CANCELLED Request for Proposal for the Provision of development of four software applications of Text-To-Speech voices

Tender cancelled on 2022-11-30 18:09 UTC. None of the bids were technically compliant and hence the evaluation team could not make an award recommendation. For any inquiries, please contact UNOPS Procurement Unit Geneva Office. Email: gva.procurement@unops.org, Phone: Switzerland 229178722

New clarification added: Q8. Is the 30TH of April strict deadline for the project deadline as developing even 2 TTS voices from scratch takes minimum of 6-month effort?A8. The expected timeframe for the work is a five and half months period as stated in the Schedule of Requirements. The 30th April deadline will be postponed only in case an unexpected delay in the contract signature occurs, or in case of unforeseen circumstances, force majeure or other events beyond the control of the contractor during the implementation of the contract.Q9. As our TTS is a commercial product, is the open-source license a must? Is sharing our model and integrating through a platform enough? A9. The open source licence is a mandatory requirement of the project. 

New amendment added #1: The RFP is hereby amended as follows:1. Deadline for clarifications (General tab) is extended  to 17 October 2022 at 15:00 (UTC).2. Pre-bid Meeting minutes and the Pre-bid Meeting presentation are uploaded in the Section "Documents".

New clarification added: Q1.  Dear ATscale,Thank you so much for considering us as bidders. We are excited by the opportunity.We are exploring the technical nature of Turkmen and Nepali, and assessing logistics and effort. We continue to examine your request with great interest, but immediately we seek clarification on the following issues:1. Collaborating Organizations chosen by ATscale. We assume that these are in Turkmenistan and Nepal. From experience, we know that quite apart from voice accuracy, naturalness and understandability there are cultural aspects which can lead to dissatisfaction with a voice. This relates to standard "accents", with regional, city, social class being detected in a voice. Different communities using the same language can have different views. In addition, a key challenge in our work is sourcing acceptable "voice talents" to make recordings for us.For both of those reasons, and given the relative remoteness of Turkmenistan in particular, can we be placed in contact with your Collaborating Organizations during the bidding process?A1: We expect the selected vendor to set up country contact for the proposed languages. ATscale doesn’t have country offices. However, ATscale may  help support the selected vendor by reaching out to country contact during the implementation if the need arises but not guaranteed.  During the bidding process, any additional information or request for clarification should be made in writing through the eSourcing System, and  the answers will also be posted on eSourcing.Q2. Options for Languages and Voicesa) There is space for two more proposed languages on the bidding forms. May we propose more than two, and allow ATscale to chose from them?A2.  a)  The bidders are allowed to propose more than two optional languages as far as they meet the requirements stipulated in the ToR.b) Voice features. We have two alternative ways to provide "language switching" to English. Their results differ qualitatively and in cost.  May we propose both options for any or all languages?A2. b) As indicated in Section “Particulars”, alternative proposals  are not allowed in this RFP.  So, Offerors shall not submit more than one Proposal per Offeror in this RFP process. (Please see Section I “Instructions to offerors” Clause 15 “ALTERNATIVE PROPOSALS”).Q3. My apologies if this is a second request, but first request of 2022-10-03 19:37 UTC seens to be stuck in DRAFT status. I can find no means on the esourcing UNOPS website where I can submit it past draft.  Hence I am using this form to further explore options.A3.  All the questions that we received were answered. In case you sent a question that was not answered, please send it again so we can receive and answer it.Q4: a) Hi, I have a further question, below. But, COULD YOU PLEASE ACKNOWLEDGE RECEIPT of my request for clarification of last week? (We have not used the UNOPS eSourcing site before, and the signalls it gives on submission status are not completely unambiguous.)A4 a) Please see above Answer number 3 (A3).b) Form B: Financial Proposal Form.  Table 2A Cost Components. May we leave this table blank without fear of our bid being dismissed?This table requests numbers of working-days and charge-out rates.  Since your bid equest is for a fixed-price, turn-key solution, we do not see why the information requested here is appropriate.Any price we propose, will consist not only of personal efforts, but extermal costs (e.g. voice artists) and, importantly,cost contingency to ensure timely and quality deliverables.Further, our key members do not have a charge-out rate, or cost-recovery rate:  This is because we do not contract them on a Time and Materials consultancy basis.We would be prepared to give a very rough indication of man-days, and to add in the nature of externalities which effect our costs.Kind regards,A4 b) The bidder must complete and submit Table 2A.Q5. Section VI - Expected ActivitiesSecond bullet point regarding 100 sentences.   In order to be an objective test, we request adoption of the following process be considered by UNOPS:The 100 sentences will be selected randomly so as not to be over-weight words which can be expected to be challenging to a TTS (those would be homographs, non-phonemic words).The sentences will be selected by someone designated by UNOPS, and then placed in an encrypted file and deposited with vendor before voice development begins. The key to decrypt  the file will be given to vendor only at commencement of final acceptance testing.A5.  Requirements remain the same. UNOPS has no interest in selecting unusually difficult sentences or words. Its interest is in selecting words which are representative of the day-to-day use of the language. Also, we cannot select sentences for languages which are not yet defined. Q6. Section VII  - Deliverables and OutputsSecond bullet point:  "01 s" we take to mean 0.1 seconds.  Can you, please, confirm?Note that currently Android introduces a sometimes random buffer delay of up to 0.05 seconds.  But 0.1 secs will be achievable.A6: Yes, this is to confirm that 0.1 seconds is correct.Q7.  Ibid.   Third bullet point - "500 words per minute"  Although we can produce speech at that rate,  we are not sure it will be interlligble.  And that is for a language like English.Turkmen, being an aggliutinative language has some rather long words. E.g. :"Okuwçylaryňkymyka?"  – English translation:  "I wonder if it belongs to the school children? "And, to be fair, Chinese and some SE Asian languages have words of typically only one or two syllables.  Would UNOPS consider re-casting this requirement in terms of syllables per minute?A7: The words per minute rather than syllables and the 500 word per minute rate requirements remain the same.