"""FAL.ai video generation backend. User-facing surface: pick a **model family** (e.g. "Pixverse v6", "Veo 3.1", "Seedance 2.0", "Kling v3 4K", "LTX 2.3", "Happy Horse"). The plugin auto-routes to the family's text-to-video endpoint when called without ``image_url``, and to its image-to-video endpoint when ``image_url`` is provided. The agent never sees the routing — it just calls ``video_generate(prompt=..., image_url=...)``. Model families (most expose both t2v + i2v; gemini-omni-flash is image-to-video only): Cheap tier: ltx-2.3 fal-ai/ltx-2.3-22b/text-to-video / fal-ai/ltx-2.3-22b/image-to-video pixverse-v6 fal-ai/pixverse/v6/text-to-video / fal-ai/pixverse/v6/image-to-video seedance-2.0-mini bytedance/seedance-2.0/mini/text-to-video / bytedance/seedance-2.0/mini/image-to-video Premium tier: veo3.1 fal-ai/veo3.1 / fal-ai/veo3.1/image-to-video seedance-2.0 bytedance/seedance-2.0/text-to-video / bytedance/seedance-2.0/image-to-video seedance-2.5 bytedance/seedance-2.5/text-to-video / bytedance/seedance-2.5/image-to-video minimax-h3 minimax/h3/text-to-video / minimax/h3/image-to-video minimax-h3-max minimax/h3-max/text-to-video / minimax/h3-max/image-to-video flux-3 blackforestlabs/flux-3/text-to-video / blackforestlabs/flux-3/image-to-video grok-imagine-1.5 xai/grok-imagine-video/v1.5/text-to-video / xai/grok-imagine-video/v1.5/image-to-video kling-v3-4k fal-ai/kling-video/v3/4k/text-to-video / fal-ai/kling-video/v3/4k/image-to-video happy-horse alibaba/happy-horse/text-to-video / alibaba/happy-horse/image-to-video Image-to-video only (no text_endpoint): gemini-omni-flash google/gemini-omni-flash/image-to-video Selection precedence for the active family: 1. ``model=`` arg from the tool call 2. ``FAL_VIDEO_MODEL`` env var 3. ``video_gen.fal.model`` in ``config.yaml`` 4. ``video_gen.model`` in ``config.yaml`` (when it's one of our family IDs or a full endpoint path that contains a family ID) 5. ``DEFAULT_MODEL`` Authentication via ``FAL_KEY`` or the managed Nous gateway. Output is an HTTPS URL from FAL's CDN; the gateway downloads and delivers it. """ from __future__ import annotations import logging import os import threading import uuid from typing import Any, Dict, List, Optional, Tuple from agent.video_gen_provider import ( VideoGenProvider, error_response, success_response, ) logger = logging.getLogger(__name__) # --------------------------------------------------------------------------- # Family catalog # --------------------------------------------------------------------------- # # Each family declares both endpoints (when available) plus a per-family # capability sheet derived from FAL's OpenAPI schemas. Capability flags # drive which keys get added to the request payload — keys a family doesn't # advertise are dropped before send. # # Capabilities: # aspect_ratios : tuple of supported ratios (None = endpoint decides) # resolutions : tuple of supported resolutions (None = endpoint decides) # durations : tuple of supported durations OR (min, max) range # (heuristic: 2-element with gap > 1 is a range) # audio : True if generate_audio is supported # negative : True if negative_prompt is supported # seed : False when the endpoint declares no `seed` field # (absent = True, so existing families keep sending it) # duration_int : True when FAL types duration as an integer rather than # the usual queue-API string FAL_FAMILIES: Dict[str, Dict[str, Any]] = { # ─── Cheap / fast tier ───────────────────────────────────────────── "ltx-2.3": { "display": "LTX 2.3 (22B)", "speed": "~30-60s", "price": "cheap", "strengths": "22B model with native audio generation. Affordable.", "tier": "cheap", "text_endpoint": "fal-ai/ltx-2.3-22b/text-to-video", "image_endpoint": "fal-ai/ltx-2.3-22b/image-to-video", # LTX docs don't expose duration/aspect/resolution enums — leave # blank so we don't send unrecognized payload keys. "aspect_ratios": None, "resolutions": None, "durations": None, "audio": True, "negative": True, "seed": True, }, "pixverse-v6": { "display": "Pixverse v6", "speed": "~30-90s", "price": "cheap", "strengths": "Affordable. Negative prompts. 