New British speech vocab!

[Barneyntd]

On another message board, I stumbled across a dump of CM62025: the speech phrom from the ICL 0ne-Per-Desk. It doesn't have a standard index, so you can't use word numbers, but it is coded for the TMS5220, so you can use it if you know the word addresses. I have just analysed all the speech data, and can report that it contains 209 words, letters and suffixes. There is relatively little overlap with the standard Acorn phrom, except for numbers and letters.

I don't know who recorded it for ICL, but it sounds like a different recording of the same person, Kenneth Kendal. Here are a couple of examples:

Number.m4a

(18.09 KiB) Downloaded 24 times

Second.m4a

(20.27 KiB) Downloaded 17 times

Certainly close enough to use in the same sentence.

I have attached the list of word addresses; if you have this phrom in slot -n you can speak the word at addr by

Code: Select all

SOUND &FFC0-<n>,<addr>,0,0

as usual. I don't know what slot the physical chip uses.

[philpem]

Very interesting - I've been playing with these chips too. I have a CM62006 ROM in the queue to be dumped.

Where did you find the ROM dump? Edit - never mind, I think I've found it: https://forum.vcfed.org/index.php?threa ... e.1241972/
Interestingly it seems to have the wordlist at the end in some form. There's text, but the last character seems to be XORed with 0xFE - but sometimes it's 0xFC. Quite strange.
(edit 2) - the last two bytes seem to contain "38 D0" hex, and at $38D0 there's a load of 4-byte entries (04 19 02 00, 03 A6 02 00, 24 C5 02 00 ...) followed by the word table, which seems to be in length-string form.

Also Bluewizard and Python Wizard can be used to encode speech for these chips:

Bluewizard: https://github.com/patrick99e99/BlueWizard
Python_wizard: https://github.com/ptwz/python_wizard/

What a shame this didn't exist back in the days of the Beeb - who knows, it might have got more people interested in them!

[Pernod]

Very interesting - I've been playing with these chips too. I have a CM62006 ROM in the queue to be dumped.

Any idea where it came from? It's an unknown on the wiki https://en.wikipedia.org/wiki/User:Lord_Nightmare

[Barneyntd]

Interestingly it seems to have the wordlist at the end in some form. There's text, but the last character seems to be XORed with 0xFE - but sometimes it's 0xFC. Quite strange.
(edit 2) - the last two bytes seem to contain "38 D0" hex, and at $38D0 there's a load of 4-byte entries (04 19 02 00, 03 A6 02 00, 24 C5 02 00 ...) followed by the word table, which seems to be in length-string form.

The last letter is the 2s complement negation, i.e. &100 minus the ascii code. If the word length is even, there is a 00 padding byte, then the two-byte address of the speech data, not bit-reversed, followed by another byte which I haven't identified. However, not all the words are listed in the index: the alphabet, numbers and ordinals are not listed with the rest.

I examined the data to determine that the first word starts at byte 08, and my program PHINDEX to work through all the words in the order they are stored. If you compare my index to the index in the phrom, you can see (for example) that EIGHT, EIGHTEEN, EIGHTY, and ELEVEN are stored between DURING and EMERGENCY, but not listed in the phrom.

If you post a dump of CM62006, I can do the same for that.

[Barneyntd]

at $38D0 there's a load of 4-byte entries (04 19 02 00, 03 A6 02 00, 24 C5 02 00 ...) followed by the word table, which seems to be in length-string form.

I don't know what 0419 is (it's in the middle of Q), but 03A6, 24C5, 3448, ... 144D are the addresses of the letter O, the numbers 1—20, 30, 40, 50 and 80. Somewhere there must be a table of letters, and a table of ordinals, since these exist but are not in the table of words. My index doesn't contain a "ty" suffix for converting 6,7 & 9 to 60, 70 & 90 resp. but one ought to exist, since they don't exist as whole words. Maybe they used the letter T. It does contain two "th" suffixes for converting numbers to ordinals.

[Barneyntd]

Any idea where it came from? It's an unknown on the wiki https://en.wikipedia.org/wiki/User:Lord_Nightmare

Someone is selling these on ebay at £15 each, which seems a bit much for something with no provenance or documentation. I contacted the seller, who said it was part of a job lot they bought from the USA, but who knows...