KEITH, Circuit Judge.

Appellants Charles Cornett, James Coldiron and Clarence Robinson, were each charged in a two count indictment with violating, and conspiring to violate, the Racketeer Influenced and Corrupt Organizations Act (RICO), 18 U.S.C. §§ 1962(c) and (d). 1 The indictment alleged that after the State of Kentucky had confiscated alcoholic beverages from illegal “bootlegging” activities, appellants surreptitiously removed the contraband from the state’s custody and resold it for private monetary gain.

The government’s case in the United States District Court for the Eastern District of Kentucky relied heavily upon conversations it had recorded between a government Alcoholic Beverage Control (ABC) agent and appellants. Since substantial portions of the tape recordings were inaudible, the government sought to aid the jury’s understanding of the tapes by providing them with a purported transcript of the recordings. The first trial resulted in a hung jury on April 14, 1980. The second trial which also employed transcripts of the taped conversations, resulted in the conviction of each appellant on October 30, 1980. Appellants Cornett and Robinson were sentenced on each count to four-year concurrent terms of confinement, while appellant Coldiron was sentenced on each count to three-year concurrent terms.

Appellants seek reversal of their convictions on two grounds. First they contend that the district court erred in allowing the government to provide the jury with a “transcript” of tape recordings. Essentially, they allege that the recordings were in many portions so greatly inaudible as to render any attempt at transcription speculative. Secondly, they argue that the district court erred in denying their motions to sever their cases. Although we find no abuse of discretion in the district court’s denial of the motion to sever, we are persuaded that the district court erred to the substantial prejudice of the appellants by permitting the jurors to read purported transcripts of electronically recorded conversations. We reverse for the reasons set forth below.

FACTS

Harlan County Kentucky, with the exception of the city of Cumberland, is a “dry” county 2 which prohibits the sale of alcohol. Those who knowingly sell alcohol in the county’s “dry” territories contrary to local law are “bootleggers.” In an effort to *875 eradicate the bootlegging business, the Harlan County Sheriff’s Office seizes liquor being sold illegally and retains possession of it pending court order. Under Kentucky law, a judge of the state’s district court is empowered to order that contraband alcohol 3 either be destroyed or sold to a licensed liquor store for the remunerative benefit of the Commonwealth of Kentucky. 4

Appellant Cornett is a judge in the Commonwealth’s district court. He is alleged by the government to be a silent partner in Ann’s Liquor Store in nearby non-dry Cumberland, Kentucky. Appellant Robinson was an employee in that store. At trial, ABC agent James Saylor testified for the prosecution that he was approached by appellant Coldiron who proposed a profit-making scheme involving the resale of contraband liquor. Under the devised plan, Judge Cornett would issue court orders empowering agent Saylor to take possession of contraband alcohol being held as evidence in bootlegging cases pending before his court. Armed with the orders, Agent Saylor could transport the contraband to appellant Robinson at Ann’s Liquor Store. Robinson would pay Saylor for the liquor and in turn sell it to the public at a profit. Meanwhile, court records would reflect that the contraband had been destroyed.

After the plan was revealed to him, Agent Saylor contacted the Kentucky State Police and the Federal Bureau of Investigation. Thereafter, all of Saylor’s conversations with appellants were electronically monitored and taped. Approximately twenty-five to thirty tape recordings were made. Prior to trial, a master reel was prepared which contained a composite of all the conversations the prosecution deemed to be relevant. The tape recordings were introduced into evidence. However, because substantial portions of the recordings were either inaudible or unintelligible, the prosecution provided the jury with what purport to be transcripts of the recorded conversations. Although the transcripts were not admitted in evidence, they were provided to the jury during the course of trial while the recordings were being played.

II.