1-15s durations.", "tier": "cheap", "text_endpoint": "fal-ai/pixverse/v6/text-to-video", "image_endpoint": "fal-ai/pixverse/v6/image-to-video", "aspect_ratios": None, "resolutions": ("360p", "540p", "720p", "1080p"), "durations": (1, 15), "audio": True, "negative": True, "seed": True, }, "seedance-2.0-mini": { "display": "Seedance 2.0 Mini", "speed": "~30-90s", "price": "cheap", "strengths": "ByteDance. Faster/cheaper Seedance tier, audio + lip-sync, 4-15s.", "tier": "cheap", "text_endpoint": "bytedance/seedance-2.0/mini/text-to-video", "image_endpoint": "bytedance/seedance-2.0/mini/image-to-video", "aspect_ratios": ("21:9", "16:9", "4:3", "1:1", "3:4", "9:16"), "resolutions": ("480p", "720p"), "durations": (4, 15), "audio": True, "negative": False, "seed": False, }, # ─── Expensive / premium tier ────────────────────────────────────── "veo3.1": { "display": "Veo 3.1", "speed": "~60-120s", "price": "premium", "strengths": "Google DeepMind. Cinematic, native audio, strong prompt adherence.", "tier": "premium", "text_endpoint": "fal-ai/veo3.1", "image_endpoint": "fal-ai/veo3.1/image-to-video", "aspect_ratios": ("16:9", "9:16"), "resolutions": ("720p", "1080p", "4k"), "durations": (4, 6, 8), "duration_suffix": "s", # FAL veo3.1 wants "4s" not "4" "audio": True, "negative": True, "seed": True, }, "seedance-2.0": { "display": "Seedance 2.0", "speed": "~60-120s", "price": "premium", "strengths": "ByteDance. Cinematic, synchronized audio + lip-sync, 4-15s.", "tier": "premium", "text_endpoint": "bytedance/seedance-2.0/text-to-video", "image_endpoint": "bytedance/seedance-2.0/image-to-video", # Seedance accepts "auto" too — we omit it from the enum so the # agent can't pass it; the endpoint defaults handle the rest. "aspect_ratios": ("21:9", "16:9", "4:3", "1:1", "3:4", "9:16"), "resolutions": ("480p", "720p", "1080p"), "durations": (4, 15), "audio": True, "negative": False, # FAL input schema has no `seed` (only returned on output). "seed": False, }, "seedance-2.5": { "display": "Seedance 2.5", "speed": "~60-180s", "price": "premium", "strengths": "ByteDance flagship. Native 30s single-pass, audio in the same latent space, lip-sync.", "tier": "premium", "text_endpoint": "bytedance/seedance-2.5/text-to-video", "image_endpoint": "bytedance/seedance-2.5/image-to-video", # i2v accepts only "auto" for aspect_ratio (it follows the input # image), so aspect_ratio is dropped for image jobs via # image_drop_keys. "image_drop_keys": ("aspect_ratio",), "aspect_ratios": ("21:9", "16:9", "4:3", "1:1", "3:4", "9:16"), "resolutions": ("480p", "720p"), "durations": (4, 30), "audio": True, "negative": False, "seed": False, }, "minimax-h3": { "display": "MiniMax H3", "speed": "~60-180s", "price": "premium", "strengths": "MiniMax frontier. Native 2K (up to 4K), 5-15s, seven aspect ratios.", "tier": "premium", "text_endpoint": "minimax/h3/text-to-video", "image_endpoint": "minimax/h3/image-to-video", # H3 takes duration as a JSON integer, not the stringified form # most FAL endpoints use. "duration_int": True, # i2v derives the aspect ratio from the input image and rejects # the key entirely. "image_drop_keys": ("aspect_ratio",), "aspect_ratios": ("21:9", "16:9", "4:3", "1:1", "3:4", "9:16"), # H3 uses capitalized/2K-style resolution enums — mapped from the # tool's usual 720p/1080p-style values via resolution_aliases. "resolutions": ("768P", "2K", "4K"), "resolution_aliases": { "480p": "768P", "540p": "768P", "720p": "768P", "768p": "768P", "1080p": "2K", "2k": "2K", "4k": "4K", "2160p": "4K", }, "durations": (5, 15), "audio": False, # no generate_audio TOGGLE — audio is always on "audio_native": True, # native audio in every generation (fal docs) # audio is native/always-on; no generate_audio key "negative": False, "seed": False, }, "minimax-h3-max": { "display": "MiniMax H3 Max (fal post-train)", "speed": "~5-30s", "price": "premium", "strengths": "fal's post-trained MiniMax H3. Top-ranked quality/prompt adherence/aesthetics, 768p in seconds, 5-15s.", "tier": "premium", "text_endpoint": "minimax/h3-max/text-to-video", "image_endpoint": "minimax/h3-max/image-to-video", # Same wire quirks as base H3: integer duration, i2v derives the # aspect ratio from the input image (t2v-only key on Max: the i2v # schema doesn't declare aspect_ratio at all). "duration_int": True, "image_drop_keys": ("aspect_ratio",), "aspect_ratios": ("21:9", "16:9", "4:3", "1:1", "3:4", "9:16"), # Max tops out at 768P (no 2K/4K tiers like base H3); map the # tool's usual values onto the two capitalized