ATH-MaaS/Pixelle-Video · error · ValueError

Prompt count mismatch: expected

Error message

Prompt count mismatch: expected {len(batch_narrations)}, got {len(batch_prompts)}

What it means

generate_video_prompts checks each batch so the number of returned video prompts equals the number of narrations; on mismatch it raises ValueError with expected and got counts. There is no retry-with-continue path here (unlike the image generator), so the first count mismatch aborts the call.

Solutions

Shrink batch size so the model reliably emits one prompt per narration.
Add retry logic (wrap the batch call like generate_image_prompts does) to regenerate mismatched batches.
Make the prompt state the exact expected count explicitly before the narrations.
Truncate/pad the array to len(batch_narrations) instead of raising, if approximate output is acceptable.
Use schema-constrained output with minItems/maxItems equal to the batch size.

Example fix

// before
if len(batch_prompts) != len(batch_narrations):
    raise ValueError(f"Prompt count mismatch: expected {len(batch_narrations)}, got {len(batch_prompts)}")
// after
if len(batch_prompts) > len(batch_narrations):
    batch_prompts = batch_prompts[:len(batch_narrations)]
elif len(batch_prompts) < len(batch_narrations):
    batch_prompts += [batch_prompts[-1]] * (len(batch_narrations) - len(batch_prompts))

Defensive patterns

Strategy: validation

Validate before calling

def validate_video_batch(result: dict, narrations: list) -> bool:
    prompts = result.get("video_prompts") if isinstance(result, dict) else None
    return isinstance(prompts, list) and len(prompts) == len(narrations)

Type guard

def is_valid_video_batch(obj: object, expected: int) -> bool:
    return (isinstance(obj, dict) and isinstance(obj.get("video_prompts"), list)
            and len(obj["video_prompts"]) == expected)

Try / catch

try:
    prompts = generator.generate_video_prompts(narrations)
except ValueError as e:
    if "count mismatch" in str(e):
        for size in (4, 2, 1):  # shrink batch until model complies
            prompts = generator.generate_video_prompts(narrations, batch_size=size)
            break
    else:
        raise

Prevention

Pass small batches so count alignment is trivial for the model
Explicitly number narrations in the prompt and require one prompt per number
Set max_tokens high enough that the JSON array is never truncated
Validate counts per batch immediately and regenerate only the failing batch

When it happens

Trigger: Any generate_video_prompts invocation where len(result['video_prompts']) != len(batch_narrations) for a batch — model omits entries, merges scenes, or adds extras.

Common situations: Large batches exceeding model reliability; narration text containing scene breaks the model splits; JSON arrays truncated by max-token limits; few-shot examples in the prompt showing a different count.

Related errors

AI-assisted analysis of ATH-MaaS/Pixelle-Video@848b054e4f (2026-08-30). Data as JSON: /api/errors/715c50e75b1e3802. Report an issue: GitHub.

Appendix: source

Thrown at pixelle_video/utils/content_generators.py:438

                response = await llm_service(
                    prompt=prompt,
                    temperature=0.7,
                    max_tokens=8192
                )
                
                logger.debug(f"Batch {batch_idx} attempt {attempt}: LLM response length: {len(response)} chars")
                
                # Parse JSON
                result = _parse_json(response)
                
                if "video_prompts" not in result:
                    raise KeyError("Invalid response format: missing 'video_prompts'")
                
                batch_prompts = result["video_prompts"]
                
                # Validate batch result
                if len(batch_prompts) != len(batch_narrations):
                    raise ValueError(
                        f"Prompt count mismatch: expected {len(batch_narrations)}, got {len(batch_prompts)}"
                    )
                
                # Success - add to all_prompts
                all_prompts.extend(batch_prompts)
                logger.info(f"✓ Batch {batch_idx} completed: {len(batch_prompts)} video prompts")
                
                # Report progress
                if progress_callback:
                    completed = len(all_prompts)
                    total = len(narrations)
                    progress_callback(completed, total, f"Batch {batch_idx}/{len(batches)} completed")
                
                break  # Success, move to next batch
            
            except Exception as e:
                logger.warning(f"✗ Batch {batch_idx} attempt {attempt} failed: {e}")
                if attempt >= max_retries:

View on GitHub (pinned to 848b054e4f)