This is probably over-powered.
Normally, there's a 10% chance of a given portent roll portending a crit, split evenly between success and failure (ie., either a natural 1 or 20). ... with some fuzziness around using the portents for rolls that don't crit; my experience is that portents are used for those rolls rarely enough to not worry about them too much.
With the expanded range, there are 24 possible values, 6 of which are crits; the probability of portending a crit becomes a whopping 25% (6/24 = 1/4); moreover, the probability of portending a critical success becomes 16.66% (4/24 = 1/6)! The expanded crit range means that portending a critical success is over 3x more likely than normal, and over 50% more likely than portending a crit either way normally.
And, that's before considering the alterations it makes to 5e's bounded accuracy model. One of the design considerations for 5e is that it's hard to get bonuses to d20 rolls and AC that differ tremendously between characters (previous editions made it relatively easy to have a difference of 10 or more - and differences of 20 aren't terribly hard - in some of those areas). This is part of why items that grant bonuses to attacks are relatively rare compared to those that, say, grant advantage or allow a skill to be used in more situations (eg., instead of granting a bonus on a Dexterity (stealth) roll, an item might grant advantage on it or allow the wearer to make such a roll without needing cover). This alternate feature allows for up to a "+3" to any roll, and will grant some "bonus" 12.5% (3/24 = 1/8) of the time. Further, imposes a "-1 penalty" on an enemy 4.166% of the time (1/24).
Another way of looking at that is that it increases the average of a d20 from 10.5 to 12, letting portents add a "+1.5" bonus.
Simply increasing the range of values a d20 can take is potent, especially since the range is weighted towards the high end. This is a great boon for must-make saves or must-hit attacks, even more so than a standard portent.
The increased crit range is incredibly powerful, especially for forcing crits on attacks.
It's possible that the remainder of the class might make up for this power in other ways, though I remain skeptical. The linked subclass, as currently written, doesn't seem to do it, though. The alternate effects (eg., The Lovers conferring disadvantage on a CHA saving throw) are weighted towards boons (that is, the vast majority of the cards grant advantage), further increasing their utility. And, the other abilities of the subclass aren't weak, either.
I very much like the feel of this feature. The alternate effects of the fortune dice feel flavorful and are well-tied to the general tendency of bards to buff their allies. However, many of the effects can be replicated without fortune dice via providing assistance, limiting their utility. And, of course, they're pretty situational. I can definitely see where the urge to augment them comes from, and I don't disagree with that urge. But, I think that the expanded range of portent dice is too powerful. I don't know how to resolve that tension, though.
I'll note also that it's unclear whether the bard chooses whether the foretelling dice are roll-replacements or grant their other effect when they roll the dice or during the adventuring day. My reading suggests that they can choose later in the day, but I'd encourage clarifying that. My gut says that choosing when the dice are rolled means that they'll almost always be used as portent dice, since the other options are pretty situational.