enums. "resolutions": ("480P", "768P"), "resolution_aliases": { "480p": "480P", "540p": "480P", "720p": "768P", "768p": "768P", "1080p": "768P", "2k": "768P", "4k": "768P", "2160p": "768P", }, "durations": (5, 15), # `prompt_expansion_mode` is in the schema's required array (with a # "balanced" default) — always send it. "static_payload": {"prompt_expansion_mode": "balanced"}, "audio": False, # no generate_audio TOGGLE — audio is always on "audio_native": True, # native audio in every generation (fal docs) # audio is native/always-on; no generate_audio key "negative": False, "seed": True, # Unlike base H3, Max declares `seed` on both endpoints. }, "flux-3": { "display": "FLUX 3 (via FAL)", "speed": "~60-120s", "price": "premium", "strengths": "Black Forest Labs frontier video. Native audio, 5-20s, 8 aspect ratios.", "tier": "premium", "text_endpoint": "blackforestlabs/flux-3/text-to-video", "image_endpoint": "blackforestlabs/flux-3/image-to-video", # FLUX 3 duration enum is "auto" | 5..20 as JSON integers. "duration_int": True, "aspect_ratios": ("21:9", "2:1", "16:9", "4:3", "1:1", "3:4", "9:16"), "resolutions": ("720p", "1080p"), "durations": (5, 20), "audio": True, "negative": False, "seed": False, }, "grok-imagine-1.5": { "display": "Grok Imagine 1.5 (via FAL)", "speed": "~30-90s", "price": "premium", "strengths": "xAI. Fast stylized video with audio, 1-15s, cheap per second.", "tier": "premium", "text_endpoint": "xai/grok-imagine-video/v1.5/text-to-video", "image_endpoint": "xai/grok-imagine-video/v1.5/image-to-video", "duration_int": True, # i2v derives aspect from the input image; the key is t2v-only. "image_drop_keys": ("aspect_ratio",), "aspect_ratios": ("16:9", "4:3", "3:2", "1:1", "2:3", "3:4", "9:16"), "resolutions": ("480p", "720p", "1080p"), "durations": (1, 15), "audio": False, # no generate_audio TOGGLE — audio is always on "audio_native": True, # native audio in every generation (fal docs) # audio is native; no generate_audio key "negative": False, "seed": False, }, "gemini-omni-flash": { "display": "Gemini Omni Flash (via FAL)", "speed": "~60-120s", "price": "premium", "strengths": "Google. Image-to-video with audio, physics-grounded motion, 3-10s.", "tier": "premium", # No text-to-video endpoint on FAL — image/reference only. "text_endpoint": None, "image_endpoint": "google/gemini-omni-flash/image-to-video", "duration_int": True, "aspect_ratios": ("16:9", "9:16"), "resolutions": None, "durations": (3, 10), "audio": False, # no generate_audio TOGGLE — audio is always on "audio_native": True, # native audio in every generation (fal docs) # audio is native; no generate_audio key "negative": False, "seed": False, }, "kling-v3-4k": { "display": "Kling v3 4K", "speed": "~120-300s", "price": "premium", "strengths": "4K output, native audio (Chinese/English), 3-15s.", "tier": "premium", "text_endpoint": "fal-ai/kling-video/v3/4k/text-to-video", "image_endpoint": "fal-ai/kling-video/v3/4k/image-to-video", # Kling 4K image-to-video uses `start_image_url` instead of # `image_url`. Handled in _build_payload via image_param_key. "image_param_key": "start_image_url", "aspect_ratios": ("16:9", "9:16", "1:1"), "resolutions": None, # 4K is implicit "durations": (3, 15), "audio": True, "negative": True, "seed": True, }, "happy-horse": { "display": "Happy Horse 1.0", "speed": "~60-120s", "price": "premium", "strengths": "Alibaba. New model, sparse public docs — conservative defaults.", "tier": "premium", "text_endpoint": "alibaba/happy-horse/text-to-video", "image_endpoint": "alibaba/happy-horse/image-to-video", # Docs don't expose duration/aspect/resolution — let the endpoint # apply its own defaults. "aspect_ratios": None, "resolutions": None, "durations": None, "audio": False, # no generate_audio TOGGLE — audio is always on "audio_native": True, # native audio in every generation (fal docs) "negative": False, "seed": True, }, } DEFAULT_MODEL = "pixverse-v6" # cheap, both modalities, sane defaults def _is_duration_range(durations: Any) -> bool: """Heuristic: a 2-tuple of ints with a gap > 1 is treated as ``(min, max)``.""" if not isinstance(durations, tuple) or len(durations) != 2: return False if not all(isinstance(d, int) for d in durations): return False return durations[1] - durations[0] > 1 def _clamp_duration(family: Dict[str, Any], duration: Optional[int]) -> Optional[int]: durations = family.get("durations") if not durations: return duration if duration is None: # Range