated - use the top-level segments array instead
</ResponseField>

<ResponseField name="input[].type" type="string" required>
  Input type (video, image, audio, or text)
</ResponseField>

<ResponseField name="input[].url" type="string">
  URL of the media to be used for generation
</ResponseField>

<ResponseField name="model" type="string">
  The name of the model used for generation
</ResponseField>

<ResponseField name="options" type="object">
  Additional options available for a Sync Labs generation
</ResponseField>

<ResponseField name="options.active_speaker_detection" type="object">
  Active speaker detection configuration
</ResponseField>

<ResponseField name="options.active_speaker_detection.auto_detect" type="boolean">
  Whether to automatically detect and apply generation to the active speaker
</ResponseField>

<ResponseField name="options.active_speaker_detection.bounding_boxes" type="integer[][]">
  Per-frame array of bounding boxes \[x1, y1, x2, y2] for the detected face
</ResponseField>

<ResponseField name="options.active_speaker_detection.bounding_boxes_url" type="string">
  URL to a JSON file containing bounding boxes
</ResponseField>

<ResponseField name="options.active_speaker_detection.coordinates" type="integer[]">
  Pixel coordinates \[x, y] in the source video frame identified by frame\_number
</ResponseField>

<ResponseField name="options.active_speaker_detection.frame_number" type="integer">
  Frame index that corresponds to the provided coordinates for manual speaker selection
</ResponseField>

<ResponseField name="options.active_speaker_detection.v3" type="boolean">
  Whether to use ASD v3
</ResponseField>

<ResponseField name="options.model_mode" type="string">
  Edit region for the model (lips, face, head); only works with react-1
</ResponseField>

<ResponseField name="options.occlusion_detection_enabled" type="boolean">
  Whether to detect occlusion during generation
</ResponseField>

<ResponseField name="options.prompt" type="string">
  Emotion prompt; only works with react-1
</ResponseField>

<ResponseField name="options.sync_mode" type="string">
  How to handle duration mismatches between video and audio (bounce, loop, cut\_off, silence, remap)
</ResponseField>

<ResponseField name="options.temperature" type="number">
  How expressive lipsync can be, 0 to 1

  Format: `double`
</ResponseField>

<ResponseField name="outputDuration" type="number">
  The duration of the output media in seconds

  Format: `double`
</ResponseField>

<ResponseField name="outputFileName" type="string">
  The sanitized filename applied to the output media
</ResponseField>

<ResponseField name="outputUrl" type="string">
  The URL of the output media
</ResponseField>

<ResponseField name="projectId" type="string">
  The id of the project this generation is attached to
</ResponseField>

<ResponseField name="segmentOutputUrl" type="string">
  The URL of the segment output media
</ResponseField>

<ResponseField name="segments" type="object[]">
  The segments of the generation
</ResponseField>

<ResponseField name="segments[].audioInput" type="object" required>
  Audio input configuration for a specific segment
</ResponseField>

<ResponseField name="segments[].audioInput.endTime" type="number">
  Optional end time in seconds to crop the referenced audio

  Format: `double`
</ResponseField>

<ResponseField name="segments[].audioInput.refId" type="string" required>
  Reference ID of the audio/text-to-speech input to use for this segment
</ResponseField>

<ResponseField name="segments[].audioInput.startTime" type="number">
  Optional start time in seconds to crop the referenced audio

  Format: `double`
</ResponseField>

<ResponseField name="segments[].endTime" type="number" required>
  Segment end time in seconds

  Format: `double`
</ResponseField>

<ResponseField name="segments[].optionsOverride" type="object">
  Override generation options for a specific segment
</ResponseField>

<ResponseField name="segments[].optionsOverride.active_speaker_detection" type="object">
  Active speaker detection configuration
</ResponseField>

<ResponseField name="segments[].optionsOverride.active_speaker_detection.auto_detect" type="boolean">
  Whether to automatically detect and apply generation to the active speaker
</ResponseField>

<ResponseField name="segments[].optionsOverride.active_speaker_detection.bounding_boxes" type="integer[][]">
  Per-frame array of bounding boxes \[x1, y1, x2, y2] for the detected face
</ResponseField>