TRANSCRIPTS OF TAPE RECORDINGS

Appellants contend that the prosecution should not have been permitted to provide the jury with transcripts that purport to reflect recorded conversations concededly of poor quality. The government, relying upon cases in this circuit, responds that the decision to employ recordings and transcriptions is a matter entrusted to the sound judicial discretion of the trial court. See, e.g., United States v. Reed, 647 F.2d 678 (6th Cir.1980), cert. denied, 454 U.S. 837, 102 S.Ct. 142, 70 L.Ed.2d 118 (1981); United States v. Vinson, 606 F.2d 149 (6th Cir.1979), ce rt. denied, 444 U.S. 1074, 100 S.Ct. 1020, 62 L.Ed.2d 756 (1980); United States v. Enright, 579 F.2d 980 (6th Cir.1978). The government argues that no abuse of discretion has been demonstrated since the tapes were at least partially audible, and since the transcripts were not themselves admitted into evidence. Finally, the government asserts there was no undue prejudice because appellants were afforded the opportu *876 nity at trial to introduce a “counter transcript” reflecting appellants’ interpretation of the inaudible portions.

It is well settled that the admission of tape recordings at trial rests within the sound discretion of the trial court. United States v. Enright, 579 F.2d at 988. United States v. Cooper, 365 F.2d 246, 250 (6th Cir.1966), cert. denied, 385 U.S. 1030, 87 S.Ct. 760, 17 L.Ed.2d 677 (1967). That discretion presumes, as a prerequisite to admission, that the tapes be authentic, accurate and trustworthy. United States v. Haldeman, 559 F.2d 31 (D.C.Cir.1976), cert. denied sub nom. Mitchell v. United States, 431 U.S. 933, 97 S.Ct. 2641, 53 L.Ed.2d 250 (1977). Moreover, they must be audible and sufficiently comprehensible for the jury to consider the contents. United States v. Bryant, 480 F.2d 785, 789 (2d Cir.1973). Recordings will be deemed inadmissible if the “unintelligible portions are so substantial as to render the recording as a whole untrustworthy.” United States v. Jones, 540 F.2d 465, 470 (10th Cir.1976), cert. denied, 429 U.S. 1101, 97 S.Ct. 1125, 51 L.Ed.2d 551 (1977). Cooper, 365 F.2d at 250.

Although appellants objected at trial to the admission of the frequently inaudible recordings, they have only parenthetically noted that argument on appeal. Rather, the thrust of their appellate challenge is directed to the jury’s use of government prepared transcripts. At various intervals during the proceedings, the prosecution provided the jury with transcripts to read while the recordings were being played aloud.

The use of transcripts is also a matter commended to the trial court’s discretion. United States v. Onori, 535 F.2d 938, 947 (6th Cir.1976). United States v. McMillan, 508 F.2d 101, 105 (8th Cir.), cert. denied, 421 U.S. 916, 95 S.Ct. 1577, 43 L.Ed.2d 782 (1974). This Court has previously stated that the preferred practice is not to submit transcripts to the jury unless the parties stipulate to their accuracy. Vinson, 606 F.2d at 155. See also United States v. Crane, 632 F.2d 663, 664 (6th Cir.1980) (per curiam); United States v. Smith, 537 F.2d 862, 863 (6th Cir.1976) (per curiam). While articulating our preferences, we have not yet established guidelines for the use of transcripts in the absence of a stipulation.

The Honorable Barrington D. Parker 5 , writing for the District of Columbia Circuit in United States v. Slade, 627 F.2d 293 (D.C.Cir.), cert. denied, 449 U.S. 1034, 101 S.Ct. 608, 66 L.Ed.2d 495 (1980), has correctly and comprehensively examined the use of transcripts and has set forth procedures to ensure their accuracy and fairness. He states for the court, “The ideal procedure for testing accuracy is to have the prosecution and defense attorneys stipulate to a transcript.” Id. at 302. If there is a dispute concerning the contents of the tape, “the second best method is for the trial court to make a pretrial determination of accuracy by reading the transcript against the tapes.” Id. The third and least preferred method is to present two transcripts to the jury, one of which contains the government’s version and the other the defense’s version. Id. Hence, the jury becomes the final arbiter of which version is most accurate.