families (e.g. pixverse-v6 (1,15)) should omit the field so # the FAL endpoint applies its own default rather than receiving the # minimum value. Enum families (e.g. veo3.1 (4,6,8)) keep sending # their first entry as the default. return None if _is_duration_range(durations) else durations[0] if _is_duration_range(durations): lo, hi = durations return max(lo, min(hi, duration)) # enum if duration in durations: return duration return min(durations, key=lambda d: abs(d - duration)) # --------------------------------------------------------------------------- # Config / model resolution # --------------------------------------------------------------------------- def _load_video_gen_section() -> Dict[str, Any]: try: from hermes_cli.config import load_config cfg = load_config() section = cfg.get("video_gen") if isinstance(cfg, dict) else None return section if isinstance(section, dict) else {} except Exception as exc: logger.debug("Could not load video_gen config: %s", exc) return {} _ENDPOINT_MODALITY_LEAVES = frozenset({"text-to-video", "image-to-video"}) def _normalize_family_key(c: str) -> Optional[str]: """Try to extract a known family ID from a model string. Handles bare IDs (``seedance-2.5``), full endpoint paths (``bytedance/seedance-2.5/text-to-video``), truncated endpoint stems (``minimax/h3``, ``bytedance/seedance-2.0/mini``), and provider-prefixed names (``bytedance/seedance-2.5``). """ c = c.strip() if not c: return None if c in FAL_FAMILIES: return c # Exact declared endpoint — unambiguous, and beats any segment scan # that would otherwise see "seedance-2.0" inside ".../seedance-2.0/mini/...". for fid, meta in FAL_FAMILIES.items(): if c in (meta.get("text_endpoint"), meta.get("image_endpoint")): return fid # Truncated stem of a declared endpoint: "minimax/h3" or # "bytedance/seedance-2.0/mini". The next path segment after ``c`` must # be a modality leaf so "bytedance/seedance-2.0" does not also match the # Mini family's deeper ".../seedance-2.0/mini/text-to-video" path. stem_hits: List[Tuple[int, str]] = [] for fid, meta in FAL_FAMILIES.items(): for endpoint in (meta.get("text_endpoint"), meta.get("image_endpoint")): if not isinstance(endpoint, str): continue if not endpoint.startswith(c + "/"): continue first = endpoint[len(c) + 1:].split("/", 1)[0] if first in _ENDPOINT_MODALITY_LEAVES: stem_hits.append((len(c), fid)) break if stem_hits: stem_hits.sort(key=lambda item: item[0], reverse=True) return stem_hits[0][1] # Longest family-id path-segment match ("bytedance/seedance-2.5" → # seedance-2.5; prefers seedance-2.0-mini over seedance-2.0 when both # somehow appear). parts = set(c.split("/")) best_fid: Optional[str] = None best_len = -1 for fid in FAL_FAMILIES: if fid in parts and len(fid) > best_len: best_fid = fid best_len = len(fid) return best_fid def _resolve_family(explicit: Optional[str]) -> Tuple[str, Dict[str, Any]]: """Decide which FAL family to use. Returns ``(family_id, meta)``.""" candidates: List[Optional[str]] = [] candidates.append(explicit) candidates.append(os.environ.get("FAL_VIDEO_MODEL")) cfg = _load_video_gen_section() fal_cfg = cfg.get("fal") if isinstance(cfg.get("fal"), dict) else {} if isinstance(fal_cfg, dict): candidates.append(fal_cfg.get("model")) top = cfg.get("model") if isinstance(top, str): candidates.append(top) for c in candidates: if isinstance(c, str) and c.strip(): fid = _normalize_family_key(c) if fid: return fid, FAL_FAMILIES[fid] return DEFAULT_MODEL, FAL_FAMILIES[DEFAULT_MODEL] # --------------------------------------------------------------------------- # Payload construction # --------------------------------------------------------------------------- def _build_payload( family: Dict[str, Any], *, prompt: str, image_url: Optional[str], duration: Optional[int], aspect_ratio: str, resolution: str, negative_prompt: Optional[str], audio: Optional[bool], seed: Optional[int], ) -> Dict[str, Any]: """Build a family-specific payload, dropping keys the family doesn't declare.""" payload: Dict[str, Any] = {} if prompt: payload["prompt"] = prompt if image_url: # Some endpoints (e.g. Kling v3 4K image-to-video) expect # `start_image_url` instead of `image_url`. The family entry can # declare an override. key = family.get("image_param_key") or "image_url" payload[key] = image_url # Several newer endpoints (seedance 2.x, minimax