affaan-m/ECC · error · FalError

no image URL in response from

Error message

no image URL in response from {ENDPOINTS['text_to_image']}

What it means

FalError raised by text_to_image() after submit() returns a result from which parse_urls() extracts no image URLs. It names the failing endpoint so you know which model's response was malformed. One image is requested (num_images=1), so an empty URL list means the provider did not return the expected output.

Solutions

Log the raw result dict to inspect what the endpoint actually returned
Rephrase the prompt if a content policy rejection is suspected and retry
Check the fal.ai model page for the endpoint's current response schema and update parsing
Wrap submit/parse in a retry for transient provider failures

Example fix

// before
urls = falapi.text_to_image(prompt)  # FalError: no image URL in response
// after
try:
    urls = falapi.text_to_image(prompt)
except falapi.FalError:
    result = falapi.submit(falapi.ENDPOINTS["text_to_image"], {"prompt": prompt, "num_images": 1})
    raise RuntimeError(f"text_to_image returned no URLs; keys={list(result) if isinstance(result, dict) else type(result)}")

Defensive patterns

Strategy: retry

Validate before calling

result = falapi.submit(falapi.ENDPOINTS["text_to_image"], {"prompt": prompt, "num_images": 1})
assert falapi.parse_urls(result), f"no image URLs; keys={list(result) if isinstance(result, dict) else type(result)}"

Try / catch

try:
    urls = falapi.text_to_image(prompt)
except falapi.FalError:
    prompt = sanitize_prompt(prompt)  # may have tripped content policy
    urls = falapi.text_to_image(prompt)

Prevention

Log raw provider responses in dev to spot schema changes early
Rephrase prompts that may trigger content filters
Add a retry with backoff around text_to_image for transient provider issues
Track the fal.ai model version; update parsing when the schema changes

When it happens

Trigger: text_to_image(prompt, ...) where the fal response lacks its image output field — provider response-shape change, content-policy rejection returning an error object, or model failure.

Common situations: Prompt tripped a provider safety filter and the endpoint returned no images; fal.ai updated the model's output schema; transient provider error surfaced as an empty result.

Understand the failure class

Background: "invalid response format", "malformed payload", "missing data field": when an API returns 200 but the response shape is wrong — this error's family across 23 libraries.

Related errors

AI-assisted analysis of affaan-m/ECC@8321021c54 (2026-09-16). Data as JSON: /api/errors/61423a1bddf5c4cd. Report an issue: GitHub.

Appendix: source

Thrown at skills/taste-distillation/scripts/taste/falapi.py:622

def quantize_duration(duration: float, lo: int = 3, hi: int = 12) -> int:
    """Round a planned shot length onto the video model's supported grid."""
    return int(max(lo, min(hi, round(float(duration)))))


def text_to_image(
    prompt: str,
    image_refs: list[str] | None = None,
    *,
    timeout: int = 300,
) -> list[str]:
    """Generate stills, optionally conditioned on reference images."""
    payload: dict[str, Any] = {"prompt": prompt, "num_images": 1}
    if image_refs:
        payload["image_urls"] = list(image_refs)
    result = submit(ENDPOINTS["text_to_image"], payload, timeout)
    urls = parse_urls(result)
    if not urls:
        raise FalError(f"no image URL in response from {ENDPOINTS['text_to_image']}")
    return urls


def extract_frame(video_url: str, timestamp: float, *, timeout: int = 300) -> str:
    """Pull a single frame out of a hosted video. Returns the image URL."""
    payload = {"video_url": video_url, "timestamp": round(float(timestamp), 3)}
    result = submit(ENDPOINTS["extract_frame"], payload, timeout)
    return first_url(result, ENDPOINTS["extract_frame"])


def compose(tracks: list[dict], *, timeout: int = 900) -> str:
    """Composite timeline tracks into one video. Returns the output URL.

    ``tracks`` is passed straight through so the caller owns the timeline
    shape; the ffmpeg-api track schema is another default worth verifying
    before a live run.
    """
    result = submit(ENDPOINTS["compose"], {"tracks": tracks}, timeout)

View on GitHub (pinned to 8321021c54)