Submitting two versions of a disputed transcript has also been approved by the Second Circuit. In United States v. Carson, 464 F.2d 424 (2d Cir.), cert. denied, 409 U.S. 949, 93 S.Ct. 268, 34 L.Ed.2d 219 (1972), the court held an in camera hearing to determine the accuracy of the transcripts. The parties generally agreed that the transcripts reflected the words on the tapes with certain exceptions. Where there was a dispute regarding particular words, it was agreed that the transcripts would contain the version believed accurate by each party. With respect to the most incriminating tape, the parties agreed to delete certain words and substitute the word “inaudible”. Id. at 437. See United States v. Bryant, 480 F.2d 785 (2d Cir.1973); and United States v. Chiarizio, 525 F.2d 289 (2d Cir.1975).

*877 The Eighth Circuit has taken a slightly different approach to the use of transcripts. That Court has stated that the best evidence of the conversation is the tape itself. McMillan, 508 F.2d at 105. Therefore, a transcript should only be used after the defendant has an opportunity to verify its accuracy and then only to assist the jury as it listens to the tape. Id. When the accuracy of the transcript is disputed, a foundation should be laid by having the person who prepared the transcript testify that he or she has listened to the recordings and accurately transcribed their contents.

It is significant to note that none of the established procedures was effectively employed in the case at bar. The government suggests that the trial judge acknowledged the tapes to be accurate. However, a careful reading of the court’s language reveals that the trial judge merely acknowledged, at the start of the second trial, that he was familiar with the tapes and transcriptions if they were in fact the same ones utilized in the first trial. 6 Moreover, the government places substantial reliance upon the fact that appellants did not avail themselves of the opportunity to introduce a “counter-transcript.” While such a course would certainly permit appellants to deny certain portions of the government prepared transcript, it would correlatively permit the jury to infer admissions by appellants as to the unchallenged portions. More importantly, this Court is of the opinion that several of the tapes are so inaudible that it would be impossible to transcribe them without an independent recollection of the conversations.

Although the enumerated procedures are not exhaustive, each entails basic safeguards to ensure reliability. However, neither the preparation nor use of the transcripts at issue bears any semblance of reliability. Kentucky State Police Sergeant John Barton was one of the custodians of the tapes. Sergeant Barton testified that the Commonwealth’s transcript was prepared by a stenographer who was unconnected with the actual proceedings. There was no apparent disruption of the custodial chain. Independently, the FBI also prepared a transcript of the recordings. According to Sergeant Barton, comparison of the two transcripts revealed “some differences” and “some deletions;” and many places marked “unintelligible” in the Commonwealth’s transcript were somehow transcribed by the FBI. Sergeant Barton characterized the transcript preparation as an “interpreted thing” where a third party listening to the tape might reasonably disagree. Even Agent Saylor, who necessarily was present at each of the actual recordings, could not repeat verbatim while listening to the tapes on the witness stand precisely what was said. Yet, somehow a transcript was prepared.

In fact, the record reveals that a pretrial conference was held for the specific purpose of interpreting and transcribing the inaudible portions of the tape. At least four of the five persons involved in the monitoring and recording of the tapes were present. An assistant United States Attorney who attended the meeting testified “whoever was there and had previously listened or monitored the conversations when they were being recorded, of course, gave their input as far as filling in what was inaudible on the tapes.... ”

The government’s attorney further testified that “a clerk-typist somewhere in Louisville generally makes the original transcripts. And, of course, that clerk-typist would not be familiar with the conversation ... and would not understand the total nature of the conversation. An agent who was present and did overhear it would be in a better position to do so.” Thus, it is clear that even the prosecution realized that portions of the recording were so inaudible that they could not be understood by any *878 one who was not present during the conversations.

The Court also noted that there were discrepancies between the tape and the transcript. The Court commented: “there may be a few words that could be slightly off, or they may have been inaudible when I can hear a word, or it may be inaudible when it looked as if somebody has put down a word.” Additionally, the Court remarked, “I, of course, would be the first to say that the quality of the tapes varies. The ones of the telephone conversations, obviously are the better ones. But the others are not so clear.” Nevertheless, the Court found the transcripts to be reliable.

That the jury relied heavily upon the transcripts is evident from the fact that one juror was observed while the recording was being played to be reading the transcript at length with the available earphones removed. When this reliance is viewed against a backdrop of some alleged 1000 to 1400 transcript spaces labeled “inaudible,” the probable impact of the transcript upon the outcome of the proceeding can be viewed as nothing less than substantial.