h3, flux-3, grok, gemini) # declare no `seed` field, and the managed gateway forwards whatever we # send — so gate it on the family rather than leaking an unknown key. if seed is not None and family.get("seed", True): payload["seed"] = seed if family.get("aspect_ratios"): if aspect_ratio in family["aspect_ratios"]: payload["aspect_ratio"] = aspect_ratio # otherwise let the endpoint auto-crop / use its default if family.get("resolutions"): # Some families use non-standard resolution enums (e.g. MiniMax H3's # "768P"/"2K"/"4K"); resolution_aliases maps the tool's usual # 720p/1080p-style values onto them. aliases = family.get("resolution_aliases") or {} resolved = aliases.get((resolution or "").lower(), resolution) if resolved in family["resolutions"]: payload["resolution"] = resolved # else: let the endpoint default clamped = _clamp_duration(family, duration) if clamped is not None and family.get("durations"): if family.get("duration_int"): # A few endpoints (MiniMax H3) require duration as a JSON integer. payload["duration"] = clamped else: # FAL exposes duration as a string in the queue API ("8" not 8). # Some families (e.g. veo3.1) require a unit suffix ("4s" not "4"). suffix = family.get("duration_suffix", "") payload["duration"] = f"{clamped}{suffix}" if family.get("audio") and audio is not None: payload["generate_audio"] = bool(audio) if family.get("negative") and negative_prompt: payload["negative_prompt"] = negative_prompt # Keys the family's image-to-video endpoint rejects outright (e.g. # Seedance 2.5 / MiniMax H3 derive aspect_ratio from the input image). if image_url: for key in family.get("image_drop_keys", ()): # type: ignore[assignment] payload.pop(key, None) # Constant keys the endpoint requires on every request (e.g. MiniMax # H3 Max lists `prompt_expansion_mode` in its required array). for key, value in (family.get("static_payload") or {}).items(): payload.setdefault(key, value) return payload # --------------------------------------------------------------------------- # fal_client lazy import (shared with image_generation_tool via fal_common) # --------------------------------------------------------------------------- _fal_client: Any = None _fal_client_lock = threading.Lock() def _load_fal_client() -> Any: """Lazy-load the ``fal_client`` SDK and cache it on this module. Delegates the actual import to :func:`tools.fal_common.import_fal_client` so the ``lazy_deps`` ensure-install handling stays in one place. Thread-safe via double-checked locking: concurrent first calls import the SDK exactly once instead of each racing thread re-running the import. """ global _fal_client if _fal_client is not None: return _fal_client with _fal_client_lock: if _fal_client is not None: # re-check inside the lock return _fal_client from tools.fal_common import import_fal_client _fal_client = import_fal_client() return _fal_client # --------------------------------------------------------------------------- # Managed FAL gateway (Nous Subscription) # --------------------------------------------------------------------------- _managed_fal_video_client: Any = None _managed_fal_video_client_config: Any = None _managed_fal_video_client_lock = threading.Lock() def _resolve_managed_fal_video_gateway(): """Resolve the FAL video route from the stored selection. Plain switch on the stored ``video_gen`` provider string — mirrors the image FAL resolver: ``"nous"`` (or legacy ``use_gateway: true``) → managed only (unentitled ⇒ selection-naming error); any other stored provider → direct only (missing FAL_KEY ⇒ selection-naming error); never-configured category → legacy credential autodetect. """ from tools.managed_tool_gateway import resolve_managed_tool_gateway from tools.tool_backend_helpers import ( NOUS_MANAGED_PROVIDER, fal_key_is_configured, read_selection, selection_error, ) selected = read_selection("video_gen") if selected == NOUS_MANAGED_PROVIDER: gateway = resolve_managed_tool_gateway("fal-queue") if gateway is None: raise ValueError(selection_error( "video_gen", NOUS_MANAGED_PROVIDER, "the Nous Tool Gateway is not available (not entitled or " "unreachable)", )) return gateway if selected is not None: if not fal_key_is_configured(): raise ValueError(selection_error( "video_gen", selected, "FAL_KEY is not set", )) return None # Never-configured category: legacy credential autodetect (do NOT persist). if