LTX text enhancer is still broken
- Dominant language
- Python
- Stars
- 133k
- Forks
- 15.7k
- Avg merge
- 1d 7h
- Merged PRs (30d)
- 158
Description
### Custom Node Testing
- [x] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)
### Expected Behavior
Text enhancer is supposed to output one enhanced prompt
### Actual Behavior
The text enhancer sometimes outputs the entire enhanced prompt twice, nearly verbatim. This can cause duplicated actions or speech in the generated video.
### Steps to Reproduce
Use the LTX 2.5 text enhancer / Generate LTX2 Prompt node with prompt enhancement enabled.
Enter a normal video prompt and run the workflow several times, optionally changing the seed or prompt text between runs.
On some runs, the enhancer outputs the entire enhanced prompt twice instead of once.
The issue is intermittent: it does not happen with every seed or every input prompt. I have reproduced it with both Gemma 4 E2B and Gemma 4 E4B FP8.
Disabling prompt enhancement and using the generated prompt manually avoids the problem.
### Debug Logs
```powershell
...
```
### Other
_No response_
Contributor guide
Research direction
Start with the LTX 2.5 text enhancer / Generate LTX2 Prompt node and its prompt-enhancement path; the issue does not name implementation files or tests. Reproduce with enhancement enabled across multiple runs, using Gemma 4 E2B or E4B FP8. Done means the node returns one enhanced prompt rather than duplicating it.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 48/100