I would second that opinion that it looks too similar to Stardew Valley. There are subtle differences, but to a certain point, if a game appears too similar, it can be distracting for the audience and the identity of your project will be lost. Maybe try altering the color pallet? It seems almost identical to Stardew Valley's
I think the palette is actually quite nice, more pleasing to my eyes than Stardew Valley's. The similarities, to me, are mostly in the (large amounts of) textures in elements like the grass and bushes. The tress look great and not SV-like, though, so consider making the other textures more in line with those - simple, with only suggested textures.
The character faces and poses also resemble SV, and not in a good way - they're rather lifeless, characterless, the characters all look like they're posing for a passport photo. Give them more personality and you'll kill two birds with one stone!
This is a 3 minute edit to remove shadows and increase spume opacity over the underwater shells.
(https://i.imgur.com/Jm33Msa.png)
Not my best work but I think they look more underwater now.
Especially the leftmost shell.
The noisy texture of the sand feels out of place with the style of all the other assets, which uses more solid colours and textures made out of larger shapes.
Here's some potential sand inspiration (https://i.imgur.com/Ih3XSCC.png) from Sword of Mana (I've edited the colours in this image to help with clarity, the original colours are washed out because the game was made for the original, non-backlit GBA).
About using different styles, be careful. If you use a consistent style throughout your game will seem tighter, more consistent and more professional. If you use different styles you run the risk of looking like a game assembled from asset packs.
The only game I've seen work with different styles well was one where each level was a different era in computer game graphics.
You're acting as de facto art director for this project, and as far as I can tell, one of the jobs of an art director is to maintain a consistent art style throughout the project...
Ambitious project!
Though I have to draw your attention to the cliffs. They act more as a texture and less as a volume - you need stronger highlights and a stronger sense of depth. The pattern itself is fine, it just needs to pop out more.
Edit:
(https://i.imgur.com/yUvRRwr.png)
The light's going to hit the edge of the cliff first, then spread from the direction it's cast from (in the edit it's centre-left).
Hope this helps and good luck with the game!
Your character moving animation towards and backwards from player`s perspective looks weird... I think? Or maybe this is ok and just adds uniqueness to the game (this comes to mind after argue about similarity between this game and Stardew Valley). Correct me, then, but my thought that this man`s legs moving too far to the left and right (cant explain this more precisely, sorry for my english).
The dynamic lighting is looking very flat right now, like you're just lighting up sticker rather than a 3D object. I'm not actually sure that convincing whole-sprite lighting can be done with normal maps alone.
Part of the issue could be that you've been pretty extreme with the lighting in your examples. This is understandable as you're trying out a new effect and you want to see how far it can go.
Here's a random video on Hollow Knight. Not pixel art, but flat 2D.
https://www.youtube.com/watch?v=QWM4J_FR7gM
I like how dynamic lighting is handled here. The background is quite dark except around the main character, but the main character stays well lit throughout. Of course, when your game isn't set in a world of bugs underground, this effect would probably be better used at night...
Depending on the effect you're going for, a subtle vignette towards white around the edges at daytime could also be very pleasant. You could do this with a simple shader on the camera maybe...
For tilesets of natural features like grass and water, I prefer to do tilesets focusing on diagonals instead of verticals and horizontals, i.e. every tile is basically a corner tile. This allows for a smaller tileset, and more importantly, its structure encourages organic shapes by making straight lines harder to create. The drawback to this style of tileset is most existing automappers don't work well with it, though it could be managed with one based on Wang tiles.
I also think those single-column tiles are kind of useless in most scenarios, as IMHO they create distracting shapes in the environment. They're useful for small-scale games to depict smaller features, but for a large-scale game where a tile is the size of a person or smaller, they're not needed. They're also nearly straight lines by necessity in smaller tiles, which doesn't look natural for things like grass.
Ah, I think you misunderstood. By diagonals, I just meant diagonal transitions, i.e. corner tiles. They're still square tiles. Here's a simple example using just 2 types of grass, with 4 diagonal transition tiles (in theory you could use just 1 or 2 with appropriate flips, but for organic objects it helps to have variety).
(https://i.imgur.com/V6ZivFJ.png)
Typically, the chief drawback to using a tileset like this is the limits it imposes on the exact shapes you can create, but some of this can be mitigated with careful tile design. For example, you wouldn't expect to create "vertical" or "horizontal" edges that are an odd number of tiles in length (since they must be created out of pairs of diagonals in a saw pattern), but for noisy organic surfaces like grass, it's possible to make two diagonals tile with each other seamlessly too, allowing for such scenarios.
Don't forget that what is "minimal" depends on what the tiles are. Some things just naturally need more tiles, or different kinds of tiles, so you won't have the same tile layout for different things. For example, you probably wouldn't need both diagonal and squared corners for terrain, but they could be useful for paved paths.
Since your game will probably have player-modified terrain, you'll probably need a tileset that can handle more weird situations than the diagonals-only one above. IIRC 16 tiles per required pair of terrains is the minimum that's procgen-friendly, which is similar to what you posted before - no isolated/narrow strips of terrains. In farming games, it's common for players to modify the terrain so a tileset where you can easily compute the proper tile for a given environment is probably more important than a minimal tileset.