fal_key_is_configured(): return None return resolve_managed_tool_gateway("fal-queue") def _get_managed_fal_video_client(managed_gateway): """Reuse the managed FAL client so its internal httpx.Client is not leaked per call.""" global _managed_fal_video_client, _managed_fal_video_client_config from tools.fal_common import _ManagedFalSyncClient client_config = ( managed_gateway.gateway_origin.rstrip("/"), managed_gateway.nous_user_token, ) with _managed_fal_video_client_lock: if _managed_fal_video_client is not None and _managed_fal_video_client_config == client_config: return _managed_fal_video_client _load_fal_client() _managed_fal_video_client = _ManagedFalSyncClient( _fal_client, key=managed_gateway.nous_user_token, queue_run_origin=managed_gateway.gateway_origin, ) _managed_fal_video_client_config = client_config return _managed_fal_video_client def _submit_fal_video_request(endpoint: str, arguments: Dict[str, Any]): """Submit a FAL video request using direct credentials or the managed queue gateway. Returns a request handle whose ``.get()`` blocks until the result is ready. """ _load_fal_client() request_headers = {"x-idempotency-key": str(uuid.uuid4())} managed_gateway = _resolve_managed_fal_video_gateway() if managed_gateway is None: return _fal_client.submit(endpoint, arguments=arguments, headers=request_headers) managed_client = _get_managed_fal_video_client(managed_gateway) try: return managed_client.submit( endpoint, arguments=arguments, headers=request_headers, ) except Exception as exc: from tools.fal_common import _extract_http_status status = _extract_http_status(exc) if status is not None and 400 <= status < 500: raise ValueError( f"Nous Subscription gateway rejected endpoint '{endpoint}' " f"(HTTP {status}). This model may not yet be enabled on " f"the Nous Portal's FAL proxy. Either:\n" f" • Set FAL_KEY in your environment to use FAL.ai directly, or\n" f" • Pick a different model via `hermes tools` → Video Generation." ) from exc raise def _check_fal_video_available() -> bool: """True if the FAL video backend selected via `hermes tools` (or, on a never-configured install, any FAL backend) is reachable. Never raises — a stored-but-broken selection reports False here; the honest selection-naming error surfaces at call time from ``_resolve_managed_fal_video_gateway``. """ from tools.managed_tool_gateway import resolve_managed_tool_gateway from tools.tool_backend_helpers import ( NOUS_MANAGED_PROVIDER, fal_key_is_configured, read_selection, ) selected = read_selection("video_gen") if selected == NOUS_MANAGED_PROVIDER: return resolve_managed_tool_gateway("fal-queue") is not None if selected is not None: return fal_key_is_configured() if fal_key_is_configured(): return True return resolve_managed_tool_gateway("fal-queue") is not None # --------------------------------------------------------------------------- # Upscaler (SeedVR2 — video upscale pass) # --------------------------------------------------------------------------- # ByteDance SeedVR2 on FAL: $0.001/megapixel of output video. A 5s 720p→1440p # 2x pass is roughly $0.44. Faithful restoration-style upscaler (the same # model family Krea exposes as its "SeedVR2" video enhancer). UPSCALER_ENDPOINT = "fal-ai/seedvr/upscale/video" UPSCALER_FACTOR = 2 def _upscale_video( video_url: str, source_request_id: Optional[str] = None, ) -> Optional[str]: """Upscale a generated video via SeedVR2; return the new URL or None. Best-effort: any failure logs and returns ``None`` so the caller falls back to the native-resolution video — an upscale failure must never destroy an already-successful generation. """ try: logger.info("Upscaling video with SeedVR2 (%dx)...", UPSCALER_FACTOR) arguments: Dict[str, Any] = { "video_url": video_url, "upscale_mode": "factor", "upscale_factor": UPSCALER_FACTOR, } if _resolve_managed_fal_video_gateway() is not None: if not source_request_id: raise RuntimeError("Managed SeedVR upscale requires the source FAL request id") arguments["source_request_id"] = source_request_id handle = _submit_fal_video_request(UPSCALER_ENDPOINT, arguments) result = handle.get() except Exception as exc: # noqa: BLE001 logger.warning("Video upscale failed: %s", exc) return None video = (result or {}).get("video") if isinstance(result, dict) else None if isinstance(video, dict) and video.get("url"): return video["url"] if isinstance(video, str) and