<ResponseField name="segments[].optionsOverride.active_speaker_detection.bounding_boxes_url" type="string">
  URL to a JSON file containing bounding boxes
</ResponseField>

<ResponseField name="segments[].optionsOverride.active_speaker_detection.coordinates" type="integer[]">
  Pixel coordinates \[x, y] in the source video frame identified by frame\_number
</ResponseField>

<ResponseField name="segments[].optionsOverride.active_speaker_detection.frame_number" type="integer">
  Frame index that corresponds to the provided coordinates for manual speaker selection
</ResponseField>

<ResponseField name="segments[].optionsOverride.active_speaker_detection.v3" type="boolean">
  Whether to use ASD v3
</ResponseField>

<ResponseField name="segments[].optionsOverride.occlusion_detection_enabled" type="boolean">
  Override occlusion detection for this segment
</ResponseField>

<ResponseField name="segments[].optionsOverride.sync_mode" type="string">
  Override the sync mode for this segment
</ResponseField>

<ResponseField name="segments[].optionsOverride.temperature" type="number">
  Override temperature (0-1) for this segment

  Format: `double`
</ResponseField>

<ResponseField name="segments[].startTime" type="number" required>
  Segment start time in seconds

  Format: `double`
</ResponseField>

<ResponseField name="status" type="string">
  The status of the generation (PENDING, PROCESSING, COMPLETED, FAILED, REJECTED)
</ResponseField>

<ResponseField name="synthesizedAudioUrl" type="string">
  The URL of the audio synthesized from a text (TTS) input
</ResponseField>

<ResponseField name="webhookUrl" type="string">
  The URL to the webhook endpoint
</ResponseField>

## Examples

### Input

```json theme={null}
{
  "input": [
    {
      "type": "video",
      "url": "https://example.invalid/synclabs/sync-3/speaker.mp4"
    },
    {
      "type": "audio",
      "url": "https://example.invalid/synclabs/sync-3/voiceover.wav"
    }
  ],
  "options": {
    "sync_mode": "bounce"
  }
}
```

### Output

```json theme={null}
{
  "createdAt": "2026-01-01T00:00:00.000Z",
  "id": "9a3d0c1e-0000-4000-8000-000000000000",
  "model": "sync-3",
  "outputDuration": 4.25,
  "outputFileName": "lipsync",
  "outputUrl": "https://example.invalid/synclabs/sync-3/output.mp4",
  "status": "COMPLETED"
}
```

## Before you ship

The SDKs create an `Idempotency-Key` and reuse it for automatic retries. For manual retries, reuse the original key. Router can hold the connection for up to 10 minutes.

When a request fails, Router sends an `X-Comfy-Error-Type` response header explaining why. A `422` means Router rejected the input before calling the provider, and a `413` means the request body was larger than Router accepts. Download generated assets promptly because [result URLs can expire](/development/comfy-router/reference#result-assets).

Any size limit named in a field description above is the provider's own bound on that field, quoted from the provider's specification. Router applies a separate cap to the whole request body, which base64-encoded media counts against: see [request body size](/development/comfy-router/limitations#request-bodies-are-capped).

This page documents one partner model called through Comfy Router. The same `comfy-sdk` / `@comfyorg/sdk` package also ships a second client, for running a whole ComfyUI workflow graph on Comfy Cloud: `Comfy(api_key=...)` / `new Comfy({ apiKey })`, with `client.workflows`, `client.assets` and `client.jobs`. See [Comfy SDKs](/development/api-development/sdks).

<CardGroup cols={3}>
  <Card title="Headers" icon="list" href="/development/comfy-router/headers">
    Authentication, idempotency, request IDs, error buckets, retry pacing, spend limits.
  </Card>

  <Card title="Using the Router API" icon="code" href="/development/comfy-router/api">
    Model discovery, validation errors, retries, and billing.
  </Card>

  <Card title="Limitations" icon="triangle-exclamation" href="/development/comfy-router/limitations">
    What Router does not do today, and what to use instead.
  </Card>
</CardGroup>