We are unpersuaded by the government’s attempts to minimize the impact of the transcripts. The government urges this Court to give deference to the fact that the transcripts are mere aids which were not introduced into evidence. This Court is keenly aware that there is a distinct difference between evidence and an aid used to assist the jury in understanding the evidence. 7 However, the distinction becomes nebulous where, as here, the evidence is unintelligible. The practical effect of using an aid to comprehend unintelligible matter is that the aid becomes the evidence.

The transcripts at issue are problematic despite the trial court’s cautionary instruction to the jury. Prior to playing the first taped conversation, the jury was admonished:

THE COURT: Now, ladies and gentlemen of the jury, before we commence playing the tape, let me advise you that the tape recordings are received into evidence. The transcripts that you have are being furnished for your guidance as you listen to the tapes in clarifying portions of the tapes which may be difficult to hear, or to identify speakers. The tapes, however, are evidence in the case and the transcripts are not evidence. If you perceive any variation, you will be guided solely by the tapes and not the transcripts. Of course, if you don’t want to use the transcripts, you are not required to do that either. If you cannot determine from the tape that particular words are spoken, you must disregard the transcripts insofar as those words are concerned. Now, that applies to all tapes in this case and all transcripts used in this case.

While this is an adequate instruction, its directives are only viable when the tape is clear enough for a juror to detect that the tape is at variance with the transcript. But where, as here, the tapes are partially inaudible, the juror is precluded from making an intelligent comparison. Hence, the likely result is that the transcript becomes the evidence. Shepherding hearsay to the jury via the transcripts, without employing traditional safeguards, can be considered nothing less than prejudicial. The better course would have been for the agents involved to testify regarding their recollection of the conversations so that the jury could be afforded an opportunity to assess their credibility.

We therefore reiterate our preference for using a transcript when the parties stipulate to its accuracy. But in the absence of a stipulation, we hold that the transcriber should verify that he or she has *879 listened to the tape and accurately transcribed its content. The court should also make an independent determination of accuracy by reading the transcript against the tape. Where, as here, there are inaudible portions of the tape, the court should direct the deletion of the unreliable portion of the transcript. This, however, assumes that the court has predetermined that unintelligible portions of the tape do not render the whole recording untrustworthy. Finally, we find submission of two versions of the transcript prejudicial when the tape is significantly inaudible. Such a practice would undoubtedly inspire wholesale speculation by the parties and engender jury confusion. It would be entirely too difficult for the jury to read two separate transcripts while listening to the tape recording. Furthermore, this method is impractical in cases such as this where the defendant has asserted his fifth amendment right to remain silent.

Given the articulated standards, the transcripts utilized at trial bear little semblance of reliability. The transcriber could not have verified their accuracy because the transcripts were the product of a government conference where several agents offered their independent recollection of the taped conversations. We are convinced after listening to the master tape employed at trial that several of the conversations were so inaudible as to preclude transcription. Thus under the facts of this case, we view the district court’s decision to permit juror use of transcripts to be a manifest abuse of discretion.

III.

SEVERANCE

Resolution of the transcript issue renders a lengthy discussion of the severance question unnecessary. Moreover, it is clear that a motion for severance of defendants is committed to the sound discretion of the trial court. United States v. Goldfarb, 643 F.2d 422 (6th Cir.), cert. denied, 454 U.S. 827, 102 S.Ct. 117, 70 L.Ed.2d 101 (1981). A general rule in conspiracy cases is that persons jointly indicted should be tried together. United States v. Dye, 508 F.2d 1226, 1236 (6th Cir.), cert. denied, 420 U.S. 974, 95 S.Ct. 1395, 43 L.Ed.2d 653 (1975). This is particularly true where the offenses charged may be established against all of the defendants by the same evidence. Id. See United States v. Echeles, 352 F.2d 892 (7th Cir.1965). The appellants, therefore, have the burden of showing that they were prejudiced by the court’s denial of the severance motion. United States v. Hamilton, 689 F.2d 1262, 1275 (6th Cir.), cert. denied, — U.S. —, 103 S.Ct. 218, 74 L.Ed.2d 174 (1982); United States v. Tanner, 471 F.2d 128 (7th Cir.), cert. denied, 409 U.S. 949, 93 S.Ct. 269, 34 L.Ed.2d 220 (1972).