video: return video logger.warning("Video upscaler returned no URL") return None # --------------------------------------------------------------------------- # Provider # --------------------------------------------------------------------------- class FALVideoGenProvider(VideoGenProvider): """FAL.ai multi-family video generation backend. Routes between text-to-video and image-to-video endpoints automatically based on whether ``image_url`` was provided. """ @property def name(self) -> str: return "fal" @property def display_name(self) -> str: return "FAL" def is_available(self) -> bool: try: return _check_fal_video_available() except Exception: # noqa: BLE001 — never break the picker return False def list_models(self) -> List[Dict[str, Any]]: out: List[Dict[str, Any]] = [] for fid, meta in FAL_FAMILIES.items(): modalities: List[str] = [] if meta.get("text_endpoint"): modalities.append("text") if meta.get("image_endpoint"): modalities.append("image") entry: Dict[str, Any] = { "id": fid, "display": meta["display"], "speed": meta["speed"], "strengths": meta["strengths"], "price": meta["price"], "tier": meta.get("tier", "premium"), "modalities": modalities, } durs = meta.get("durations") if durs: if _is_duration_range(durs): entry["min_duration"], entry["max_duration"] = durs else: entry["min_duration"] = min(durs) entry["max_duration"] = max(durs) out.append(entry) return out def default_model(self) -> Optional[str]: return DEFAULT_MODEL def get_setup_schema(self) -> Dict[str, Any]: return { "name": "FAL", "badge": "paid", "tag": "LTX, Pixverse, Seedance 2.0/2.5/Mini, Veo 3.1, MiniMax H3, FLUX 3, Kling 4K, Happy Horse, Grok Imagine, Gemini Omni — text-to-video & image-to-video", "env_vars": [ { "key": "FAL_KEY", "prompt": "FAL.ai API key", "url": "https://fal.ai/dashboard/keys", }, ], } def capabilities(self) -> Dict[str, Any]: # Active-model-aware (mirrors the image_gen fal plugin, #97057): # report the RESOLVED family's actual surface so the dynamic tool # schema gates params on what the selected model honors, not a # union that overstates every axis. Falls back to the cross-family # union if resolution fails (never raises). try: _family_id, family = _resolve_family(None) except Exception: # noqa: BLE001 family = None if family: modalities = [] if family.get("text_endpoint"): modalities.append("text") if family.get("image_endpoint"): modalities.append("image") durs = family.get("durations") or (1, 1) if _is_duration_range(durs): lo, hi = durs else: lo, hi = min(durs), max(durs) return { "modalities": modalities or ["text"], "aspect_ratios": list(family.get("aspect_ratios") or []), "resolutions": list(family.get("resolutions") or []), "max_duration": hi, "min_duration": lo, "supports_audio": bool(family.get("audio")), # Always-on native audio (no toggle): surfaces as a # description line, not a param. Verified per-family # against fal model pages (H3, Grok 1.5, Happy Horse, # Gemini Omni Flash all return native audio every run). "audio_always_on": bool(family.get("audio_native")), "supports_negative_prompt": bool(family.get("negative")), # Explicit per-family key (contract-tested); absent would # mean a catalog bug, so fail closed here. "supports_seed": bool(family.get("seed", False)), # SeedVR upscaler chains for any FAL video family. "supports_upscale": True, "max_reference_images": 0, } # Fallback: union across families (legacy shape). max_dur = 1 min_dur: Optional[int] = None for meta in FAL_FAMILIES.values(): durs = meta.get("durations") if not durs: continue if _is_duration_range(durs): lo, hi = durs else: lo, hi = min(durs), max(durs) max_dur = max(max_dur, hi) min_dur = lo if min_dur is None else min(min_dur, lo) return { "modalities": ["text", "image"], "aspect_ratios": ["16:9", "9:16", "1:1"], "resolutions": ["360p", "540p", "720p", "1080p"], "max_duration": max_dur, "min_duration": min_dur if min_dur is not None else 1, "supports_audio": True, "supports_negative_prompt": True, "supports_seed": True, "supports_upscale": True, "max_reference_images": 0, } def generate( self, prompt: str, *, model: Optional[str] = None, image_url: Optional[str] = None, reference_image_urls: Optional[List[str]] = None, duration: Optional[int] = None, aspect_ratio: str = "16:9", resolution: str = "720p", negative_prompt: Optional[str] = None, audio: Optional[bool] = None, seed: