Generate images with deterministic output using 'seed' and 'generationid' in dalle3

Question

I'm trying to create a comic-book style story using dalle3 and gpt35turbo. While I understand gpt35 supports reproducible output (deterministic) by using 'seed' but am not able to find any reference on how this can be used for image generation. I have recently read somewhere that you could use the 'generationid' (I am not sure how to obtain that) to generate reproducible images with dalle3 but am not sure how to.

Answer 1

Hello @Paarth Aggarwal ,

Currently, there is no usage of GenerationID on Azure OpenAI or OpenAI's DALL-E. Additionally, reproducible output is not supported on DALL-E. For models that do support this feature, please refer to the Microsoft documentation for further details.

Best regards,
Charlie

If you find my response helpful, please consider accepting this answer and voting 'yes' to support the community. Thank you!