# Dreamverse streaming demo for `fastvideo serve --config`.
#
# This public example mirrors the Dreamverse product defaults while avoiding
# private repo paths or deployment-specific comments. Hosts without
# flashinfer / NVFP4 can comment out `engine.quantization` to fall back to bf16.

generator:
  model_path: FastVideo/LTX2-Distilled-Diffusers

  engine:
    num_gpus: 1
    offload:
      dit: false
      dit_layerwise: false
      text_encoder: false
      vae: false
      pin_cpu_memory: true
    compile:
      enabled: true
      text_encoder_enabled: true
      backend: inductor
      fullgraph: true
      mode: max-autotune-no-cudagraphs
      dynamic: false
    quantization:
      transformer_quant: NVFP4

  pipeline:
    workload_type: t2v
    vae_tiling: false
    components:
      config_root: FastVideo/LTX2-Distilled-Diffusers
    preset_overrides:
      refine:
        enabled: true
        num_inference_steps: 2
        guidance_scale: 1.0
        add_noise: true

server:
  host: 0.0.0.0
  port: 8009
  output_dir: outputs/

default_request:
  sampling:
    num_frames: 121
    height: 1088
    width: 1920
    num_inference_steps: 5
    fps: 24

streaming:
  session_timeout_seconds: 300
  generation_segment_cap: 6
  stream_mode: av_fmp4
  warmup:
    enabled: true
    prompt: "A cinematic drone shot over coastal cliffs at sunrise, golden light, gentle ocean waves, ultra detailed"
    timeout_seconds: 2400
  pool:
    num_workers: 1
    enable_audio_reencode: true
    conditioning_num_frames: 9
    conditioning_end_offset: 0
  prompt:
    enabled: true
    provider: cerebras
    model: gpt-oss-120b
    timeout_ms: 20000
  safety:
    enabled: false