Ask a native English speaker from America to pronounce todas las noches and they will likely say
tʰoɾaz laz notʃes
While a native speaker from Lima, Peru, would instead say
toðaz laz notʃes
And I, from Buenos Aires, Argentina, would say
toðah lah notʃes.
Assuming the speaker gets all the vowel sounds right or close enough, and leaving aside the aspiration on the t, what sticks out like a sore thumb is the pronunciation of the D as ɾ (that is, like a Spanish ere, i.e. an R, specifically the R in hora) instead of as ð (that is, like a Spanish intervocalic de, i.e. a D).
The reason this happens is because of a feature of English pronunciation known as intervocalic alveolar flapping or tapping. To make a long story short, tapping is the pronunciation of /t/ and /d/ as ɾ when the sound is between vowels, or after /r/ and before a vowel, like in the words butter, metal/medal, howdy, writer/rider, barter, Hardy. For some speakers, it also happens after /l/ and before a vowel, provided the syllable is not immediately post-tonic (happening in faculty but not in altar). Flapping does not happen if /t/ or /d/ are starting a stressed syllable (in potato, the second T is flapped but the first one isn’t). It also doesn’t happen when the following vowel is unreduced and there is no word boundary (like in latex). For more details, follow the wikipedia link above and this link.
Tapping is widespread in the United States, Canada, Australia and New Zealand. For some reason (probably the fact that tapping produces a voiced sound), the tapping sound is associated by English speakers with a D sound and not with a T sound. Ask an American how the T in writer is different from the T in ten and they will tell you that “Americans actually pronounce writer with a D instead of a T”. This explanation makes little sense to a Spanish speaker – they would be better served by hearing “Americans pronounce the T in writer like the Spanish R in hora”.
The thing is, when an English speaker whose native accent has tapping learns Spanish, they quickly learn not to tap their Ts, but they do not learn not to tap their Ds. Thus the pronunciation of todas as toras, todo as toro, etc. This is a potential problem, because todo and toro are different words, and this pronunciation of D is very alien to Spanish native speakers.
Let’s take a closer look at Spanish and English pronunciation regarding /t/ and /d/. In English, these sounds are pronounced as alveolar sounds (pronounced with the tip of the tongue against the alveolar ridge, i.e. the bump in the palate slightly to the back of the top front teeth). In Spanish, they are pronounced as denti-alveolars, i.e. with the tip of the tongue against the back of the top front teeth and the center of the tongue against the alveolar ridge. This small difference might not appear terribly significant, all the more so because there isn’t much of an auditory difference between an alveolar and a denti-alveolar plosive. The biggest difference between the dental T in Spanish toco and the alveolar one in English ten is aspiration, not place of articulation*. Likewise, there is little difference between the dental D in Spanish diez and the alveolar one in English dice.
However, the fact that Spanish D is dental might help explain why its intervocalic pronunciation is a dental approximant ð̞, usually transcribed without the undertack, as ð. Those familiar with English phonetic transcription will remember that symbol as the one used to transcribe the TH sound in though, rather and tithe. However, the Spanish sound is approximant, not fricative – the tip of the tongue only approaches the teeth, but doesn’t quite touch it. Usually, to get learners to reproduce this sound, I tell them it’s like the TH in southern – soft, faint, relaxed, not quite touching. With that, and a few listens and a few tries, they finally say todo the way natives do.
The result is that this small detail will make a learner sound a lot more native. I have found that it is the most effective way to avoid tapping. Asking them to “not tap” their Ds (i.e. to pronounce their intervocalic Ds like the word-initial D in dice) seldom works. Tapping is a very automatic process for natives, and hard to avoid. Even if successful, the result will still come out sounding non-native. They are better served with substituting with the native equivalent, which is a totally different sound articulatorily speaking and runs no risk of being tapped. In this case, to avoid the non-native sounding pronunciation, the best bet is to aim for a fully native one.
* Likewise, there is little perceivable difference between the denti-alveolar tʰ of Mandarin and Cantonese and the alveolar tʰ of English.