video/avatar

Avatar talking-head generation — standalone workflow.

Category: video
Source: workflows/video/avatar.py

Input Schema

Field	Type	Default	Description
`actor`	`any`	required	Reference to the actor portrait image (PNG/JPG). Face must be clearly visible.
`audio`	`any`	required	Reference to the audio clip to lip-sync (MP3/WAV). Drives the talking-head duration.
`avatar_model`	`string`	`"fal-ai/kling-video/ai-avatar/v2/standard"`	Provider/model id. Supported families: `fal-ai/kling-video/ai-avatar/`, `seedance/`, `bytedance/omnihuman*`, or any local model registered via `_LOCAL_AVATAR_MAP`.
`prompt`	`string`	`"Person speaking directly to camera, relaxed posture, natural expression, warm soft key light, shallow depth of field, authentic vlog energy."`	Stylistic prompt sent to the avatar model. Ignored by pure lipsync providers.
`regenerate`	`object`	—	When set, this run is a regeneration. Workflows may read `direction` / `keep` / `extra_instructions` to modulate prompts; the engine persists `parent_run_id` and `parent_variant_index` as run lineage columns.
`variants`	`integer`	`1`	Number of independent variant executions (1–10). When > 1, the engine runs the workflow N times with different sampling, producing N outputs.

Output Schema

Field	Type	Default	Description
`asset_id`	`object`	—	Fabric asset id, populated when the run had server access.
`avatar_model`	`string`	required	Model that produced the clip.
`kind`	`object`	—	Variant card shape: video / carousel / image / text. Surfaced on the per-variant entry of the run-output API and used by gallery UIs to pick the right layout.
`video_path`	`string`	required	Local path (or output_dir copy) of the generated mp4.

Task Pipeline

render_avatar

Task	Description
`render_avatar`	Resolve inputs → call the avatar stage → persist the result.

Run-spec example

Save the YAML below as my-run.yaml, edit the values, and run with the CLI or POST it to the API. Required fields are uncommented; optional knobs are documented above the input: block — copy any line under input: and uncomment to set.

workflow: video/avatar

# Optional fields — copy any line(s) under `input:` and uncomment to set:
# Provider/model id. Supported families: `fal-ai/kling-video/ai-avatar/*`, `seedance/*`, `bytedance/omnihuman*`, or any local model registered via `_LOCAL_AVATAR_MAP`.
#   avatar_model: fal-ai/kling-video/ai-avatar/v2/standard
#
# Stylistic prompt sent to the avatar model. Ignored by pure lipsync providers.
#   prompt: "Person speaking directly to camera, relaxed posture, natural expression, warm soft key light, shallow depth of field, authentic vlog energy."
#

input:
  # Typed reference to a file input.
  actor: {}

  # Typed reference to a file input.
  audio: {}

Run it locally:

fab-workflow --from-file my-run.yaml

Or submit over the wire — the same file is the request body:

curl -X POST 'https://gofabric.dev/v1/workflows/runs?name=video/avatar' \
  -H 'Authorization: Bearer fab_xxx' \
  -H 'content-type: application/yaml' \
  --data-binary @my-run.yaml

Every workflow also accepts the universal WorkflowInput fields — variants (1–10 fan-out) and regenerate (creative-direction hints with run lineage). See Run-specs (YAML / TOML / JSON) for the full top-level shape (metadata, priority, bundle, parent, etc.).