4 ms·

I hope they have a plan to release DALL-E 3 sometime this year, MidJourney seems to be pulling ahead.

by jasfi 4y ago

jutrewag 4y ago
Midjourney is already far ahead. I can’t get nearly the same quality with DALLE2
Der_Einzige 4y ago
Open source is still so far ahead of midjourney it's not even funny. Like, a racist RimWorld mod author (Automatic1111) built a UI for stable diffusion which unlocks far more capabilities out of it than midjourney will ever have.
- ftxbro 4y ago
  I'm not sure if you are trolling or not, but if you aren't then you haven't seen Midjourney v5. But I wouldn't blame you because your information is only like one month out of date which is short in normal timespans but so long in AI timespans.
  - Der_Einzige 4y ago
    No, midjourney V5 is not close to bridging the gap. I can fix hands on any model within stable diffusion by using several control nets for the hands. Luckily a 3D openpose extension which makes that super easy already exists for the Automatic1111 SD UI!
    - speedgoose 4y ago
      MidJourney v5 doesn’t fail hands on most pictures though.
  - alienreborn 4y ago
    Out of the box midjourneyv5 generates great images but SD has controlnet, plethora of custom models to pick from, can run locally on most machines gives it an edge in my opinion.
    - speedgoose 4y ago
      ControlNet is great and is missing in MidJourney. I find most of the community models for Stable Diffusion to be pretty poor. A lot of models made from merges, waifu only models, models made from very small training datasets, and still stuck on 1.5 while 2.1 is much better overall. However you have the amazing LoRas for 2.1 that are very powerful. The community is a bit ignoring them, but I think the potential is great. But MidJourney is a lot better at text to image and image to image for now.
Imnimo 4y ago
I feel like Bing Image Creator is at least DALL-E 2.5, it feels like it has higher quality outputs for the same prompt. Could also just be some form of post-processing, though.
dougmwne 4y ago
I think Dall-e 3 basically exists in Bing. It is significantly better than Dall-e 2 and is close but not quite at the quality of Midjourny v5. I just generated a series of about 40 portraits and even the hands are significantly closer.

I hope they have a plan to release DALL-E 3 sometime this year, MidJourney seems to be pulling ahead.

by jasfi 4y ago

I hope they have a plan to release DALL-E 3 sometime this year, MidJourney seems to be pulling ahead.

jutrewag 4y ago

Midjourney is already far ahead. I can’t get nearly the same quality with DALLE2

Der_Einzige 4y ago

Open source is still so far ahead of midjourney it's not even funny. Like, a racist RimWorld mod author (Automatic1111) built a UI for stable diffusion which unlocks far more capabilities out of it than midjourney will ever have.

ftxbro 4y ago
I'm not sure if you are trolling or not, but if you aren't then you haven't seen Midjourney v5. But I wouldn't blame you because your information is only like one month out of date which is short in normal timespans but so long in AI timespans.
- Der_Einzige 4y ago
  No, midjourney V5 is not close to bridging the gap. I can fix hands on any model within stable diffusion by using several control nets for the hands. Luckily a 3D openpose extension which makes that super easy already exists for the Automatic1111 SD UI!
  - speedgoose 4y ago
    MidJourney v5 doesn’t fail hands on most pictures though.
- alienreborn 4y ago
  Out of the box midjourneyv5 generates great images but SD has controlnet, plethora of custom models to pick from, can run locally on most machines gives it an edge in my opinion.
  - speedgoose 4y ago
    ControlNet is great and is missing in MidJourney. I find most of the community models for Stable Diffusion to be pretty poor. A lot of models made from merges, waifu only models, models made from very small training datasets, and still stuck on 1.5 while 2.1 is much better overall. However you have the amazing LoRas for 2.1 that are very powerful. The community is a bit ignoring them, but I think the potential is great. But MidJourney is a lot better at text to image and image to image for now.

Imnimo 4y ago

I feel like Bing Image Creator is at least DALL-E 2.5, it feels like it has higher quality outputs for the same prompt. Could also just be some form of post-processing, though.

dougmwne 4y ago

I think Dall-e 3 basically exists in Bing. It is significantly better than Dall-e 2 and is close but not quite at the quality of Midjourny v5. I just generated a series of about 40 portraits and even the hands are significantly closer.