I suggest a few options; For videos: * Lower credits usage for longer videos * Add Camera options * Add sound effects For images: * Keep Krea 2 Turbo at 2 credits * Put ImagineArt, Nano Banana 2 and Grok at 12 credits
Declined: Model prices follow what each model costs us to run, so we cannot set them below that. Cheaper and longer video is also discussed in #95.
Could members receive a little energy back for trying another dreamer's shared prompt? Exploring the same prompt through different models teaches us so much about colour, mood and composition, but energy cost can limit experimentation. Perhaps a small daily rebate for a limited number of completed tries, with credit staying with the original prompt creator, could encourage more sharing, learning and discovery.
Declined: We will not be giving energy back for Tries: a generation costs us the same to run whoever wrote the prompt. What we did add: every Try is now credited to its author, and "Tried by" shows who tried a dream.
Log in to join in.
Since all voters must rate every image shown (with at least 1 point), fake accounts cannot make their own image win unless that image is also recognized as a winner by the other 1,000 plus voters. A fake account voter seeing their own image with all the fake accounts they have would only add five points per account — 5 accounts would add 25 points, hardly winner material unless most of the other voters share the enthusiasm for their images. They must also assign scores to the other 39 images in each batch in order to move on to the end. If the system randomly selects which 40 images each voter sees, they may not even encounter their own image.
Declined: Thanks for laying this out. Voting stays as it is: the server picks each entry you rate, fewest votes first, so every entry gets about as many ratings and nobody chooses what they see. Ballots that rate nearly everything the same no longer count.
Log in to join in.
Concerning voting in the challenge: it should be possible for everyone to vote as long as the challenge is active. Day 1: everyone votes on the first batch of participating dreams, Day 2: everyone votes on the second batch, Day 3: everyone votes on the last batch. That way everyone would see all the dreams and could vote on all of the dreams.
Declined: We are keeping how voting works: the server picks each entry you rate, fewest votes first, so every entry gets about the same number of ratings over the whole challenge. The earlier decision is explained in #48.
Log in to join in.
The way the current system works, we use up energy before publishing. This can make it difficult to put up what we really want. Waiting hours to publish something slows down productivity. If we can improve a dream before publishing it and not use up energy, the final result comes out better and faster.
Declined: A preview is itself a generation, so it costs the same energy as the image it shows; there is no cheaper way to see the result first. Publishing never costs energy, only generating does.
Log in to join in.
Oh thank god it helps me because that limited 82% OFF cannot revive it all it looks like some of 777 slot machine
Declined: This is not BUG report.
Log in to join in.
H3 SuperFast's 82% OFF unfortunately dies? ARE YOU FUCKING KIDDING ME!?!?
Log in to join in.
Bring back some sunsetted models, DuckDuckGo!?
Declined: There was never a model named DuckDuckGo on DDG.
Log in to join in.
It seems that it always interprets the source image as a prompt instead of taking the source image directly. 1 year ago the results looked like the style of the source, now it became useless. I used it with to adapt the syles of drawings that I drew myself some years ago . The image with the musicians is based on my drawing the other one is the boring one from now, using the same additional prompt.
Declined: Yes, it is now working only by generating prompt and then use it to generate the image. The old option is no longer supported. You can sue the Deep Style model to transfer style though.
seabird
·
Log in to join in.
Either would be an improvement. I enjoy voting, so I'd rather keep it than lose it.
Log in to join in.
The author suggests that DDG could program the AI assistant to handle the judging. They note they have no idea if it is possible, but the idea has stuck with them and they keep thinking about it.
Log in to join in.
To ensure that the second image is never identical to the original, the platform should handle the seed value dynamically. By automatically regenerating a new random seed whenever a user clicks "Try," the platform guarantees a unique variation.
Log in to join in.
https://deepdreamgenerator.com/ai-tools/kiss-generator - "Page not found." Mentioned here https://deepdreamgenerator.com/support I was very disappointed :D
Declined: We had to put it down because of compliance issues with the payment provider. You still can generate similar video in the Video Generator though.
Gronk
·
Log in to join in.
https://deepdreamgenerator.com/ai-tools/kiss-generator - "Page not found." Mentioned here https://deepdreamgenerator.com/support I was very disappointed :D
Declined: We had to put it down because of compliance issues with the payment provider. You still can generate similar video in the Video Generator though.
Gronk
·
Log in to join in.
Log in to join in.