In the instant case, appellants claim that they were denied their sixth amendment right to confrontation because they were not able to call their co-conspirators to testify. However, the record clearly shows that two of the co-conspirators declined to testify on the ground of the fifth amendment. Before this Court finds an abuse of discretion, it must at least be demonstrated that one of the co-conspirators was willing to testify. In the absence of any showing of prejudice, this Court finds no abuse of discretion in the trial court’s denial of severance.

CONCLUSION

After reviewing the record, listening to the tapes and comparing them to the transcripts, we are persuaded that the court erred in permitting the use of government prepared transcripts which concededly “interpreted” inaudible portions of tape recordings. Accordingly, the convictions are reversed and the case remanded for further proceedings consistent with this opinion.

Notes

. The pertinent provisions of 18 U.S.C. § 1962 are as follows:

(c) It shall be unlawful for any person employed by or associated with any enterprise engaged in, or in the activities of which affect, interstate or foreign commerce, to conduct or participate, directly or indirectly, in the conduct of such enterprise’s affairs through a pattern of racketeering activity or collection of unlawful debt.

(d) It shall be unlawful for any person to conspire to violate any of the provisions of subsections (a), (b), or (c) of this section.

The indictment alleged that the enterprise through which the appellants affected interstate commerce was “the office of District Judge of the Harlan District Court, Harlan County, Kentucky.” Initially, the defense challenged the legality of referring to a governmental entity as an enterprise within the meaning of the Statute. This issue was ultimately resolved in United States v. Thompson, 685 F.2d 993 (6th Cir.1982) (en banc) rev’g, 669 F.2d 1143. In Thompson, the full court held that the language and plain meaning of the statute did not exclude identification of a government office as a RICO enterprise. 685 F.2d at 1000.

. Ky.Rev.Stat. § 241.010 defines a dry territory as “a county, city district or precinct in which a majority of voters have voted in favor of prohibition ... . ”

. Contraband alcohol is defined as “any alcoholic beverages in the possession of anyone not entitled by law to possess them.” Ky.Rev.Stat. § 244.180(3).

. The disposition of seized alcohol is generally governed by Ky.Rev.Stat. § 242.380 which states:

(1) No property rights other than those in the alcoholic beverage commission pursuant to KRS 244.195(1) shall exist in any alcoholic beverages obtained, possessed, held or used in violation of this chapter.

(2) The court, upon conviction of the person arrested, shall order the alcoholic beverages to be sold by the department of alcoholic beverage control or its representative in accordance with the provisions of KRS 244.-195.

This revised provision went into effect June 17, 1978. The alleged fraudulent orders were issued subsequent to that date and were presumably issued in accordance with this statute. The orders themselves do not cite the statutory authority under which they were issued. See Exhibits 25 and 26. For a general discussion of the precursor to the above statute, see Ky.Op. Atty.Gen. 65-651 (September 1965).

. United States District Judge for the District of Columbia, sitting by designation.

. In response to a defense objection to the transcripts the Court stated:

I have listened to the tapes and read the transcripts, and although the transcripts are not one hundred percent reliable, I don’t find any material error that would affect the outcome of the case.

In light of our review of the tapes and transcripts, we hold this finding to be clearly erroneous.

. It is interesting to note that the Fifth Circuit has rejected the notion that transcripts are mere “aids” to help the jury follow the tapes. In United States v. Onori, 535 F.2d 938, 947 (5th Cir.1976), the Court stated that the use of a transcript as a guide is analogous to the use of expert testimony as a device aiding a jury in understanding other types of real evidence. As such, a transcript is evidence of what is recorded on the audio, and should be authenticated like all other evidence. Id. See also United States v. Sutherland, 656 F.2d 1181, 1200 n. 15 (5th Cir.1981), cert. denied, 455 U.S. 949, 102 S.Ct. 1451, 71 L.Ed.2d 663 (1982).