Optional[int] = None, upscale: Optional[bool] = None, **kwargs: Any, ) -> Dict[str, Any]: if not _check_fal_video_available(): from tools.tool_backend_helpers import read_selection if read_selection("video_gen") is not None: # A stored selection that cannot run gets the honest # selection-naming error from the strict resolver. try: _resolve_managed_fal_video_gateway() except ValueError as exc: return error_response( error=str(exc), error_type="auth_required", provider="fal", prompt=prompt, ) return error_response( error=( "No FAL backend available. Either set FAL_KEY " "(run `hermes tools` → Video Generation → FAL to configure) " "or sign in to Nous (`hermes setup`) for managed gateway access." ), error_type="auth_required", provider="fal", prompt=prompt, ) try: _load_fal_client() except ImportError: return error_response( error="fal_client Python package not installed (pip install fal-client)", error_type="missing_dependency", provider="fal", prompt=prompt, ) prompt = (prompt or "").strip() family_id, family = _resolve_family(model) # Route: image_url → image-to-video endpoint; else → text-to-video. image_url_norm = (image_url or "").strip() or None if image_url_norm: endpoint = family.get("image_endpoint") modality_used = "image" if not endpoint: return error_response( error=( f"FAL family {family_id} has no image-to-video " f"endpoint. Pick a family with image-to-video support " f"via `hermes tools` → Video Generation." ), error_type="modality_unsupported", provider="fal", model=family_id, prompt=prompt, ) else: endpoint = family.get("text_endpoint") modality_used = "text" if not endpoint: return error_response( error=( f"FAL family {family_id} has no text-to-video " f"endpoint. Pass an image_url to use its " f"image-to-video endpoint, or pick a different family." ), error_type="modality_unsupported", provider="fal", model=family_id, prompt=prompt, ) if not prompt: return error_response( error="prompt is required.", error_type="missing_prompt", provider="fal", model=family_id, prompt=prompt, ) payload = _build_payload( family, prompt=prompt, image_url=image_url_norm, duration=duration, aspect_ratio=aspect_ratio, resolution=resolution, negative_prompt=negative_prompt, audio=audio, seed=seed, ) try: handle = _submit_fal_video_request(endpoint, payload) source_request_id = getattr(handle, "request_id", None) result = handle.get() except Exception as exc: logger.warning( "FAL video gen failed (family=%s, endpoint=%s): %s", family_id, endpoint, exc, exc_info=True, ) return error_response( error=f"FAL video generation failed: {exc}", error_type="api_error", provider="fal", model=family_id, prompt=prompt, aspect_ratio=aspect_ratio, ) video = (result or {}).get("video") if isinstance(result, dict) else None url: Optional[str] = None if isinstance(video, dict): url = video.get("url") elif isinstance(video, str): url = video if not url: return error_response( error="FAL returned no video URL in response", error_type="empty_response", provider="fal", model=family_id, prompt=prompt, ) # Optional high-resolution pass (SeedVR2). Explicit agent/user opt-in # only; best-effort — failure falls back to the native-resolution # video rather than failing the generation. upscaled = False if upscale: upscaled_url = _upscale_video(url, source_request_id) if upscaled_url: url = upscaled_url upscaled = True else: logger.warning( "Video upscale pass failed — returning native-resolution video" ) extra: Dict[str, Any] = {"endpoint": endpoint, "upscaled": upscaled} if upscaled: extra["upscale_factor"] = UPSCALER_FACTOR if isinstance(video, dict): if video.get("file_size") and not upscaled: extra["file_size"] = video["file_size"] if video.get("content_type"): extra["content_type"] = video["content_type"] return success_response( video=url, model=family_id, prompt=prompt, modality=modality_used, aspect_ratio=aspect_ratio if "aspect_ratio" in payload else "", duration=int("".join(c for c in str(payload["duration"]) if c.isdigit()) or "0") if "duration" in payload else 0, provider="fal", extra=extra, ) # --------------------------------------------------------------------------- # Plugin entry point # --------------------------------------------------------------------------- def register(ctx) -> None: """Plugin entry point — wire ``FALVideoGenProvider`` into the registry.""" ctx.register_video_gen_provider(FALVideoGenProvider())